Effects of Cooperative and Competitive Outcome Dependency on Attention and Impression Preferences

Effects of Cooperative and Competitive OutcomeDependency on Attention and Impression Preferences

Roos Vonk

Leiden University

Received: April 29, 1997; revised: October 23, 1997; accepted: February 17, 1998

These studies examined the effects of outcome dependency on the motive to formaccurate versus desirable impressions, as reflected in attention, liking, and judgments oftask-related social (Exp. 1 and 2) or intellectual (Exp. 3) qualities of the target person. Bothcooperatively and competitively dependent subjects formed more favorable impressionsthan control subjects (Exp. 1), but this effect was due primarily to the expected interactionand not to the dependence (Exp. 2 and 3). Cooperative dependence did produce optimisticjudgments of the target’s abilities compared with other conditions (Exp. 3). Competitivelydependent subjects did not derogate the abilities of the target person; instead, theirjudgments were moderate and relatively unaffected by initial expectancies in all threeexperiments. Further, reading times of these subjects were longer than in the otherconditions. These results suggest that a competitor is perceived as a potential threat,producing caution, careful processing, and conservative judgments.r 1998 Academic Press

People are motivated to form accurate impressions of persons they depend onfor desired outcomes. This motivation produces enhanced attention to behavioralor trait information about such persons. Initially, these effects were demonstratedin studies where subjects received information about a person they expected tomeet or date (e.g., Berscheid et al., 1976). In later studies, Fiske and hercolleagues obtained similar results with various forms of outcome dependence,such as cooperative dependence, where the outcomes of the perceiver and thetarget person depend on their collaborative accomplishment (Erber & Fiske,1984; Neuberg & Fiske, 1987; see also Devine, Sedikides, & Fuhrman, 1989;Srull & Brand, 1983); competitive dependence, where the outcomes of theperceiver are inversely related to those of the target (Ruscher & Fiske, 1990; cf.

Thanks to Ingeborg Mulders and Marjolijn Olde Monnikhof for their assistance in conducting thesestudies, and to Ellen de Bruin, Eric van Dijk, Susan Fiske, and two anonymous reviewers for helpfulcomments on an earlier draft.

Address correspondence and reprint request to: Roos Vonk, Dept. of Social Psychology, Wasse-naarseweg 52, 2333 AK Leiden, The Netherlands. E-mail: [email protected].

JOURNAL OF EXPERIMENTAL SOCIAL PSYCHOLOGY34, 265–288 (1998)ARTICLE NO. JS981350

265

0022-1031/98 $25.00Copyrightr 1998 by Academic Press

All rights of reproduction in any form reserved.

JESP 1350@xyserv3/disk3/CLS_jrnl/GRP_jesp/JOB_jesp34-3/DIV_405a03 rich

No. of Pages—24 First page no.—265 Last page no.—288

Ruscher et al., 1991); and asymmetrical dependence, where a target person hasunilateral power over the perceiver (Fiske, 1993). In these studies, it was alsoshown that subjects individuate the person they depend on and form non-stereotypic impressions. These results support the view that outcome dependencyinduces accuracy motivation, which in turn leads to increased attention andthoughtful processing, especially of stereotype-inconsistent information (forreviews, see Fiske & Depret, 1996; Fiske & Neuberg, 1990).

It is intuitively sensible that people want to have accurate impressions of thosethey depend on, regardless of the nature of their dependence (e.g., cooperative vscompetitive, symmetrical vs asymmetrical): Knowing what the other person islike allows them to predict and control their outcomes (Fiske, 1993). Neverthe-less, outcome dependency may also induce entirely different motives—motivesthat may, in some cases, interfere with accuracy, because they produce prefer-ences towards a particular impression. First, impression preferences aroused byoutcome dependency may concern the likeability of the target person. A robustfinding in the literature is that people are motivated to like persons they expect tointeract with and depend on (e.g., Berscheid et al., 1976; Darley & Berscheid,1967; Tyler & Sears, 1977). It has not yet been examined whether this increasedliking for future interaction partners is similar across different forms of depen-dence. On the one hand, it is possible that this tendency is restricted to cooperativeinteractions, because competition may induce hostility towards the person whocan thwart one’s own goals (Ruscher & Fiske, 1990, p. 832). On the other hand,when competition implies an interaction with the opponent, people presumablywant this interaction to be pleasant and friendly regardless of the outcome, whichcould motivate them to like the opponent just as much as they are motivated tolike their collaborator.

A second impression preference concerns the abilities of the person onedepends on. Klein and Kunda (1992, Study 1) have shown that subjects whoexpect to compete with a person are motivated to develop more unfavorableimpressions of this person’s abilities than subjects who expect to cooperate.Subjects in this study rated a target person’s competence with respect to anintelligence-related task they expected to perform. Subjects who were told that thetarget was going to be their partner formed more favorable impressions of thetarget’s abilities with regard to the task than subjects who expected the target to betheir opponent. This difference converges with the notion of unrealistic optimism(e.g., Weinstein, 1980), according to which people expect to acquire favorableoutcomes and avoid unfavorable outcomes. Thus, when perceivers’ own out-comes can be improved in part by the performance of another person, they maydevelop a more favorable view of this person’s capacities than when theiroutcomes will be damaged by a good performance of the other person.

In sum, it seems that the anticipation of outcome dependency may induce threesorts of motives. First, outcome-dependent subjects want to form accurateimpressions, which results in increased attention to information about the target.Presumably, this tendency does not vary with the type of dependence, because it is

266 ROOS VONK


always functional to have an accurate view of the person who can affect one’soutcomes, regardless whether this person is a partner or a competitor (Ruscher &Fiske, 1990). Second, people want to like the person whom they are going to meetand depend on. This motive, too, may be similar across different forms ofdependence, but it may also be reversed in case of competitive dependence,because of hostility towards the opponent. Finally, outcome dependency maymotivate perceivers to form optimistic impressions of the target’s abilities. Thismotive produces differential impression preferences as a function of the nature ofthe outcome dependence, because it is more desirable to have a positive view ofthe abilities of one’s partner than of one’s opponent.

The assumption of impression preferences, such as described above, does notviolate the assumption that outcome-dependent perceivers are driven by accuracymotives: Outcome dependence may arouse multiple motives simultaneously(Klein & Kunda, 1992). However, in some cases these may conflict with eachother. First, in the case of cooperative dependence, perceivers aim to like theirpartner and form favorable impressions of their partner’s qualities. When theavailable information about the target reflects negative qualities, these prefer-ences run counter to the goal of forming an accurate impression. In the presentstudies, this conflict will be examined by manipulating subjects’ expectancies:Subjects are induced to expect either positive or negative information about thetarget person. Presumably, the accuracy motive leads outcome-dependent subjectsto carefully examine the available evidence (resulting in longer reading times,compared with control subjects) and to base their judgments on this evidence,rather than on their initial expectancy (resulting in a smaller expectancy effectamong outcome-dependent subjects). On the other hand, cooperatively dependentsubjects may also be motivated to hang on to a positive expectancy, and use it toassimilate subsequent target information, whereas they may be more willing toabandon a negative expectancy (cf. Tyler & Sears, 1977). This would result in adifferential expectancy effect among these subjects, depending on whether theexpectancy is positive or negative.

Second, in the case of competitive dependence, a conflict may arise betweenmultiple impression preferences. On the one hand, competitively dependentsubjects may be motivated to like their competitor because they are going tointeract with this person and they do not want the interaction to be hostile. On theother hand, they are also motivated to derogate the abilities of their competitor.1

These motives may run counter to each other when the abilities required for theanticipated task are strongly related to likeability. In earlier studies that have

1 People may also be motivated to ascribe high ability to their opponent because it serves theirself-esteem: If they lose, it can be attributed to the opponent’s exceptional ability, and if they win, itfollows from the augmentation principle that they are exceptionally capable themselves (cf. Goethals& Darley, 1977). However, in the Klein and Kunda studies (1992) as well as the present studies, thisparticular motive was largely eliminated by making the tasks of the subject and the opponentincomparable, so that the opponent’s performance was uninformative about the subject’s ability (seeMethod section).

267EFFECTS OF OUTCOME DEPENDENCY


addressed the effects of competitive dependence on impressions (Klein & Kunda,1992; Ruscher & Fiske, 1990; Ruscher et al., 1991), the task required intellectualor creative abilities that are only weakly related to likeability. In these cases, it ispossible to form a negative impression of the opponent’s abilities withoutnecessarily having to dislike this person. On the other hand, when the requiredabilities are socially related (e.g., social skills, the ability to empathize withothers’ feelings), derogation of these abilities implies that the opponent becomesless likeable, because these skills are strongly related to likeability. This maycreate a conflict between the motive to downplay the abilities of one’s opponentand the motive to like the person one is going to interact with. Note thatcooperatively dependent subjects do not face this dilemma, because they shouldbe motivated to both like their partner and form favorable judgments of theirpartner’s abilities.

The present experiments address these possible conflicts between the differentmotives induced by cooperative and competitive dependence. Subjects receiveinformation about a target person on whom they are either cooperatively depen-dent, competitively dependent, or not dependent. The particular dependencemanipulation used here is similar to that used by Klein and Kunda (1992, Exp. 1;cf. Ruscher et al., 1991, Exp. 2), but the target information and the anticipated taskare related to the social-evaluation dimension of personality (likeability), ratherthan the intellectual-evaluation (competence) dimension (cf. Rosenberg, Nelson,& Vivekananthan, 1968). Specifically, the task anticipated by subjects is such thata person with socially positive, likeable qualities is more likely to perform wellthan a person with socially negative, dislikeable characteristics. This allows anexamination of two possibly rival motives among competitively dependentsubjects.

The information about the target consists of a general introduction, inducingeither a positive or a negative expectancy of the target’s likeability and task-related social abilities, and subsequent specific descriptions, both favorable andunfavorable. The dependent variables are reading times of these descriptions(used as a measure of attention; cf. e.g., Fiske, 1980; Ruscher & Fiske, 1990) andlikeability-related judgments.

The first hypothesis concerns the effects of dependence on reading times. Itfollows from the theoretical arguments of Fiske and her associates that bothcooperative and competitive dependence should produce enhanced attention totrait information about the target person, compared with a control condition inwhich no dependence is expected. This prediction implies a main effect ofDependence on reading times (Hypothesis 1).2

2 Initially, it was also predicted that outcome-dependent subjects would be especially attentive toexpectancy-inconsistent descriptions, compared with non-dependent control subjects, because accu-racy motivation produces additional elaboration on inconsistent information (e.g., Fiske & Neuberg,1990). However, none of the experiments supported this prediction. In retrospect, this may beexplained by the fact that the expectancy was target-based (i.e., it pertained to the target person as anindividual), not category-based, as in the studies by Fiske and her associates (e.g., Erber & Fiske,

268 ROOS VONK


As noted, another way to capture accuracy motivation is by means of the effectof subjects’ expectancy on their judgments of the target (cf. Ruscher & Fiske,1990; Ruscher et al., 1991). Presumably, the easiest way to form a personimpression is to rely on an initial expectancy as an interpretation frame andassimilate subsequent information towards this expectancy (cf. Fiske & Neuberg,1990; Schwarz & Bless, 1992). It may be assumed that judgments of non-dependent subjects are more strongly driven by their initial expectancy thanjudgments of dependent, accuracy-motivated subjects, because the latter shouldbe more willing to expend attentional and cognitive resources by examining andusing each piece of evidence about the target person. Thus, it is predicted thatjudgments of the target person reflect an interaction of Expectancy by Depen-dence, such that the Expectancy effect is weaker in both dependence conditionsthan in the control condition (Hypothesis 2).

An alternative hypothesis follows from the motive to form desired impressions,rather than accurate impressions. As described above, this motive may producedifferential effects of positive and negative expectancies, depending on the natureof the anticipated dependence. Cooperatively dependent subjects who are inducedto form a favorable expectancy of their partner may be motivated to cling to thisexpectancy (and counterargue or assimilate information that is inconsistent withit), because it agrees with their preference towards a favorable impression. On theother hand, these subjects may be more ready to drop a negative expectancy as aninterpretation frame in processing subsequent target information. Reversely,ifcompetitively dependent subjects are motivated to derogate the target (as will beexamined in Hypothesis 4), these subjects may be perfectly happy to stick with anegative expectancy and more willing to abandon an initial positive expectancy.According to this argument, judgments should be affected by an interaction ofDependence by Expectancy, such that the Expectancy effect is stronger when theexpectancy converges with the impression preference aroused by the dependencerelationship toward the target (Hypothesis 3).

Finally, consider the two divergent impression preferences that may be heldamong competitively dependent subjects. On the one hand, deducing from theresults obtained by Klein and Kunda, it follows that these subjects form relativelyunfavorable judgments of the target’s social qualities: It is in these subjects’interest that their opponent does not have the socially desirable qualities required

1984; Neuberg & Fiske, 1987). Information that is inconsistent with target-based expectanciesgenerally evokes longer processing times than stereotype-inconsistent information, even amongnon-dependent perceivers, because inconsistencies within a single person are more compelling thaninconsistencies within a group (e.g., Vonk & Van Knippenberg, 1995; cf. Wyer & Gordon, 1982).There is no theoretical reason to assume that outcome dependency should further increase thisparticular effect, because a target-based expectancy implies that the target person is perceived as anindividual as soon as the expectancy is induced. Thus, whereas additional attention to stereotype-inconsistent information serves the accuracy goal because it produces individuation of the targetperson (by undoing the influence of stereotypic expectancies), this argument does not apply totarget-based expectancies.



for this particular task, so they may be motivated to derogate the opponent. On theother hand, just as cooperatively dependent subjects, subjects in the competitivecondition are expecting to meet their competitor and interact with this person,which should motivate them to like this person. Because of the strong correlationbetween likeability and the socially desirable traits required for successfulperformance, it would be difficult to simultaneously like the competitor and forma negative impression of the competitor’s social qualities, or vice versa. Thus, twocompeting hypotheses can be drawn about the main effect of Dependence onliking. One possibility is that impressions are most favorable in the cooperationcondition and least favorable in the competition condition (Hypothesis 4a),because subjects in the latter condition are motivated to downgrade the oppo-nent’s social qualities required for successful performance. Alternatively, impres-sions in the competition condition may be similar to those in the cooperationcondition, with both being more favorable than in the control condition (Hypoth-esis 4b), because people are motivated to like the person they are going to interactwith and depend on.

EXPERIMENT 1

MethodSubjects and design.Subjects were 93 psychology undergraduates (80 women, 13 men), recruited

during classes. They were paid for participation. A 3 (Dependence)3 2 (Expectancy) factorial designwas used.

Introduction and dependence manipulation.Upon arrival, subjects were asked to read and sign a‘‘confidentiality form,’’ to make it credible that they would later receive personal information aboutanother participant. The form inquired whether subjects had previously participated in a study wherethey had filled out a personality test (all subjects responded correctly that they had not) and if so,whether they had given permission to use their personality data for the present study. Finally, the formexplained that some subjects would receive personal information about another participant who hadfilled out the personality test, and they were asked to sign a statement declaring that they would notdiscuss this information with anyone. All subjects complied with this request.

Subsequently, they were seated in individual cubicles with a computer that paced them through theentire experiment. By pressing the return key, the subject determined the pace of all text on the screen.All subjects were told that they would later go to a different room with three other participants, wherethey would play a question-and-answer game in which two couples would compete against each other.By acquiring game points, a couple could earn a bonus between 5 and 50 Dutch guilders (i.e., between$2.50 and $25.00), depending on how many points they gained. A couple could never receive anybonus if the competing couple had more points than they had.

The anticipated game, as it was announced to subjects, was adapted from Klein and Kunda (1992,Study 1). Each of the two couples playing against each other would consist of a questioner and aresponder. The task of the questioner was to select a psychological problem from a list (e.g.,depression, eating disorder, relation problems) and to present this problem to the responder of thecompeting couple. The task of the responder was to indicate how he or she would respond to thisproblem as if it were a real problem, with the goal of helping the other person as adequately aspossible. A good response would be understanding and would require good social skills from theresponder. Each response was going to be judged by a professional psychologist. Based on thesejudgments, the responding couple could gain or lose points. So, the responder could gain points for hisor her team by responding adequately to the problems. The questioner could not gain points directly.

After the questioner of the first couple had presented a problem and the responder of the secondcouple had responded, it would be the second couple’s turn to present a problem selected by the

270 ROOS VONK


questioner of this couple, and the responder of the first couple would reply. This way, there would beseveral turns.

After this introduction, the computer ostensibly initiated a program that randomly determinedwhich participants were going to enter the game with each other and what their roles during the gamewould be. After a brief pause, four first names appeared on the computer screen, indicating who wouldbe working in the same team and what each person’s task was going to be. In all conditions, thesubject’s own name was attached to the role of questioner. The other three names were female names(because the large majority of subjects were women). In the cooperative dependence condition, thename of the target person, Josey, was mentioned as the partner of the subject, that is, the responder inthe subject’s team. Two other names were listed as the questioner and responder in the competingteam. In the competitive dependence condition, the name Josey was listed as the responder in thecompeting team, and two other names were given for the subject’s partner and Josey’s partner. In thecontrol condition, the name Josey was not mentioned at all. Thus, in the cooperation condition, itwould be beneficial for subjects if Josey was an understanding person with good social skills, becausethat would help Josey to gain points for their team; in the competition condition, it would be harmfulfor subjects if Josey possessed these qualities, because that would increase the chances of thecompeting team acquiring more points than their own team, so that they could not earn any bonus.3

All subjects were then told that they would get to see a series of test results of another participantwho had filled out a personality test a few weeks ago. It was explained that this test was reasonablypredictive of the performance of a responder in the question-and-answer game. At this point, the texton the screen revealed a brief simulated search for the personality files of the two responderparticipants, indicating ‘‘File not found’’ for one of them, and ‘‘Ready’’after searching for the filename‘‘Josey.’’ In the control condition, the computer allegedly could not find either of the responder’s filesand started searching for other personality files. Subjects in this condition were then told that none ofthe participants in their game had filled out the test, and that they, therefore, would see the test resultsof someone else, Josey, who was playing as a responder in a game with three other participants.

In sum, Josey, the participant whose test results all subjects would get to see, was going to besubjects’ partner in the cooperative condition, subjects’ opponent in the competitive condition, andsomeone they had nothing to do with in the control condition. As in the Klein and Kunda study, theentire procedure of the anticipated game, as it was described to subjects, serves several purposes. First,the outcomes of a team are largely dependent upon the responder (i.e., the target person). Second, thetasks of the subject and the target person are entirely different and incomparable. In the cooperationcondition, this reduces the chance that subjects develop competitive feelings towards their partner.Third, in the competition condition, subjects cannot infer their own abilities from the performance oftheir opponent, because the target’s performance is uninformative about the skills required for thesubject’s task. Thus, their self-esteem cannot be served by perceiving the target as particularly capable.

Manipulation of expectancy and target descriptions.The information presented about Joseyconsisted of a general summary of her personality test results and the results from a selection of 20specific test scales. To provide subjects with some latitude in forming preferred impressions anddiscounting undesired information, these materials were preceded by a warning that results ofpersonality tests should be taken with care and reservation, because they reflect a momentaryassessment. For the same reason, all specific test results included qualifiers such as ‘probably’ and‘likely’, rather than definite statements about the target person.

3 As in the studies by Klein and Kunda (1992) and Ruscher et al. (1991), the dependencerelationship toward the target person (cooperative vs competitive) was deliberately confounded withteam membership (ingroup vs outgroup). Since people are generally motivated toward less favorableviews of outgroup members, it was expected that this would increase subjects’ tendency to derogatethe competitor. If this expectancy had been borne out, it would have been necessary to subsequentlyexamine the isolated role of both variables involved in the manipulation. However, as will becomeclear shortly, subjects did not derogate the opponent in spite of their conflicting interests anddifferential group membership.



The general summary was used to manipulate subjects’ expectancy. Depending on condition, itread: ‘‘Generally, Josey’s test results (do not) provide a favorable picture, although there are a fewexceptions on some test scales. Judging the overall pattern of results, it may be expected that Josey will(not) respond adequately to the questions of the psychological-problems-game. She will probablyscore above (below) average.’’

Out of the 20 specific test results that followed, the first 2 were always expectancy-consistent(positive or negative, depending on Expectancy) in order to bolster the expectancy. The remaining 18items consisted of a mixture of 6 positive (e.g., ‘‘Is likely to be interested in other people’’), 6 negative(e.g., ‘‘Is probably intolerant towards other opinions’’), and 6 neutral items (e.g., ‘‘Tends to be orientedtowards the future’’). The items were selected from a pretest in which 40 respondents had judged 42items on a 7-point dislikeable-likeable scale. The neutral items had a mean rating of 4.17. The positiveand negative items were similar in extremity (5.45 vs. 2.57) and word length (mean of 7.67 and 7.17words per item).

Counterbalanced across conditions, the items were presented in two different orders. Because thisvariable did not produce any effects, it was dropped from the analyses.

Dependent variables and manipulation checks.Subjects were urged to concentrate while readingthe test results. For each item, they had to press the return key to see the next item, so that readingtimes could be assessed. After reading the items, a 5-minute filler task was presented, to avoid thatjudgments of the target would be based directly on the evidence just seen. Subjects were asked to writedown all European countries they knew on a sheet of paper. Subsequently, they were asked to givetheir impression of Josey on a general likeability scale (15 dislikeable, 75 likeable) and a series oftraits, including the two traits required for the anticipated task (‘understanding’ and ‘has good socialskills’; 1 5 not at all, 75 highly). It was stressed that subjects’ responses were anonymous and thatneither Josey nor any other participant would get to see them.

Finally, manipulation checks and other checks were assessed. To check for Dependence, subjectswere asked whether they had received information about (a) their partner, (b) a member of thecompeting couple, or (c) someone else. To check whether they had generally understood the procedureof the game, they were asked what their role during the game would be (questioner or responder), whatJosey’s role would be, and which qualities were required for a good performance on the part of theresponder. Subjects were also asked to rate their expectancy of Josey’s performance (15 bad,5 5 good) and of their team’s performance during the game (15 worse than the other team,5 5 better than the other team).

After these questions, subjects were paid and debriefed. It was explained that the experiment wasover and that the question-and-answer game would not be conducted. To avoid frustration about theabsence of the anticipated opportunity to earn additional money, all subjects were given 2.50 guildersin addition to their standard compensation.

Results and Discussion

Manipulation checks.Four subjects (3 men, 1 woman) erred on the manipula-tion check for Dependence (2 were in the competition condition, 2 in the controlcondition). One of them also erred on the check for the role of the target personduring the game. In the analyses described here, these subjects were discarded.(Analyses with and without these subjects produced the same pattern of results.)The remaining 89 subjects responded correctly to all of the checks.

Subjects’ ratings of their expectancies regarding Josey’s performance wereexamined in a 3 (Dependence)3 2 (Expectancy) analysis of variance, whichproduced a strong main effect of Expectancy (F(2, 83)5 72.11, p , .001):Estimates were more favorable in the positive than in the negative expectancycondition (4.51 vs. 3.09). An interaction with Dependence (F(2, 83)5 5.26,p , .01) indicated that this difference was more pronounced in the control

272 ROOS VONK


condition (4.33 vs. 2.46;F(1, 26)5 30.85, eta5 .74, p , .001) and the cooperationcondition (4.81 vs. 3.00;F(1, 29)5 44.87, eta5 .78,p , .001) than in the competi-tion condition (4.36 vs. 3.69;F(1, 28)5 5.63, eta5 .41, p , .05). Thus, estimateswere more moderate in the latter condition. This result will be discussed later.

Subjects’ estimates of their team’s performance produced a Dependence3Expectancy interaction (F(2, 83)5 3.65,p , .05). The main effects were nonsig-nificant (F# 1). In cooperation conditions, expectancies were more favorable incase of a positive than a negative expectancy (3.19 vs. 2.80,p , .05). Incompetition conditions, the reverse applied, although this difference was smallerand nonsignificant (2.93 vs. 3.07). In control conditions, the ratings were the same(both 3.00). These results indicate that subjects adequately inferred the implica-tions of the target’s characteristics and their relationship with the target during thegame.

Because a small minority of subjects were men (10 in the present analyses),subject sex was not included as a variable in the analyses below. (Analyses withand without the male subjects produced the same results.)

Reading times.Before the reading times were collapsed across the six positiveand the six negative items (i.e., the items that were presented in both Expectancyconditions), outliers were examined. Two items from 2 different subjects werediscarded because the reading times indicated that the subject had accidentallyheld the return key for too long, so that the item was skipped. Another 19 itemsfrom 11 different subjects were discarded because the reading time was more than3 standard deviations above the mean, indicating a temporary lapse of attention(cf. Vonk & Van Knippenberg, 1995). Subsequently, the mean reading time of thepositive and the negative items was computed for each subject.

These variables were examined in a Dependence3 Expectancy analysis ofvariance, with Item (positive, negative) as a within-subjects variable. ConfirmingHypothesis 1, a main effect of Dependence (F(2, 83)5 4.65,p , .05) indicatedlonger reading times in the cooperation (3.82 sec. per item) and the competitioncondition (4.08) than in the control condition (3.09). Although reading timestended to be longer in the competition than the cooperation condition, thisdifference was nonsignificant. Both means differed significantly from the controlcondition. No other effects involving Dependence emerged.

The analysis also revealed a main effect of Item (F(1, 83)5 5.41,p , .05),qualified by an interaction with Expectancy (F(1, 83)5 4.25, p , .05). In thepositive-expectancy condition, expectancy-inconsistent (i.e., negative) items werestudied longer than consistent (positive) items (4.10 vs. 3.63). In the negative-expectancy condition, this difference was absent (3.51 vs. 3.47). Thus, subjectspaid more attention to expectancy-inconsistent items (cf. Belmore, 1987), butonly when the expectancy was positive. This may be explained by the ‘double’inconsistency of negative items in this condition: Socially undesirable informa-tion is inconsistent with a favorable target-based expectancy, but it is alsoinconsistent with general expectancies about other people (e.g., Jones & Davis,1965; Fiske, 1980).



Liking and ability judgments.Separate analyses of variance were performed forthe overall likeability rating and the mean rating on the two traits that had beenspecifically mentioned as abilities required for the anticipated task: understandingand socially skilled. Hereafter, these variables will be referred to as likeability andability (i.e., the social abilities required for this particular task). As expected, theywere significantly correlated in all conditions (r 5 .83, r 5 .77, andr 5 .66 forthe cooperation, competition, and control condition). Because the results forlikeability and ability were highly similar, they will be described jointly.

A Dependence3 Expectancy analysis produced a main effect of Expectancyon likeability (F(1, 83)5 27.73, p , .001) and on ability (F(1, 83)5 26.95,p , .001). Judgments were more favorable in positive (5.11 and 5.04) than innegative expectancy conditions (3.95 and 3.77). A main effect of Dependence onlikeability (F(2, 84)5 4.67,p , .05) and on ability (F(2, 83)5 4.30,p , .05)indicated that judgments were more favorable in the cooperative (4.84 and 4.77)and competitive (4.67 and 4.50) than in the control condition (4.07 and 3.93).Both experimental conditions differed significantly from the control conditionand not from each other. This supports Hypothesis 4b, that cooperative andcompetitive dependence both produce relatively favorable impressions. Appar-ently, then, outcome-dependent subjects were motivated to like the target and toascribe socially desirable qualities to this person, even when these qualities werenot to their own benefit, as in the competitive condition.

This implies that Hypothesis 3, according to which the Expectancy effectwould vary as a function of whether the expectancy converges with the impres-sion preference, cannot be tested adequately, because subjects in the two experi-mental conditions did not manifest opposing impression preferences. RegardingHypothesis 2, the Dependence3 Expectancy interaction on likeability judgmentsdid not reach significance (F(2, 83)5 2.47,p , .10). However, a closer look atthe means, presented in the first column of Table 1, indicated that the Expectancyeffect was weaker in the competition condition than in the other conditions.Simple main effects revealed that Expectancy strongly affected judgments in thecooperation condition (F(1, 29)5 23.02, p , .001) and the control condition(F(1, 26)5 12.39,p , .01), but not in the competition condition (F(1, 28)5 1.37,ns). This pattern was more pronounced in the analysis of ability judgments, whichrevealed a significant Dependence3 Expectancy effect (F(2, 83)5 5.61,p , .01).Cell means are presented in the second column of Table 1.

These results provide partial support for Hypothesis 2, which predicted that theExpectancy effect would be reduced in both experimental conditions. Instead, theExpectancy effect was weaker only for competitively dependent subjects. Notethat a parallel finding was obtained on the check for these subjects’ expectanciesabout the target’s performance: The Expectancy manipulation produced a smallereffect on these estimates than in the other conditions. As noted in the Introduction,this reduced Expectancy effect could mean that competitively dependent subjectswere more accuracy-oriented. Indeed, it may be concluded that these subjectsbased their judgments more strongly on the test results they received, which

274 ROOS VONK


altogether consisted of eight expectancy-consistent and six inconsistent items.The substantial effect of Expectancy in the cooperation and control condition isless in agreement with the actual test data presented than the much smallerdifference found in the competition condition. Considered concurrently with thelatter subjects’ tendency to study the items somewhat longer, these results suggestthat subjects who anticipated competition engaged in more thoughtful, accuracy-oriented processing, compared with subjects in the other two conditions, whosejudgments were more strongly affected by the initial expectancy.

In sum, the present results suggest that both competitive and cooperativedependence produce increased attention and more favorable impressions. Inaddition, there is some indication that competitively dependent subjects basedtheir judgments more heavily on the actual target information presented, thusexhibiting a stronger accuracy motive than cooperatively dependent and non-dependent subjects.

EXPERIMENT 2

Experiment 2 was a replication of Experiment 1 involving two improvements.First, given that subjects in the experimental conditions of Experiment 1 paidmore attention to the target and formed more favorable impressions, it remainsunclear whether these effects were produced primarily by the expected interactionor by the expected outcome dependency: In the control condition, subjectsexpected neither of these. Although each interaction implies some degree ofmutual dependency (i.e., people want their interactions to be pleasant, and thisdepends in part on the interaction partner), the outcome dependency introducedhere clearly adds something to the mere anticipation of an interaction: In bothexperimental conditions, subjects know that their chances of earning moneydepend largely on the target person. The increased attention and liking in these

TABLE 1MEANS AND STANDARD DEVIATIONS (BETWEEN BRACKETS) OF LIKEABILITY (1 5 DISLIKEABLE,

7 5 LIKEABLE) AND TASK-RELATED ABILITIES (UNDERSTANDING AND SOCIALLY SKILLED ; 1 5 NOT AT

ALL, 7 5 HIGHLY) AS A FUNCTION OFDEPENDENCE ANDEXPECTANCY IN EXPERIMENTS1 AND 2

Experiment 1 Experiment 2

LikeabilityUnderstanding1socially skilled Likeability

Understanding1socially skilled

CooperationPositive exp. 5.63 (.72) 5.75 (1.07) 5.23 (.96) 5.30 (.98)Negative exp. 4.00 (1.13) 3.73 (1.39) 3.78 (1.05) 3.62 (.98)

CompetitionPositive exp. 4.93 (.92) 4.57 (1.16) 4.58 (.67) 4.92 (.76)Negative exp. 4.44 (1.31) 4.44 (1.32) 4.00 (1.00) 4.33 (1.08)

ControlPositive exp. 4.73 (.70) 4.73 (1.10) 5.28 (.95) 5.41 (.92)Negative exp. 3.31 (1.38) 3.00 (.91) 3.28 (.96) 3.21 (1.03)



conditions could either be due to this dependence or to the anticipated interaction.In Experiment 2, therefore, the control condition was altered, such that subjects inthis condition did expect to interact with the target person, without beingdependent on this person. Assuming that outcome dependency is the crucialvariable in the differential attention and impressions obtained in Study 1, it waspredicted that the main effects of Dependence would be replicated in Study 2.

A second alteration in Experiment 2 concerns the ratio of expectancy-inconsistent to consistent target descriptions. In Experiment 1, this ratio was 6 to8, that is, subjects received almost as many inconsistent as consistent items. Thisimplies that, if judgments are based on the target information, it would be accurate(i.e., in agreement with the information presented) to judge the target quitemoderately in both Expectancy conditions, producing a small difference betweenthese two conditions. Thus, the moderate judgments in the competition conditionmay be interpreted as a sign of accuracy-oriented processing.

Alternatively, moderate judgments may be an indication of uncertainty orcaution. Thus, it is possible that competitively dependent subjects were merelymore cautious and less willing to arrive at closure in their inferences about thetarget. Possibly, these subjects wanted to remain prepared for any strengths orweaknesses their opponent might later reveal, regardless of the information theyreceived. This would imply that their judgments remain moderate even when thelarge majority of target descriptions is expectancy-consistent and, hence, clearlypoints to one direction of the underlying trait.

To examine this possibility, the number of expectancy-inconsistent items wasreduced in Experiment 2. If competitively dependent subjects are more conserva-tive because they simply want to be prepared for anything, their judgments shouldremain moderate even when the majority of items is expectancy-consistent;consequently, the Dependence3 Expectancy effect should be replicated inExperiment 2. On the other hand, if competitively dependent subjects base theirimpressions primarily on the target descriptions, a smaller proportion of inconsis-tent items should increase the Expectancy effect on their judgments, in compari-son with Experiment 1. In this case, a combined analysis on the competitionconditions of Experiments 1 and 2 should produce an Expectancy3 Experimentinteraction, such that the Expectancy effect is stronger in Experiment 2, reflectingthese subjects’ sensitivity to differences in the number of expectancy-inconsistentitems.

MethodSubjects were 90 students (34 men, 56 women) from different faculties, mostly law and medicine.

The design and procedure were the same as in Experiment 1, except for two changes. First, subjectswere told that after the question-and-answer game, they would discuss and evaluate the game with aparticipant who was going to play the game at the same time as the subject, but in another room withthree other participants. So, in addition to the subject’s partner and the two participants in thecompeting couple, another participant was introduced, one that subjects were going to talk with afterthe game. In the control (i.e., anticipated interaction) condition, this participant’s name was Josey, thetarget person. It was stressed that the conversation with Josey would take place after the game pointshad been assigned to everyone and that it could not affect the subject’s chances of earning a bonus. So,

276 ROOS VONK


subjects in the control condition expected an interaction with the target person, but their outcomes didnot depend on this interaction in any way.

A second change was that each subject received 8 expectancy-consistent, 4 expectancy-inconsistent, and 4 neutral descriptions of Josey. In the positive-expectancy condition, subjects readthe same 8 positive items as in Experiment 1; in the negative-expectancy condition, they read the same8 negative items as in Experiment 1. The 4 expectancy-inconsistent items were selected such that wordlength and extremity were the same as for the consistent items. As in Experiment 1, the first 2 itemspresented were always expectancy-consistent. Presentation order of the other 14 items was random-ized for each subject.

Dependent variables and manipulation checks were the same as in Experiment 1. Before thequestion about subjects’ expectancy of their team’s performance and of Josey’s performance, theywere also asked to indicate how they expected themselves to perform.


Manipulation checks.One subject in the competitive condition was discardedbecause he erred on the Dependence check. The other 89 responded correctly onall of the checks.

Subjects’ expectancies of the target’s performance, their own performance, andtheir team’s performance were analyzed in separate 3 (Dependence)3 2 (Expec-tancy) analyses of variance. Estimates of subjects’ own performance (15 bad,5 5 good) were not affected by the manipulations (allF , 1; grand mean5 3.34,SD5 .67). Estimates of the target’s performance were higher in positive than innegative Expectancy conditions (4.36 vs. 2.82; Expectancy main effectF(1,83) 5 90.25,p , .001). Although this difference was again somewhat smaller inthe competition condition (4.38 vs. 3.00) than in the cooperation condition (4.53vs. 2.81), the interaction with Dependence was nonsignificant in this experiment(F , 1).

As in Experiment 1, subjects’ estimates of their team’s performance revealed aDependence3 Expectancy interaction (F(2, 83)5 5.12,p , .01). In the coopera-tion condition, subjects with a positive expectancy were more optimistic that theywould beat the competing team than subjects with a negative expectancy (3.53 vs.3.00, p , .05). In the competition condition, the reverse applied; again, thisdifference was smaller and nonsignificant (2.92 vs. 3.07). In the control condition,the means did not differ (3.06 vs. 3.00).

Initially, subject sex was included in the analyses, but because it did not affectany of the results, it was dropped from the analyses reported here.

Reading times.The reading time analysis was based on the 4 positive and 4negative items that were presented in all conditions. One reading time wasdiscarded because the item was accidentally skipped. Eleven reading times from 8other subjects were discarded because they were more than 3 standard deviationsabove the mean for that item.

A Dependence3 Expectancy analysis of variance, with Item (positive, nega-tive) as a within-subjects variable, produced two significant effects. A main effectof Dependence (F(2, 83)5 8.30, p , .001) was obtained, but the pattern ofsignificant differences was not the same as in Experiment 1. Reading times in thecompetition condition (4.29 sec. per item) were significantly longer than in both



the cooperation (3.10) and control condition (3.32), which did not differ fromeach other. Note that, in comparison with Experiment 1, reading times in thecontrol condition were somewhat longer, which is most likely due to theanticipated interaction introduced in this condition. As a result, the differencebetween this condition and the cooperation condition disappeared. Note also that,in Experiment 1, there was a nonsignificant tendency for reading times to belonger in the competition than the cooperation condition. This effect was morepronounced in the present experiment. Thus, whereas the effects of cooperativedependence do not add to the effects of mere anticipated interaction, the effects ofcompetitive dependence do. These results suggest that cooperative and competi-tive dependence may not have the same effects on attention, as has been argued byRuscher and Fiske (1990). This issue will be considered further in the GeneralDiscussion.

A substantial Expectancy3 Item interaction (F(1, 83)5 25.96, p , .001)indicated longer reading times for expectancy-inconsistent than expectancy-consistent items. This difference occurred in the positive-expectancy condition(4.15 for negative items vs. 3.41 for positive items;F(1,41)5 11.24,p , .01) aswell as in the negative-expectancy condition (3.06 for negative items vs. 3.60 forpositive items;F(1, 42)5 20.57,p , .001). Thus, subjects paid more attention toexpectancy-inconsistent items overall.

Likeability and ability judgments.A main effect of Expectancy on likeability(F(1, 83)5 44.56,p , .001) and ability (F(1, 83)5 52.39,p , .001) reflectedmore favorable judgments in positive than in negative expectancy conditions(5.06 vs. 3.70 for likeability; 5.23 vs. 3.73 for ability). Differing from Experiment1, the Dependence main effect was nonsignificant (F , 1 for both dependentvariables). This suggests that the anticipated interaction in the control conditionproduced the same tendency towards a favorable impression (m 5 4.35 and 4.38)as the anticipated outcome dependency in the cooperation (4.48 and 4.44) andcompetition (4.27 and 4.61) condition.

As in Experiment 1, the effects of Expectancy differed across Dependenceconditions. The interaction was significant for both likeability (F(1, 83)5 4.11,p , .05) and ability (F(1, 83)5 5.14,p , .01). Cell means are presented in thethird and fourth column of Table 1. Again, judgments in the competition conditionwere relatively moderate, and the Expectancy effect did not reach significance(F(1, 26)5 3.10 for likeability; F(1, 26)5 2.71 for ability) whereas it wassubstantial in the cooperation (F(1, 29)5 16.09,p , .001 andF(1, 29)5 22.78,p , .001) and control condition (F(1, 28)5 32.85,p , .001 andF(1, 28)5 37.95,p , .001).

Moreover, a combined analysis on the dependence conditions of Experiment 1(six inconsistent items) versus Experiment 2 (four inconsistent items) producedno effects of Experiment whatsoever (all effects involving Experiment wereF , 1 for both dependent variables), and merely reproduced the Dependence(cooperative, competitive) by Expectancy interactions (F(1, 112)5 5.97,p , .05for likeability; F(1, 112)5 12.71, p , .001 for ability). Thus, subjects were

278 ROOS VONK


moderate in their judgments of their opponent, regardless of the ratio of inconsis-tent to consistent descriptions.

These results suggest that competitively dependent subjects were not moresensitive to the target information than subjects who judged their partner; if theywere, their judgments would have been affected by variations in the number ofinconsistent items presented. Instead, the data show that their judgments remainedmoderate even when the large majority of items pointed into the same evaluativedirection. Note also that in both experiments, this same conservatism wasmanifested in these subjects’ estimates of whether their team would performbetter than the competing team. Combined with the longer reading times in thecompetition condition, these results provide a picture of the competition-anticipating subject as a thoughtful, careful perceiver, who is not at all ready tojump to conclusions about the opponent and, instead, wants to be prepared foranything.4

An alternative interpretation of these results follows from the two competingmotives described in the introduction. Given the socially desirable qualitiesrequired for successful performance in the anticipated game, subjects in thecompetition condition may face a dilemma. On the one hand, they may bemotivated to downgrade the social qualities of their opponent, because they wantto win the game. On the other hand, they are expecting an interaction with thisperson, which is a powerful incentive to like the opponent. If these two rivalmotives are aroused simultaneously, it could lead subjects to become hesitant andto form judgments that meet both motives to some extent (with somewhat moreweight exerted by the motivation towards a favorable impression, as the datashow).

This possibility was examined in Experiment 3, by altering the nature of theanticipated game, such that the abilities required for successful performance bythe target person are less strongly related to likeability (as in earlier studies onoutcome dependency). Subjects were told that the responder in the question-and-answer game would have to be intelligent and knowledgeable in order to respondadequately and gain points. This would allow competitively dependent subjects toderogate the intelligence-related abilities of the opponent, without having todislike her and, hence, without having to expect an unpleasant interaction. Thus, itwas predicted that ability judgments would be most favorable in the cooperation

4 These results appear to deviate from those reported by Ruscher et al. (1991, Exp. 2). In thisexperiment, all subjects received information about their partner as well as both opponents (i.e., thecomparison was within-subjects), so they had to distribute their attentional resources across thesedifferent persons. Reading time analyses suggested that subjects assigned higher priority to forming anaccurate impression of their partner than of their opponents. It is possible that, when confronted withmultiple targets, imposing restraints on cognitive resources, subjects prefer to focus on forming anaccurate impression of their partner. Ruscher et al.’s analysis of think-aloud protocols did indicate that,insofar as subjects considered expectancy-inconsistent information about their opponents, thisinformation evoked more elaborations than did expectancy-inconsistent information about theirpartner.



condition and least favorable in the competition condition; likeability judgments,on the other hand, should be favorable in both conditions.

EXPERIMENT 3Method

Subjects were 89 students (62 women, 27 men) from different faculties, mostly law, medicine, andeducation. The design and procedure were the same as in Experiment 2, except that the anticipatedgame was described as a task in which the questioners would select intelligence and generalknowledge questions; thus, in order to respond adequately, the responders would have to be intelligentand possess good general knowledge. The test taken by the target person was described as anintelligence test that was reasonably predictive of the performance of responders in this game. The testresults presented were intelligence-related. As in Experiment 2, subjects received 8 expectancy-consistent, 4 expectancy-inconsistent, and 4 neutral items. These items were selected from a pretest inwhich they had been rated by 50 respondents on a 7-point scale (15 unintelligent, not knowledgeable,7 5 intelligent, knowledgeable). The selected positive items (e.g., ‘‘Appears to be well-informedabout what goes on in the world’’) and negative items (e.g., ‘‘Seems to find it difficult to rememberthings’’) were similar in extremity (5.52 and 2.46) and word length (6.17 and 5.67 words per item).The neutral items were the same as in Experiment 2, after having established from the pretest resultsthat these items were also neutral with respect to intelligence/knowledgeability.

After the filler task, subjects were asked to rate the target person on likeability (15 dislikeable,7 5 likeable) and ability with regard to the anticipated task (15 unintelligent, not knowledgeable,7 5 intelligent, knowledgeable). The order of these two questions was counterbalanced acrossconditions. Because this did not affect the results, question order was dropped from the analysesdescribed below.


Manipulation checks.Three subjects (2 in the cooperation condition, 1 in thecontrol condition) were discarded because they responded incorrectly to thedependence manipulation check. One subject in the competition condition wasdropped because he did not know what roles he and the target person had beenassigned to. The remaining 85 subjects (60 women, 25 men) responded correctlyon all the checks.

As in Experiment 2, subjects’ expectancies of their own performance were notaffected by the manipulations (F, 1; grand mean5 3.48 on a 5-point scale,SD5 .72), whereas expectancies of the target’s performance were more favorablein positive than in negative Expectancy conditions (4.22 vs. 2.80;F(1, 79)5 60.54,p , .001). This effect was not qualified by Dependence (F , 1). Expectancies ofthe performance of the subject’s team in comparison with the other team reflecteda Dependence3 Expectancy effect (F(2, 79)5 3.26,p , .05). In the cooperationcondition, estimates of subjects with a positive expectancy were more favorablethan of subjects with a negative expectancy (3.27 vs. 2.80,p , .05). In both thecompetition and control condition, this difference was absent (3.33 vs. 3.23, and3.00 vs. 3.25). Thus, even more clearly than in the earlier experiments, theseresults appear to reflect the cautiousness of competitively dependent subjects.

Subject sex was excluded from the analyses after having established that it didnot affect the results.

280 ROOS VONK


Reading times.One reading time was discarded because the item was acciden-tally skipped. Nine reading times from seven different subjects were more than 3standard deviations above the mean for that item and were discarded as well.

A Dependence3 Expectancy analysis of variance, with reading times acrossthe four positive versus four negative items as a within-subjects variable,produced similar results as in Experiment 2. A main effect of Dependence (F(2,79) 5 6.55,p , .01) reflected longer reading times in the competition condition(3.58) than in the cooperation (2.84) and control (2.55) condition. A main effect ofItem (F(1, 79)5 8.90, p , .01) and an Expectancy3 Item interaction (F(1,79) 5 14.84,p , .001) indicated that negative items were studied longer thanpositive ones in the positive-expectancy condition (3.32 vs. 2.72); the reverseapplied in the negative-expectancy condition, but this difference was smaller andnonsignificant (2.86 vs. 2.95). Thus, as in Experiment 1, higher attention forexpectancy-inconsistent items was found only when the expectancy was positive.These results are remarkable, because the inordinate weight of negative informa-tion is known to be restricted to the morality and social desirability domain; in theability realm, positive information tends to carry more weight than negativeinformation (e.g., Reeder & Brewer, 1979). Nevertheless, the results indicate thatin the present study, information reflecting low ability drew more attention,especially when it was expectancy-inconsistent. This issue will be addressedfurther in the General Discussion.

Likeability and ability judgments.Likeability judgments were not affectedeither by Expectancy or Dependence (allF , 1). The absence of an Expectancyeffect indicates that, as intended, subjects did not generalize the ability-relatedinformation about the target towards likeability (indeed, the correlation betweenlikeability and ability ratings was .05). In all conditions, the target was perceivedas moderately likeable (5.03 for cooperation, 4.80 for competition, and 4.93 formere anticipated interaction). Note that these means are somewhat higher thanthose in Experiment 2. Possibly, subjects’preference for a favorable impression ofthe person they were going to meet was given more latitude in the presentexperiment because they did not receive any likeability-related information and,hence, no information indicating that the target might be dislikeable.

Analysis of variance on the ability judgments produced a main effect ofExpectancy (F(1, 79)5 27.68, p , .001), a main effect of Dependence (F(2,79) 5 3.54,p , .05), and a marginal interaction (F(2, 79)5 2.96,p , .06). Thetarget’s intelligence and knowledgeability was rated as higher in the positive thanin the negative expectancy condition (4.48 vs. 3.23). Judgments were also morefavorable in the cooperation condition (4.28) than in the competition (3.50) andcontrol condition (3.80). Thus, in this experiment competitively dependentsubjects did form less favorable judgments of the task-related abilities of theiropponent, but their ratings remained close to the neutral scale point and did notdiffer significantly from the control group. The moderate judgments of thesesubjects were again reflected in the interaction effect. In the competition condi-tion, the effect of Expectancy was small and nonsignificant (3.86 vs. 3.13;F(1,



28) 5 3.39, ns), whereas it was more substantial in the cooperation (5.42 vs. 3.23;F(1, 23)5 25.19, p , .001) and the control condition (4.36 vs. 3.31;F(1,28) 5 5.11,p , .05).

SUMMARY AND GENERAL DISCUSSION

Different types of dependence may arouse different motives. People may bedependent on others’ evaluations of them (e.g., in job applications), in which casethey want to be judged fairly and be seen as competent. People may depend onothers’ abilities and behaviors (e.g., when using a lawyer to represent one’sinterests), in which case they want the other person to be competent. People mayhave to work with others to solve a collaborative problem, in which case theywant these others to be friendly and cooperative in addition to being competent; orthey may compete with others, which arouses multiple motives that may runcounter to each other (e.g., the motive to have a friendly interaction and a sportingmatch, to downplay the opponent’s abilities so as to be optimistic about winning,or to inflate the opponent’s abilities so as to pre-emptively protect one’s self-esteem). All of these motives affect person perception in multiple ways. Forinstance, the motive to predict and control the person one depends on producesaccuracy motivation, which in turn affects stereotyping (e.g., Fiske, 1993) andother cognitive shortcomings, such as correspondence bias (Vonk, in press); themotive to get along with one’s collaborator and to beat one’s opponent affects thesalience of affiliation-related and dominance-related traits (Battistich & Aronoff,1985). Thus, dependence-based motives may impinge on different aspects ofimpression formation and trait inference.

In sum, there are many different types of dependence; each type may inducemultiple motives; and each of these motives may affect multiple aspects of personperception. At present, we lack a systematic analysis of the many relationshipsand interactions that are conceivable within this complex conglomerate. Thepresent results provide some pieces of the puzzle that we will eventually be able tosolve, when more pieces have been gathered. Specifically, these studies yield ananalysis of the different motives that are induced by cooperative and competitivedependence, the possible conflicts between these motives (i.e., the conflictbetween accuracy orientation and impression preferences, and the conflict be-tween multiple impression preferences), and three components of person percep-tion that are affected by these motives, namely, attention, liking, and judgments ofoutcome-related abilities.

Regarding the effects of attention, Experiment 1 showed that both cooperativeand competitive dependence produce enhanced attention to trait informationabout the target. However, the results from Experiments 2 and 3, in which thecontrol condition involved an anticipated interaction, showed that this effect ispartially due to the expected interaction that accompanies outcome dependency inmost situations: In these experiments, the differential attention between thecooperation and the control condition disappeared. Competitive dependence, onthe other hand, seems to add something above and beyond the effects of mere

282 ROOS VONK


anticipated interaction: Subjects spent more time processing information abouttheir opponent than about their collaborator or interaction partner.

The other variables assessed in these experiments suggest that this additionalprocessing is not aimed at forming a negative impression of the opponent, asmight have been expected from Klein and Kunda’s (1992) results. In all threeexperiments, likeability judgments of the opponent were similar to judgments ofthe partner. In Experiment 1, both were more favorable than judgments in thecontrol condition. Experiments 2 and 3, in which control subjects expected aninteraction with the target, indicated that this increased liking results primarilyfrom anticipated interaction, rather than outcome dependency per se.

In part, this motivation to like the interaction partner was responsible for thefavorable judgments given by competitively dependent subjects in Experiments 1and 2. In these experiments, the qualities required for successful performancewere strongly associated with likeability, in order to examine the role of twopossible rival motives among competitively dependent subjects. On the one hand,they may be motivated to derogate the qualities of the competitor, so that they canexpect favorable outcomes for themselves; on the other hand, they are motivatedto like the person they are going to interact with. The results showed that the lattermotive predominated. However, the relatively conservative judgments and longreading times among these subjects suggested that they may have been hesitantand reluctant to draw definite conclusions about the target, possibly because theysomehow felt that liking the target would interfere with their motive of beingoptimistic about winning the game.

In Experiment 3, these two motives were disentangled by introducing theanticipated game as a task that required intellectual, rather than social abilities.This would enable competitively dependent subjects to downplay the target’sintellectual skills, without having to dislike the person they were going to meet. Inpart, this was indeed what they did: Subjects in the competition condition likedthe target person as much as subjects in the cooperation condition, yet providedless favorable judgments of her intellectual abilities. However, this difference wascaused primarily by the cooperation condition, in which subjects tended tooverrate their partner’s abilities; in the competition condition, ability judgmentswere similar to those in the control condition. Moreover, judgments of competi-tively dependent subjects remained relatively unaffected by their expectancy,even though the large majority of the target items confirmed the expectancy(unlike Experiment 1), and even though subjects had every opportunity toderogate the target’s abilities without necessarily having to dislike her (unlikeExperiments 1 and 2).

Altogether, these results suggest that, contrary to expectations, competitivedependence may not have the same effects as cooperative dependence. Both havein common that one expects an interaction, which produces increased liking andattention, but that is where the similarity ends. An important difference betweenthe two is probably that a competitor is perceived as more of a potential threatthan a collaborator. Note that this difference is not ‘logical’: If subjects in the



present studies were to assume that the target person would respond incorrectly tothe questions, they would perceive their partner as someone who could damagetheir outcomes (i.e., a potential threat), and their opponent as someone whoseinability would be beneficial for them. However, given the moderately positiveview that people have of others in general (see e.g. Skowronski & Carlston,1987), subjects were probably more strongly focused towards the possibility thatthe target would perform reasonably well. This was also evidenced by the grandmean of subjects’ expectancies of the target’s performance (on a 5-point scale,3.81 in Exp. 1; 3.58 in Exp. 2; and 3.50 in Exp. 3).

Given this moderately positive expectancy, it follows that the opponent targetwas perceived as more threatening than the collaborator target. In addition, themere label of ‘competitor’ probably activates associations with threat, whereasthe label of ‘partner’ does not. Thus, the opponent target may have been perceivedas potentially more harmful than the collaborator target. This would explain whycompetitively dependent subjects engaged in longer processing and arrived atmore conservative judgments, since threatening stimuli tend to evoke attention,caution, and other responses that aid in controlling or minimizing the possibledamage (e.g., Peeters & Czapinski, 1990; Pratto & John, 1991; Taylor, 1991).Cooperatively dependent subjects, on the other hand, seemed more careless intheir willingness to assume that their partner would be skilled with respect to theupcoming task.

In sum, it is possible that competitive dependence induces a state of caution andconservatism that serves to prepare for a possible threat. Note that this stateproduces more attention to all information about the target person, both positiveand negative. This makes intuitive sense, because the aim of controlling a possiblenegative event requires that one knows both the strengths and the weaknesses ofthe opponent.

An alternative interpretation of the present results is that subjects in the twodependence conditions held differential expectancies about their degree of controlover the target’s behavior. Specifically, competitively dependent subject may haveinferred that they would be able to reduce the opponent’s performance byselecting difficult questions for the target. Cooperatively dependent subjects, onthe other hand, may have inferred that they would not be able to affect theirpartner’s performance. If this is true, subjects in the competitive conditionbelieved that they had more control over the outcomes than those in thecooperation condition. This would imply that subjects in the latter condition usedthe information about the target only to predict their outcomes, whereas subjectsin the competition condition used it in the service of both prediction and control.

Although this possibility cannot be ruled out conclusively, it is not corroboratedby subjects’ estimates of their own performance or their team’s performance. Ifsubjects felt the information about their opponent was useful to improve their ownperformance during the task, one would expect increased performance estimatesin competitive, compared with cooperative, conditions. However, subjects’ esti-mates of their own performance (as assessed in Experiments 2 and 3) were not

284 ROOS VONK


affected by the manipulations at all. Further, estimates of their team’s perfor-mance were quite moderate and conservative in the competitive conditions,showing no signs of the optimism that would seem to result from gaining a senseof control. In fact, it seems likely that subjects in all conditions did not expect tohave a lot of control over their outcomes, because the role assignment was suchthat the outcomes depended on the target much more than on the subject, andsubjects’ behavior during the task was pretty much fixed by their role ofquestioner and the pre-designed questions.

This may explain why the effects of cooperative dependence were smaller inthe present studies than in studies by Fiske and her colleagues (e.g., Neuberg &Fiske, 1987), in which subjects interact with their partner to jointly derive asolution to a problem. Possibly, subjects’ motive in this setting is primarily tocontrol their outcomes (e.g., by stimulating their partner, or by compensating theirpartner’s weaknesses), whereas their primary motive in the present studies mayhave been prediction: Early studies suggest that people are motivated not only tocontrol their outcomes, but also to predict them even in the absence of anypossibility of control. Predictability and control tend to be confounded, but theability to regulate aversive stimuli is beneficial only when it is accompanied byreduction of uncertainty regarding the threatening event (see e.g., Averill, 1973;Schulz, 1976; cf. Pervin, 1963). Because predictability appears to be especiallybeneficial in case of threat, the prediction motive in the present studies may havebeen stronger in competition than in cooperation conditions.

The distinction between prediction and control is clearly important and shouldbe addressed more systematically in future research. More generally, currentresearch on the effects of outcome dependence lacks a systematic analysis of theeffects of different types of dependence. For instance, results obtained by Stevensand Fiske (1995, described in Fiske & Depret, 1996) indicate that subjects whodepend on someone’s evaluation of them, and are unable to influence the otherperson, are more concerned with self-protection than with forming an accurateimpression of the evaluator. This result may be attributed to the absence ofcontrol, but also to the nature of the dependence. In this context, it is useful todistinguish evaluative from behavioral dependence (Vorauer & Ross, 1993). Incase of evaluative dependence, people depend on the other’s evaluation orimpression of them, for instance, in the way that job applicants depend onselection officers. Thus, their self-concept is at stake. Behavioral dependenceemerges when one’s outcomes are affected by the other’s behavior, for instance, inthe way that ill people depend on their physician (in case of asymmetricbehavioral dependence) and sports people depend on the performance of theirteam mates and their opponents (in case of symmetric behavioral dependence).Cross-cutting this distinction, the dependent person can be either able or unable toinfluence the target and, thereby, control the outcomes. In future research onoutcome dependency, it will be crucial to distinguish not only cooperative versuscompetitive behavioral dependence, but also to systematically disentangle behav-ioral versus evaluative dependence, and predictability versus control.



A final point of consideration concerns the present results regarding attention topositive versus negative target information. None of the experiments providedindication that positive and negative items are processed differently as a functionof dependence. Subjects generally paid more attention to negative information,especially when it was expectancy-inconsistent. This may be explained by themoderately positive view that people have of others in general, which rendersnegative information unexpected to begin with. Interestingly, the pattern of resultsfor the ability-related information presented in Experiment 3 was highly similar tothe pattern obtained in the other experiments. In research on positive-negativeasymmetries, it has been established that socially negative information carriesmore weight in evaluative judgments than socially desirable information, whereasthe reverse applies to ability-related information (e.g., Reeder & Brewer, 1979;Skowronski & Carlston, 1987; Vonk, 1995). It has also been found that sociallynegative information evokes more attention and is studied longer than sociallypositive information (e.g., Fiske, 1980), so here too, one might have expected areversed pattern regarding attention to ability-related information. The presentresults, however, indicate that when it comes to attention, the asymmetries forlikeability-related and ability-related information parallel each other.

A possible explanation is that subjects in these studies received trait informa-tion, not behavioral information. An important cause of the positivity effectsgenerally found in the ability domain is that competent behavior is perceived asmore informative about the underlying ability than behavior reflecting incompe-tence, because persons cannot perform better than their abilities allow them to(e.g., Reeder & Brewer, 1979). In the present study, however, the target descrip-tions presented to subjects already implied general trait ascriptions (e.g., ‘‘Finds itdifficult to remember things,’’ rather than ‘‘Could not remember the name of thevice-president’’). Thus, there was no reason for subjects to assume that thenegative ability information was less informative. In this case, it seems likely thatthe moderately positive expectancy of others led subjects to perceive negativeability-related traits as relatively novel and unexpected.

In addition, it seems not coincidental that subjects generally spent more timeprocessing (a) negative information (regardless of whether it was likeability-related or ability-related), and (b) information about their opponent: Both can beregarded as threatening. In everyday life, negative traits are generally morethreatening than positive traits. People can be damaged by others’ hostility orinsensitivity, but also by others’ stupidity and ignorance. As a consequence,perceivers may be automatically vigilant with respect to such potentially harmfultraits (Pratto & John, 1991), even in situations where these traits could actuallybenefit them (as in the competition conditions of the present experiments). Ifpeople are ‘‘wired’’ to be wary of threatening stimuli, it makes sense that thisresponse is not affected by more thoughtful considerations of one’s possiblebenefits and losses (Vonk, 1996). Viewed in this way, higher attention for negativetraits may be interpreted as a cautious, careful response to potential threat (cf.Peeters & Czapinski; Taylor, 1991). Reversely, higher attention for one’s oppo-

286 ROOS VONK


nent can be interpreted as a negativity effect, one that may serve the same functionas trait-related negativity effects, namely, the goal of preparing for negativeoutcomes.

REFERENCESAverill, J. R. (1973). Personal control over aversive stimuli and its relationship to stress.Psychological

Bulletin,80,286–303.Belmore, S. M. (1987). Determinants of attention during impression formation.Journal of Experimen-

tal Psychology: Learning, Memory, and Cognition,13,480–489.Berscheid, E., Graziano, W., Monson, T., & Dermer, M. (1976). Outcome dependency: Attention,

attribution, and attraction.Journal of Personality and Social Psychology,34,978–989.Battistich, V. A., & Aronoff, J. (1985). Perceiver, target, and situational influences on social cognition:

An interactional analysis.Journal of Personality and Social Psychology,49,788–798.Devine, P. G., Sedikides, C., & Fuhrman, R. W. (1989). Goals in social information processing: The

case of anticipated interaction.Journal of Personality and Social Psychology,56,680–690.Erber, R., & Fiske, S. T. (1984). Outcome dependency and attention to inconsistent information about

others.Journal of Personality and Social Psychology,47,709–726.Fiske, S. T. (1980). Attention and weight in person perception: The impact of negative and extreme

behavior.Journal of Personality and Social Psychology,38,889–906.Fiske, S. T. (1993). Controlling other people: The impact of power on stereotyping.American

Psychologist,48,621–628.Fiske, S. T., & Depret, E. (1996). Control, interdependence and power: Understanding social cognition

in its social context. In W. Stroebe & M. Hewstone (Eds.),European Review of Social Psychology(Vol. 7, pp. 31–61). New York: Wiley.

Fiske, S. T., & Neuberg, S. L. (1990). A continuum of impression-formation, from category-based toindividuating processes. In M. P. Zanna (Ed.),Advances in experimental social psychology(Vol.23, pp. 1–74). San Diego: Academic Press.

Goethals, G. R., & Darley, J. M. (1977). Social comparison theory: An attributional approach. In J. M.Suls & R. L. Miller (Eds.),Social comparison processes: Theoretical and empirical perspectives(pp. 259–278). Washington, DC: Hemisphere.

Jones, E. E., & Davis, K. E. (1965). From acts to dispositions: The attribution process in personperception. In L. Berkowitz (Ed.),Advances in experimental social psychology(Vol. 2, pp.220–266). San Diego: Academic Press.

Klein, W. M., & Kunda, Z. (1992). Motivated person perception: Constructing justifications fordesired beliefs.Journal of Experimental Social Psychology,28,145–168.

Neuberg, S. L., & Fiske, S. T. (1987). Motivational influences on impression formation: Outcomedependency, accuracy-driven attention, and individuating processes.Journal of Personality andSocial Psychology,53,431–444.

Peeters, G., & Czapinski, J. (1990). Positive-negative asymmetry in evaluations: The distinctionbetween affective and informational negativity effects. In W. Stroebe & M. Hewstone (Eds.),European Review of Social Psychology(Vol. 1, pp. 33–60). New York: Wiley.

Pervin, L. A. (1963). The need to predict and control under conditions of threat.Journal of Personality,31,570–587.

Pratto, F., & John, O. P. (1991). Automatic vigilance: The attention-grabbing power of negative socialinformation.Journal of Personality and Social Psychology,61,380–391.

Reeder, G. D., & Brewer, M. B. (1979). A schematic model of dispositional attribution in interpersonalperception.Psychological Review,86,61–79.

Rosenberg, S., Nelson, C., & Vivekananthan, P. S. (1968). A multidimensional approach to thestructure of personality impressions.Journal of Personality and Social Psychology,9, 283–294.

Ruscher, J. B., & Fiske, S. T. (1990). Interpersonal competition can cause individuating processes.Journal of Personality and Social Psychology,58,832–842.



Ruscher, J. B., Fiske, S. T., Miki, H., & Van Manen, S. (1991). Individuating processes in competition:Interpersonal versus intergroup.Personality and Social Psychology Bulletin,17,595–605.

Schwarz, N., & Bless, H. (1992). Constructing reality and its alternatives: An inclusion/exclusionmodel of assimilation and contrast effects in social judgment. In L. L. Martin & A. Tesser (Eds.),The construction of social judgment.Hillsdale, NJ: Erlbaum.

Schulz, R. (1976). Effects of control and predictability on the psychological well-being of theinstitutionalized aged.Journal of Personality and Social Psychology,33,563–573.

Skowronski, J. J., & Carlston, D. E. (1987). Social judgment and social memory: The role of cuediagnosticity in negativity, positivity, and extremity biases.Journal of Personality and SocialPsychology,52,689–699.

Srull, T. K., & Brand, J. F. (1983). Memory for information about persons: The effect of encodingoperations upon subsequent retrieval.Journal of Verbal Learning and Verbal Behavior,22,2190–230.

Taylor, S. E. (1991). Asymmetrical effects of positive and negative events: The mobilization–minimization hypothesis.Psychological Bulletin,110,67–85.

Tyler, T. R., & Sears, D. O. (1977). Coming to like obnoxious people when we must live with them.Journal of Personality and Social Psychology,35,200–211.

Vonk, R. (1995). Effects of inconsistent behaviors on person impressions: A multidimensional study.Personality and Social Psychology Bulletin,21,674–685.

Vonk, R. (1996). Negativity and potency effects in impression formation.European Journal of SocialPsychology,26,851–865.

Vonk, R. (in press). Effects of outcome dependence on correspondence bias.Personality and SocialPsychology Bulletin.

Vonk, R., & Van Knippenberg, A. (1995). Processing attitude statements from ingroup and outgroupmembers: Effects of within-group and within-person inconsistencies on reading times.Journal ofPersonality and Social Psychology,68,215–227.

Vorauer, J. D., & Ross, M. (1993). Making mountains out of molehills: An informational goalsanalysis of self- and social perception.Personality and Social Psychology Bulletin,19,620–632.

Weinstein, N. D. (1980). Unrealistic optimism about future life events.Journal of Personality andSocial Psychology,39,806–820.

Wyer, R. S., & Gordon, S. E. (1982). The recall of information about persons and groups.Journal ofExperimental Social Psychology,18,128–164.

288 ROOS VONK


Effects of Cooperative and Competitive Outcome Dependency on Attention and Impression Preferences

Documents