Outcome Dependency: Attention, Attribution, and Attraction

Journal of Personality and Social Psychology1976, Vol. 34, No. 5, 978-989

Outcome Dependency: Attention, Attribution, and Attraction

Ellen Berscheid, William Graziano, Thomas MonsonUniversity of Minnesota

Marshall DermerUniversity of Wisconsin—Milwaukee

Theoretical and empirical work on the processes by which we attribute dis-positional characteristics to others has focused almost exclusively on how suchprocesses proceed once the perceiver has been motivated to initiate them. Theproblem of identifying the factors which prompt the perceiver to engage in anattributional analysis in the first place has been relatively ignored, even thoughthe influence of such factors may extend beyond the initiation of the causalanalysis to affect the manner in which it unfolds and, ultimately, the formand substance of its conclusion. From the assumption that the function of anattributional analysis is effective control of the social environment, it washypothesized that high outcome dependency upon another, under conditionsof high unfamiliarity, is associated with the initiation of an attributional anal-ysis as evidenced by increased attention to the other, better memory of theother's characteristics and behavior, more extreme and confidently given evalua-tions of the other on a variety of dispositional trait dimensions, and increasedattraction to the other. These hypotheses were tested within the context of astudy of heterosexual dating relationships in which men and women volunteersanticipated varying degrees of dependence upon another for their dating out-comes. The findings support the view that the data processing operations ofthe social perceiver—from attention to memory to attribution—are part of aunified whole and may be viewed as manifestations of an underlying motivationto predict and control the social environment.

Virtually all of the empirical examination of machinery is incomplete. In addition, and atthe processes by which we attribute disposi- worst, it may be misleading in relation to itstional characteristics to others, as well as most usual operation in daily life; the very factorsof the theoretical work, has been directed to- which influence whether attribution processesward an understanding of the "machinery" of are initiated may also steer the manner inthe attribution process, how such processes which these processes unfold and determine,proceed once the perceiver has been motivated as well, the point at which they cease,to initiate them; the question of when he Where do attribution processes begin? Per-fires up his attribution apparatus has been ceivers, according to attribution theorists, arerelatively ignored. rational constructors of causal schemata who

That the study of attribution processes has analyze two types of data, the behavior ofnot begun at the beginning is not without another and the context in which his behaviorconsequence. We are, of course, unable to occurred, in order to extract such invariancespredict when a perceiver will try to make at- from these data as the other's dispositionaltributions about a specific other; thus, at best, characteristics (e.g., Jones & Davis, 1965;the emerging blueprint of the attribution Kelley, 1972b). It follows, then, that just as

in any other analysis of data, the attributionprocess must necessarily start with the col-

This research was supported in part by National lection of data about the object of analysis.Science Foundation Grant GS 35157X to Ellen T h beginning of the attribution processBerscheid. We would like to thank Mark Snyder for , 5 6 , . , ,, ,his helpful comments. m a ^ b e P r e s u m e d t o be> therefore, some-

Requests for reprints should be sent to Ellen Ber- w h e r e a r o u n d t h e P o i n t a t w h i c h t h e Per"scheid, Department of Psychology, Elliott Hall, Uni- ceiver attends to a particular other, observesversity of Minnesota, Minneapolis, Minnesota 55455. his behavior and observes its context (or at-

978

OUTCOME DEPENDENCY 979

tends to such artifacts of the other as theother's responses to a questionnaire, an ac-counting of his behavior and its context by athird person, etc.).

When do attribution processes begin? Whyis the perceiver sometimes willing to spendhis time and energy in an effort to arrive at acausal understanding of another's behavior?On the perceiver's motivation for performingan attributional analysis, theorists agree;they endorse the Brunswickian-Heiderian viewthat perceptual acts function to stabilize andmake predictable the phenomenal world. Kel-ley (1972a), for example, observes, "The at-tributor is not simply an attributor, a seekerafter knowledge. His latent goal in gainingknowledge is that of effective management ofhimself and his environment" (p. 22). Hecomments further, in fact, that "The attribu-tional process can be well understood only inthe context of a comprehensive analysis ofthe exercise of control" (p. 22). It follows,then, that the attribution machinery is mostlikely to be activated for a person in the per-ceiver's social environment who representsa potential control problem to him; a personwho does not present the perceiver with prob-lems of control should be unlikely to be thesubject of an attributional analysis andshould, therefore, be unlikely to invite atten-tion in the service of such an analysis.

Confidence in the designation of attentionto another as the beginning point of the at-tribution process is increased by the apparentsimilarity between the factors which reflectthe degree to which a person in the perceiver'ssocial environment may represent a potentialcontrol problem to him and those which areknown to influence attentional processes. Thefactors perception theorists and researchershave identified as governing attentional pro-cesses generally may be subsumed under thetwo general principles of novelty and impor-tance (cf. Berlyne, 1960, 1974; Lanzetta,1971).

In terms of the perceiver's exercise of con-trol over his social environment, it seems rea-sonable to assume that, other things beingequal, novel or unfamiliar people are morelikely to represent potential problems of con-trol than are familiar (and predictable) others.While the determinants of importance have

not been strictly specified, the degree of impor-tance of social stimuli should increase withanother's ability to control the perceiver's re-wards and punishments. Thus, the importanceof a social stimulus to the perceiver should bea function of the perceiver's outcome de-pendency upon that other (cf. Thibaut &Kelley, 1959).

Thus, we can predict that, other thingsbeing equal (particularly the degree of "nov-elty" of persons in the perceiver's social en-vironment), persons on whom the perceiver'soutcomes are dependent should be especiallylikely to trigger the attribution machinery; asthe perceiver's outcome dependency upon an-other increases, there should be an increasedtendency for the perceiver to be attentive tothe other, to collect data about the other'sbehavior and the context in which the be-havior occurred, and to perform an attribu-tional analysis upon these data.

We speculated earlier that the same factorswhich influence whether attribution processesare initiated in the first place may also in-fluence the manner in which the attributionalanalysis subsequently unfolds. If it is truethat outcome dependency is one determinantof whether an attributional analysis is begun,then we may make at least two predictionsabout how this particular factor might reignover the whole causal analysis.

First, we may predict that as his outcomedependency on the person who is the object ofhis analysis increases, so does the perceiver'spreference that the analysis conclude in a dis-positional rather than a situational attribu-tion. Knowledge of another's personal andinternal predispositions to behave in certainways would seem to be a more useful bit ofinformation to the perceiver who hopes to ex-ercise control over that part of his social en-vironment than an attribution of the other'sbehavior to nonpersonal situational factorswould be (cf. Shaver, 1975). Thus, the factortriggering the analysis may influence the formof the attribution arrived at.

Second, when the triggering factor is out-come dependency, the social perception lit-erature allows us to predict that it will in-fluence the substance of the attribution aswell. The results of several studies investigat-ing the influence of perceiver need states upon

980 BERSCHEID, GRAZIANO, MONSON, AND DERMER

social perception may be interpreted to sup-port the thesis that persons who are outcomedependent upon another tend to arrive atwish-fulfilling conclusions about the other'sdispositions (e.g., Berscheid, Boye, & Darley,1968; Darley & Berscheid, 1967; Levine,Chein, & Murphy, 1942; Pepitone, 1950;Stephan, Berscheid, & Walster, 1971). Thus,as the perceiver's outcome dependency uponanother increases, the dispositional attribu-tions made about the other should be in thedirection of wish fulfillment (even despite thefact that dependency should lead the perceiverto pay close attention to the other's behaviorand the context in which it occurred).

We can make, therefore, the following pre-dictions: As the perceiver's outcome depen-dency on another in his social environmentincreases, the perceiver should (a) attendmore closely to the other; (b) remember andrecognize more details about the other's be-havior; (c) be more likely to make disposi-tional attributions about the other (as evi-denced by more extreme and more confidentresponses on trait scales, Jones & Davis,1965); and (d) ascribe more positive dis-positional characteristics to the other and,thus, (e) like the other more.

METHOD

Overview

To investigate the hypotheses outlined above, itwas necessary to devise a situation in which (a) thedegree to which persons in the perceiver's social en-vironment represented potential problems of controlto the perceiver could be manipulated; specifically,the perceiver's outcome dependency on each could bevaried, and novelty could be held constant; (b)the degree of attention the perceiver subsequentlyawarded to each of these others; as well as (c) theextent to which the perceiver arrived at dispositionalinferences about each could be ascertained; and fi-nally, (d) the degree to which these dispositional at-tributions reflected wish-fulfilling distortion could bespecified a priori.

We also wished the situation to permit administra-tion of an individual difference measure, Snyder'sSelf-Monitoring Scale (Snyder, 1974), to assess theperceiver's general dependency upon persons in hissocial environment for cues to guide his or her ownbehavior. Snyder has hypothesized that high self-monitoring persons are particularly concerned withthe "appropriateness" of their own behavior and, forthis reason, are especially attentive to the behaviorsemitted by others. Thus, the behaviors hypothesized

to characterize pcrccivers scoring high on this indi-vidual difference measure resemble the behaviors wehypothesize will characterize perceivers who findthemselves in a situation in which their outcomes arehighly dependent on another.

All of these conditions were satisfied, for our col-lege age population, by a situation in which the per-ceiver agreed to turn over his (or her) heterosexualdating life to us, the experimenters, for a period of 5weeks, promising that he would date whenever, andwhomever, we instructed, fully aware that it couldbe his lot to date no one for the period, the sameperson many times over the period, or many differentpeople over the period. We were thus able to situa-tionally manipulate the perceiver's outcome depen-dency upon a person by manipulating his expectationthat he or she would be unlikely to date that par-ticular person but rather would be dating others forthe 5-week period (zero exclusiveness) ; would datethat person once and also others during the period(low exclusiveness); would date that person exclu-sively over the entire 5-week period (high exclusive-ness) .

Participants

Twenty-nine males and 28 females who werebetween 18 and 22 years of age volunteered for anadvertised experimental study of "heterosexual datingrelationships." These individuals also had indicatedthat they were not currently romantically involvedwith any one person (i.e., married, engaged, goingsteady, etc.) and were seriously interested in meetingpersons of the other sex. All but three were willing tocommit themselves to the dating terms describedbelow, leaving 27 males and 27 females who sub-sequently participated in the study.

Procedure

Each participant was interviewed individually andasked if he or she would agree to three stipulations:(a) to participate in a dating study to last 5 weeksand to date only those individuals assigned to themon the basis of a random lottery and to date ab-solately no one else; (b) to report their impressionsof their assigned date following each contact; and(c) to engage in a short videotaped discussion of"dating problems on campus" with two other par-ticipants of the same sex and to view a videotape ofa similar discussion made by three discussants of theother sex.

To assure commitment to these procedures andconditions, each participant was asked to sign aformal guarantee to the above stipulations.

Exclusiveness manipulation. Participants were ran-domly assigned to either the high-, the low-, or thezero-exclusiveness condition. In the high-exclusive-ness condition, the participant was given the prospec-tive date's name and phone number, and was in-formed that he or she would be dating this personexclusively—once a week—for the entire 5-week pe-riod. In the low-exclusiveness condition the pro-


cedure was the same, except that the participant wasinformed that he or she would be dating this persononly once, and that a different person would be datedfor each of the remaining 4 weeks. The procedure forthe zero-exclusive ness condition was identical to thelow-exclusiveness condition, except that the namegiven as the prospective date was not among thenames of the discussants whom the participant wouldshortly view on the videotape. Since none of thevideotaped discussants in the zero-exclusiveness condi-tion would be a date, this condition was not includedas part of the factorial design, but was included toprovide base-rate levels of response on the principledependent measures (cf. Himmelfarb, 1975).

Stimulus materials and apparatus. After the par-ticipant received the date's name and number, theexperimenter asked whether he or she presently hadenough time to view a videotaped discussion of someparticipants of the other sex for the purpose of gain-ing perspective on dating problems.

Participants viewed the discussion over a speciallydesigned system which consisted, in part, of a tele-vision monitor and three switches. The experimenterexplained that due to the limitations of the recordingequipment, it would be possible to view only one ofthe three discussants on the videotape at a time. Heindicated that when any one of the switches was de-pressed, only the corresponding discussant wouldappear on the monitor; for example, if the left switchwere depressed, only the participant who had beensitting in that location of the discussion room wouldappear. The audio track was common to all threeswitches. If no switch was depressed, there wasneither video nor audio. If more than one switch wasdepressed, the video was scrambled. For a switch tobe operative, it had to be actively held down; re-lease resulted in loss of both video and audio. Con-nected to each switch was a remote cumulative timerwhich measured the time a participant attended toeach of the discussants. The experimenter was notpresent while the participant viewed the videotape.

Male participants viewed a videotape of three fe-male confederates who engaged in an apparentlyspontaneous IS-minute discussion of dating problems;female participants viewed a tape of three male con-federates. The confederates were selected, and thetapes were edited, on the basis of extensive pretesting,such that (a) the stimulus persons within each tapeall appeared to be slightly above average in physicalattractiveness and equivalent to each other on thisdimension; (b) male confederates were as attractiveto females as female confederates were to males; (c)each person spoke for about the same length of time;and (d) in general, they appeared to be of aboutequivalent dating desirability.

Each stimulus person discussant was dressed dis-tinctively yet stylishly, and each mentioned his orher interests, hobby, academic major, etc., in thecourse of the dialogue. For the first 7.5 minutes ofconversation, however, the discussants did not men-tion their names. Thereafter, the discussants intro-duced themselves to each other using their "names."

Participants in the high- and the low-exclusivenessconditions anticipated dating one of the three stimu-lus persons on the videotape. The prospective date'sname had been randomly selected for each participant.Each of the stimulus persons on each tape was desig-nated a prospective date for an equal number ofparticipants in each of the high- and low-exclusive-ness conditions. In the zero-exclusiveness condition,participants expected to date a person whose namewas given to them, but that person was not amongthe discussants on the videotape viewed.

Questionnaires. After the participant had viewedthe videotape, the experimenter returned and askedthe participant to complete an "impression formationquestionnaire." This measure was ostensibly intro-duced as a means of determining if men and womendiffered in the way they responded to videotaped dis-cussions and the impressions they derived from them.

The first items on the questionnaire were memoryitems, both recall and recognition, for each of thethree stimulus persons. They required the participantto answer questions pertaining to specific things saidor articles of clothing worn by each of the stimuluspersons.

Next, participants were asked to complete 15 bi-polar trait scales (e.g., warm-cold, strong-weak) foreach of the three stimulus persons. Immediately be-neath each trait scale was a bipolar "confidence"scale. Thus, after each trait rating was made, par-ticipants rated their confidence in the rating they hadmade. According to Jones and Davis (1965), attribu-tions of dispositions to a stimulus person are evi-denced by perceiver ratings of the stimulus personwhich are extreme on trait dimensions, and ratings inwhich the perceiver indicates confidence.

In addition, participants were asked to rate theirliking for each stimulus person on a 41-point bipolarliking scale. And finally, participants were asked tocomplete the self-monitoring scale (Snyder, 1974).

After participants had completed all questionnaires,they were interviewed extensively. No participant re-ported suspicion, nor were any able to guess the truenature of the experiment. Participants were debriefedand thanked for participating in the experiment,which was terminated at this point. To fulfill thepromise of a date, however, each male was randomlyassigned to a female participant.

RESULTS

Independent Variables and Design

The independent variables were:

1. Date versus nondate. Since there werethree stimulus persons, and only one of thesewas the anticipated date, the remaining twowere designated as "nondates." For each de-pendent measure, scores for the two nondateswere summed and divided by two, yielding anaverage nondate score. Thus, each participantproduced two scores for each dependent mea-


sure: a score for the anticipated date and ascore representing the average of the twonondates. For the analyses, this within-sub-jects variable was nested within the other twoindependent variables for two reasons: (a) Itwas predicted that the other two independentvariables would affect responses to the an-ticipated date, but not responses to the non-date; and (b) on some dependent variables,there was a direct dependency of certain datescores on nondate scores; for example, if aparticipant attended to his anticipated date,he could not simultaneously attend to hisnondate.

2. Exclusiveness of relationship. Althoughthere were three levels of exclusiveness in thestudy (zero exclusiveness, 0-dates; low exclu-siveness, 1-date; and high exclusiveness, 5-dates), only low and high exclusiveness wereincluded in the major analyses. Since therecould be no date-versus-nondate factor in thezero-exclusiveness condition (all three stimu-lus persons were nondates), data obtainedfrom the zero-exclusiveness condition wereanalyzed in separate auxiliary comparisons.

3. Self-monitoring. Each of the participantsin the low-exclusiveness and high-exclusive-ness conditions were classified as either highin self-monitoring or low in self-monitoring onthe basis of a median split of scores (Mdn =12) on the Snyder Self-Monitoring Scale.1

4. Sex of participant. Due to the multipledependent measures and the likelihood ofsubstantial intercorrelations between thesemeasures, the data were analyzed in a singlemultivariate analysis of variance (MANOVA).Since there were no systematic differences(main effects or interactions) between maleand female participants on any of the de-pendent variables (all MANOVA Fs < 1.70),data for male and female participants werecombined for all subsequent analyses. Thus,subsequent MANOVAs were based on a 2X(2x2) design and were followed by separateunivariate ANOVAs for each dependent mea-sure (Hummel & Sligo, 1971).

Dependent Variables

Each participant produced seven dependentmeasures for the anticipated date and sevendependent measures for the average nondate.

These were:

1. Attention. The attention dependent mea-sure was the percentage of time the antici-pated date and the average nondate werewatched after the stimulus persons had men-tioned their names.

2. Recall. The recall dependent measureconsisted of "fill-in-the-blank" items, such as"What was the academic major of the personon the left (on the videotape) ?"

3. Recognition. The recognition dependentmeasure consisted of multiple choice items re-ferring to details of the videotaped discussionof the stimulus persons. For example, par-ticipants were asked, "Who said he liked to goto sports events on a first date?" Choices werethe person on the right (on the videotape),person on the left, person in the middle, andno one. Since males and females saw differentvideotapes, items drawn from the one tapemay have been more difficult than itemsdrawn from the other tape. For this reason,recognition items were converted to stan-dardized z scores before the recognition scoreswere analyzed. By converting to z scores,mean difficulty across items and across tapeswas equated.

4. Extremity of trait ratings. This depen-dent measure was the sum of the absolute de-viation from the midpoint of 21 on each of the15 bipolar 41-point trait scales.

5. Confidence of trait ratings. After par-ticipants had completed each rating of a stim-ulus person on a bipolar trait scale, they wereasked to indicate their confidence in that rat-ing on a 41-point scale which was placed im-mediately beneath the trait scale. Scores onthe IS confidence scales were summed.

6. Positivity of trait ratings. Each of the 15trait scales had a positive pole (e.g., "warm,""sensitive") and a negative pole (e.g., "cold,"

1 The range of self-monitoring scores in the low-exclusiveness condition was 5 to 18, with 11 partici-pants at or above the median of 12 and 7 below themedian of 12. The range of self-monitoring scores inthe high-exclusiveness condition was 5 to 22, with 11at or above the median of 12 and 7 below the me-dian. Thus, cell frequencies are proportional. Thecorrelation between the total range of self-monitoringscores and each of the dependent variables is reportedin Table 3.


TABLE 1

MANOVA AND ANOVA SUMMARY TABLE FOR THREE OUTCOME DEPENDENCY VARIABLES

Variable

MANOVA

Univariate ANOVAsAttentionMemory

RecallRecognition

AttributionExtremityConfidence

AttractionLikingPositivity

Date/Nondate 'F

44.71****

58.66****

259.54****31.83****

4.32*5.53*

6.43*12.63****

Exclusiveness condition

DateF

2.20*

8.96***

.00

.05

2.46.27

3.50*4.51*

NondateF

2.50*

8.43**

.082.23

9.13***1.86

1.074.11*

Self-monitoring

DateF

2.49*

2.02

6.77**.62

10.31***4.38*

4.51*5.67*

NondateF

1.48

1.97

1.03.01

8.22**1.11

.28

.81

7, 26; for univariate ANOVAs, <// = !, 32.Note. For MANOVA effects, df ••*p < .05.

** j> < .01.*** p < .005.

****p < .001.

"insensitive"). Each trait scale was scoredsuch that a response nearest the negative polereceived a score of 1, and a response nearestthe positive pole received a score of 41. Scoreson all 15 scales were summed.

7. Attraction. The liking dependent measureconsisted of the response to the question"How much did you like this person?" on a41-point bipolar scale.

MANOVA and Univariate Analyses

No significant interactions were found be-tween the exclusiveness variable and the self-monitoring variable on any of the dependentmeasures for either the anticipated date orthe nondate (all Fs < 1.14). Thus, the analy-ses reported below describe the main effectswhich were evident for the three outcome de-pendency factors: date/nondate, high/low ex-clusiveness, high/low self-monitoring. Theseanalyses are presented in Table 1.

First, there was a highly significantMANOVA date-versus-nondate effect, F(7,26) = 44.71, p < .00001. This result sup-ported our prediction that an anticipated datewould elicit different responses on the de-pendent measures than would a nondate.

Second, there was a significant MANOVAexclusiveness effect for the responses to theanticipated date, F(7, 26) = 2.20, p < .06.This finding supported our prediction that aperson who was expected to be a date five

times was responded to differently than was astimulus person who was expected to be adate only one time. Contrary to our predic-tions, however, there was also a significantMANOVA exclusiveness effect for the non-date, F(7, 26) = 2.60, p < .04. Exclusivenessof the relationship with the date affected reac-tions to the nondate as well.

There was a significant MANOVA self-monitoring effect for the anticipated date,F(1, 26) = 2.49, p < .04. This finding sup-ported our prediction that participants highin self-monitoring would respond to an an-ticipated date differently than low self-moni-tors would. As predicted, there was no signifi-cant MANOVA self-monitoring effect for thenondate, F(7,26) = 1.48, ns.

The univariate ANOVAs provide more spe-cific tests of our hypotheses of participants'responses on each of the dependent measures.

Attention

A significant univariate date-nondate ef-fect for attention, F{\, 32) = 58.66, p <.000001, indicated that the anticipated datewas awarded more of the participant's atten-tion (M = 40.53%) than was the nondate(Jlf = 29.69%).

As predicted, there also was a significantunivariate exclusiveness effect for the antici-pated date on the attention dependent mea-sure, JF(1 , 32) = 8.96, p < .005; an antici-


pated date was given more attention in thehigh-exclusiveness condition (M = 43.33%)than in the low-exclusiveness condition (M =37.72%).

The univariate self-monitoring effect for theanticipated date was not significant on thisdependent measure, F(l, 32) = 2.02, p < .16;men and women high in self-monitoring didnot spend more of their time watching theiranticipated date than those low in self-moni-toring did.

Memory

A highly significant univariate date-versus-nondate effect for recall, F(l,32) =259.54,p < .000001, was found; this finding indicatedthat, as predicted, participants recalled moredetails about their anticipated date (M =2.83) than about the nondate (Af=1.33).

There was a highly significant date-non-date effect for recognition as well, F(l, 32) —31.83, p < .00003, indicating that participantsrecognized more details pertaining to theiranticipated date ( M = + . 6 6 ) than to thenondate (M = — .17). To investigate whetheraccuracy in recognition was due to a responsebias toward attributing all statements to thedate, analyses of "false positive" scores wereconducted. The results of these analyses didnot support this interpretation of the recogni-tion finding; false positive scores did not varywith condition (all Fs < 1.26).

The univariate exclusiveness effect for re-call for the anticipated date was not signifi-cant, 7^(1,32) = .00, nor was the univariateexclusiveness effect for recognition for theanticipated date, F{ 1,32) = .05.

The univariate self-monitoring effect for re-call for the anticipated date was significant,77(1,32) = 6.77, p < .01; those high in self-monitoring recalled more about their antici-pated date (M = 3.00) than did those low inself-monitoring (M = 2.57).

Trait Ratings

Extremity. There was a significant uni-variate date-nondate effect for extremity oftrait rating, F(l,32) =4.32, p < .05; par-ticipants rated their anticipated date more ex-

tremely (M = 132.06) than they did theirnondates (M = 120.72).

Although the means were ordered as pre-dicted, the univariate exclusiveness effect forthe anticipated date was not significant onthe extremity of trait rating measure, F(l, 32)= 2.46, p < .13. The univariate exclusivenesseffect for extremity for the nondate was sig-nificant, however, F(l, 32) = 9.13, p < .005;this finding indicated that a nondate was ratedmore extremely in the high-exclusiveness con-dition (M = 135.83) than in the low-exclu-siveness condition (M = 105.61).

There was a significant univariate self-monitoring effect for extremity for the an-ticipated date, F(l, 32) = 10.31, p < .003,indicating that those high in self-monitoringrated their anticipated date more extremely(M = 150.36) than did persons low in self-monitoring (M = 103.29). High self-monitor-ing individuals were also more extreme inrating their nondates (M = 132.16) than werethe low self-monitoring participants (M =102.75); F(l,32) = 8.22, p < .007.

Confidence. For the confidence in trait rat-ing dependent measure, there was a significantunivariate date-nondate effect, F(l,32) =5.53, p < .03; this finding indicated that theanticipated date was rated more confidently(M = 393.67) than were the nondates (M =372.86).

The univariate exclusiveness effect for con-fidence in rating the anticipated date was notsignificant, F(l, 32) = .27. The univariateself-monitoring effect for confidence in traitrating of an anticipated date was significant,F(l, 32) = 4.38, p < .04, indicating that per-sons high in self-monitoring were more con-fident in rating an anticipated date (M =418.19) than were persons low in self-monitor-ing (M = 354.00).

Dispositional attributions. Following Jonesand Davis' operational definition of disposi-tional attributions, when a perceiver rated astimulus person both extremely and with con-fidence on trait rating scales, we assumed hehad made dispositional attributions to the per-son. It will be recalled that the extremity andconfidence data analyzed in the two separateANOVAs indicated that perceivers were bothmore extreme and more confident in rating


persons they anticipated dating than theywere in rating persons they did not expect todate. These analyses do not tell us, however,whether extremity and confidence covaried inratings of the same stimulus person. To in-vestigate whether this was indeed the case,within each exclusiveness condition, we cor-related the extremity score with the con-fidence score for the ratings of the date, andcompared this correlation with the correlationbetween average nondate extremity scores andconfidence scores.

These analyses revealed that within thelow-exclusiveness (1-date) condition, extrem-ity and confidence were correlated +.77 (df =1.6) for the date, and +.41 (df = 16) for theaverage nondate. Within the high-exclusive-ness (5-date) condition, extremity and con-fidence were correlated +.85 (df = 16) forthe date, and +.63 (df == 16) for the averagenondate. Thus, extremity and confidence weremore highly correlated for dates than theywere for the average nondates. In both ex-clusiveness conditions, the correlation betweenextremity and confidence of trait rating wassignificantly higher for the date than for thenondate.2 (A similar pattern was found forboth those high and those low in self-monitor-ing in both conditions.)

Extremity and confidence ratings from thezero-exclusiveness (0-date) condition providedespecially valuable information relevant tothis issue. Since all three stimulus persons inthis condition were nondates, dispositional at-tributions should have had relatively lessutility for the perceivers in this condition. Asa consequence, we expected the correlation be-tween extremity and confidence scores to belower for all stimulus persons in the zero ex-clusiveness condition than for the anticipateddate in the other conditions. As predicted,there was no significant correlation (r =+ .07, df = 16) between extremity and con-fidence of trait ratings in the zero exclusive-ness condition.

It might be argued that the lack of correla-tion between extremity and confidence in thezero-exclusiveness (base-rate) condition merelyreflects a restriction in range since we had ex-pected ratings in this condition to be lessextreme and less confidently given than in acondition in which a date was anticipated. To

investigate this possibility, the variances as-sociated with the extremity and confidence oftrait ratings in the zero exclusiveness condi-tion were compared with the variances associ-ated with the extremity and confidence rat-ings of an anticipated date, collapsing acrossthe high- and low-exclusiveness conditions.There was no reliable evidence of differentialvariability on the extremity ratings betweenthe zero-exclusiveness condition (SD = 45.65)and the conditions in which a date was an-ticipated (SD = 48.96), F(35, 17) = 1.15,ns; nor was there differential variability on theconfidence ratings between the zero exclusive-ness condition (SD = 75.68) and the an-ticipated date conditions (SD = 93.62), F(35,17) = 1.53, WJ.

It should be noted that, contrary to ourexpectation, the stimulus persons in the zero-exclusiveness condition were not rated lessextremely than were anticipated dates. In ad-dition, the stimulus persons were not ratedany less confidently in the zero-exclusivenesscondition than in the anticipated date con-ditions. The ANOVA summary and marginalmeans for extremity, confidence, and the otherdependent variables are presented in Table 2.

2 Due to the dependency between ratings for thedate and ratings for the nondate, these correlationscould not be directly contrasted. We would like tothank Auke Tellegen for suggesting a test based onthe following reasoning: When two variables x and yare perfectly correlated (rxy = ±1.00) and when eachpair of * and y scores are converted to z scores,within each pair, zx —zv; the smaller the relationshipbetween x and y, the greater discrepancy within pairsof z scores. Therefore, within both high- and low-exclusiveness conditions, extremity scores and con-fidence scores were converted to z scores for the dateand for the nondate. The z scores for extremity weresubtracted from the z scores for confidence for thedate, producing a date discrepancy score. A similardiscrepancy score was produced for the nondates. Itwas hypothesized that if there were a higher correla-tion between extremity and confidence for the datethan for the nondate, the discrepancy score for thedate should be significantly smaller than for the non-date. This hypothesis was tested using t tests for cor-related measures. There was a significantly smallerdiscrepancy score for the date than for the nondatein both the low-exclusiveness condition, t(l7) =2.36,p < .05, and in the high-exclusiveness condition, t (17)= 2.96, p < .01. (Both t tests were two-tailed.)


TABLE 2SUMMARY TABLE AND MEANS FOR AN ANTICIPATED

DATE (COLLAPSING ACROSS HIGH- ANDLOW-EXCLUSIVENESS CONDITIONS) VERSUS

NONDATE IN ZERO-EXCLUSIVENESS CONDITION

ANOVA effect

Attention

MemoryRecallRecognition

AttributionExtremityConfidence

AttractionLikingPositivity

AnticipateddateM

40.53%

2.83+.66

132.06393.67

29.22401.33

NondateM

33.33%

1.61- . 2 1

132.46400.43

25.54366.98

F

24.30***

60.70***4.89*

.00

.07

5.63**4.82*

Note. For ANOVA effects, df*P < .05.

** p < .025.***p < .001.

' 1, 52.

In summary, there is no evidence that par-ticipants responded less extremely or less con-fidently in the zero-exclusiveness condition.This suggests that the lack of correlation be-tween extremity and confidence in the zero-ex-clusiveness condition might better be attrib-uted to less coherence and correspondence inthe inference process in this condition than toa restriction in range artifact.

Attraction

To investigate our hypotheses on the at-traction dependent measure, we return now tothe ANOVA analyses. There was a significantunivariate date-nondate effect for liking,F(l,32) = 6.43, p < .02. As predicted, par-ticipants liked their anticipated date signifi-cantly more (M = 26.22) than they likednondates (Af = 25.44).

The univariate exclusiveness effect for lik-ing was marginally significant, J F ( 1 , 3 2 ) =

3.5O; p < .07; it suggested that an anticipateddate was liked more under high-exclusiveness(M = 30.83) than under low-exclusiveness(M = 27.61) conditions.

The univariate self-monitoring effect forliking was also significant, JP(1 , 32) =4 .51 ,p < .04, indicating that those high in self-monitoring liked their anticipated dates more(M = 30.68) than did those low in self-moni-toring (M = 36.93).

When we examine the positivity of traitratings, we find a significant univariate date-nondate effect, F ( l , 32) = 12.63, p < .001,indicating that the anticipated date was per-ceived more positively (M — 401.33) thanwas the nondate (M = 363.11).

The univariate exclusiveness effect for posi-tivity of trait ratings for the anticipated datewas also significant, F(l, 32) =4 .51 , p <.04, indicating that an anticipated date wasrated more positively under high-exclusive-ness (M = 421.50) than under low-exclu-siveness (M = 381.77) conditions. Unexpect-edly, but in accord with some results on otherdependent measures, the univariate exclusive-ness effect for the nondate was also significant,F(l, 32) = 4.11, p < .05, indicating nondatesin the high-exclusiveness condition were ratedmore positively (M = 380.06) than were non-dates in the low-exclusiveness condition (M= 346.17).

The univariate self-monitoring effect for theanticipated date was also significant, F(l, 32)= 5.67, p < .02; those high in self-monitoringrated their anticipated date more positively(M = 419.36) than did those low in self-monitoring (M = 373.00).

It will be recalled that we hypothesizedthat in general, under conditions of high out-come dependency, a person should be highlymotivated to predict the behavioral predisposi-tions of the person on whom he or she is de-pendent. We expected this motivation to bereflected in high intercorrelations among allthe dependent variables under conditions ofhigh outcome dependency, but low intercor-relations under low outcome dependency.Since each of the independent variables in theMANOVA analyses we conducted may beviewed as convergent conceptualizations ofoutcome dependency (cf. Messick, 1975), wemay interpret our finding that each of theseMANOVA effects were significant as suggest-ing that a single factor, outcome dependency,was underlying participants' responses on thedependent measures. In addition, Table 3presents the intercorrelations among the sevendependent measures, as well as their correla-tions with self-monitoring scale scores, for theanticipated date and for the nondate (col-lapsing across the 1-date and 5-date condi-tions) and the zero-date stimulus persons aswell. It can be seen that responses on the de-


TABLE 3INTERCORRELATION MATRIX FOR DEPENDENT VARIABLES FOR THE ANTICIPATED DATE, NONDATE,

AND NONDATE FROM ZERO-EXCLUSIVENESS CONDITION ("ZERO DATE")

Attention

DateNondate0-date

Memory

RecallDateNondate0-date

RecognitionDateNondate0-date

Attribution

ExtremityDateNondate0-date

ConfidenceDateNondateO-date

Attraction

LikingDateNondate0-date

PositivityDateNondate0-date

Recall

.17

.10- .06

Recognition

- . 01- . 23

.08

- . 0 6- .10

.06

Extremity

.35*- .19- .18

.38*

.02- .18

- . 2 4- . 2 0

.27

Confidence

.34*- .17- . 0 1

.28

.06- . 05

- .19- .10

.21

.80***

.57***

.07

Liking

.22

.20.21

.19

.12- .14

.02.11

- .12

.67***

.17- . 04

.50**

.20

.02

Positivity

.31

.23.11

.28

.03- .22

- .12- . 09

.02

.84***

.45***- .32

.61***

.32

.01

.58***

.70***

.69**

Self-monitorScore

.39*- .18

.06

.38*

.11

.27

- .01.04.13

.45**

.31

.05

.36*

.26

.08

.35*.05

- .11

.36*

.11

.08

the correlations of the two nondate stimulus persons and are based on 36 observations (df = 34). Correlations for O-date are derivedfrom the average of the correlations of the three O-date stimulus persons and thus are based on 18 observations {dj — 16).

*p <.O5.**p <.O1.

***p <.001.

pendent measures tended to be more highly is effective control" (p. 22)? we reasoned thatintercorrelated for the anticipated date than outcome dependency upon another, under con-for the nondate and zero-date stimulus per- ditions of high novelty, may be one such initi-sons. ating factor.

With divergences only in some particulars,^ ^ our results generally support the specific hv-DlSCUSSION AND CONCLUSIONS ., , ° A / *TL u •

potneses deduced from the above reasoning.We hypothesized, it will be recalled, that With increases in a person's dependency upon

the same factors which prompt the initiation another (for his or her dating outcomes, inof the attribution process may guide, in pre- this case), he or she tended to award moredictable ways, how the process unfolds, in- attention to the other, remember more aboutfluencing both the form and the substance of the other, evaluate the other more extremelyits conclusion. Following the view that all and more confidently on a variety of disposi-perceptual acts are performed in the service tional traits, and be more attracted to theof rendering the environment stable and pre- other as indicated by the positivity of traitdictable, and particularly Kelley's (1972a) ratings of the other and by self-reported likingview that "The purpose of causal analysis . . . of the other.


These last findings do not appear to be duesolely to the fact that perceivers paid moreattention to, and thus should have had moreinformation about, the other on whom theyanticipated dependence. Several lines of evi-dence argue against a purely informational in-terpretation of the memory and attraction re-sults. First, there were few significant cor-relations between attention time and the otherdependent variables (see Table 3). Second,while scores on the self-monitoring scaleproved to be associated with attribution, theywere not related to attention. Third, nondatesin the high-exclusiveness condition were ratedmore extremely and more confidently thanwere nondates in the low-exclusiveness condi-tion even though they were awarded less at-tention.

Consideration of the putative function ofattribution processes to make the social en-vironment more predictable, and hence moresusceptible to control, led us to hypothesizethat there may be stable individual differencesin the desire for such predictability. We hopedthat scores on Snyder's Self-Monitoring Scalemight be indicants of such differences; we rea-soned that the evidence which demonstratesthat high self-monitoring perceivers dependmore upon cues provided by others to guidetheir own social behavior than do low self-monitoring perceivers suggests that highscorers on this scale may be habitually moreconcerned with making accurate predictionsabout their social environment.

The positive and significant relationshipswhich were found between an individual's self-monitoring score and the attention he or sheawarded the prospective date, the extremityand the confidence of ratings of that other, aswell as memory of and attraction to the other,are consistent with our reasoning. It might benoted, too, that these results suggest that highself-monitoring individuals do not straight-forwardly and simple-mindedly copy the cuesprovided by another's behavior to guide theirown; rather, our findings indicate that theseindividuals expend perceptual/cognitive timeand effort to discern the contingencies whichunderlie the other's behavior.

While the attention and memory results aregenerally straightforward and as expected, theattribution results invite further comment. It

will be recalled that while neither the ex-tremity ratings nor the confidence ratings ofthe zero-exclusiveness condition differed fromthose of the other conditions, within the zero-exclusiveness condition extremity and con-fidence of trait ratings were not systematicallyrelated to each other. By comparison, extrem-ity and confidence of trait ratings for theprospective date were highly correlated in theother conditions.

These findings suggest that extremity andconfidence of trait ratings served differentfunctions in the two situations in which rat-ings were made. In particular, it suggests thatthose perceivers who were dependent upon theother for their dating outcomes tried to con-struct a coherent perception of that other (cf.Neisser, 1966) as evidenced by the kind ofcognitive constructions (i.e., dispositional at-tributions) they produced, but that those per-ceivers who did not anticipate dependence onthe other did not attempt to construct sucha systematic picture.

Perhaps more importantly, the results ofthese internal analyses investigating the rela-tionship between extremity and confidencesupport the primary findings of this study andthe conclusion that schemata of the attribu-tion process which are developed from in vitroobservations of the laboratory, and whichneglect the motivational bases of these pro-cesses, may be at variance with how theseprocesses operate in vivo. It seems fair to saythat our zero-exclusiveness condition closelyapproximates the typical person perceptionlaboratory situation. In such situations, theperceiver typically observes, or is given in-formation about, stimulus persons with whomhe anticipates little or no future interaction(and, in some cases, he is even aware that thestimulus person does not exist). In addition, itis typical of these situations that the per-ceiver might not even notice the stimulus per-son but for the experimenter's request that heread about or watch the stimulus person toevaluate him on a questionnaire. If the typicalperson perception study participant performssimilarly to our zero-exclusiveness conditionperceivers, their ratings of the stimulus per-son may be completed unsystematically withrespect to confidence and extremity. Thus,whether their evaluations are representative


of the type of evaluations they would spon-taneously make under more naturalistic con-ditions is in doubt. At the least, the presentfindings suggest that the terms "dispositionalattribution" and "rating" should not be usedinterchangeably as they often are in the per-son perception literature.

Finally, we might comment upon the gen-eral role of motivational factors in the at-tribution process. In his discussion of attribu-tion processes in the context of control, Kel-ley (1972a) states, "We might simply leaveour analysis at the point to which we havebrought it—with the recognition that at-tributions are often reasonable but that theyare also often influenced by personal interest,emotional considerations, and such" (p. 22).The results of the present study allow us toentertain the thesis that the conditions whichprovide the motivation for the perceiver toinitiate the attribution process in first placemay be the very conditions under which he islikely to invest "personal interest, emotionalconsiderations, and such" in the outcome ofthat causal analysis (cf. Berscheid & Gra-ziano, in press, for further discussion of, andevidence relating to, this proposition).

In addition, and in accord with contempo-rary recognition that perception, memory, andcognition are so inextricably interwoven thatit is difficult to study each separately (e.g.,Flavell, 1970; Neisser, 1966), the presentstudy demonstrates that all of the data-pro-cessing operations of the social perceiver—thecollection of data (attention), the selectivestorage and retrieval of data (memory), aswell as inferential extrapolation on the basisof the data (attribution)—are part of aunified whole and may be viewed as mani-festations of an underlying motivation topredict and control the social environment.

REFERENCES

Berlyne, D. Conflict, arousal and curiosity. NewYork: Academic Press, 1960.

Berlyne, D. Attention. In E. Carterette & M. Fried-man (Eds.), Handbook of perception (Vol. 1). NewYork: Academic Press, 1974.

Berscheid, E., Boye, D., & Darley, J. Effect of forcedassociation upon voluntary choice to associate.Journal of Personality and Social Psychology, 1968,8, 13-19.

Berscheid, E., & Graziano, W. The initiation of socialrelationships and social attraction. In R. L. Burgess& T. L. Huston (Eds.), Social exchange in develop-ing relationships. New York: Academic Press, inpress.

Darley, J., & Berscheid, E. Increased liking as a resultof the anticipation of personal contact. Human Re-lations, 1967, 20, 29-40.

Flavell, J. Cognitive development. In P. Mussen(Ed.), Carmichael's manual of child psychology(Vol. 2). New York: Wiley, 1970.

Himmelfarb, S. What do you do when the controlgroup doesn't fit into the factorial design? Psy-chological Bulletin, 1975, 82(3), 363-368.

Hummel, T., & Sligo, J. Empirical comparison ofunivariate and multivariate analysis of varianceprocedures. Psychological Bulletin, 1971, 76(1),49-57.

Jones, E., & Davis, K. From acts to dispositions: Theattribution process in person perception. In L.Berkowitz (Ed.), Advances in experimental socialpsychology (Vol. 2). New York: Academic Press,1965.

Kelley, H. Attribution in social interaction. In E.Jones, D. Kanouse, H. Kelley, R. Nisbett, S. Valins,& B. Weiner (Eds.), Attribution: Perceiving thecauses of behavior. Morristown, N.J.: GeneralLearning Press, 1972. (a)

Kelley, H. Causal schemata and the attribution pro-cess. In E. Jones, D. Kanouse, H. Kelley, R. Nis-bett, S. Valins, & B. Weiner (Eds.), Attribution:Perceiving the causes of behavior. Morristown,N.J.: General Learning Press, 1972. (b)

Lanzetta, J. T. The motivational properties of uncer-tainty. In H. I. Day, D. E. Berlyne, & D. E. Hunt(Eds.), Intrinsic motivation: New direction in edu-cation. Toronto, Canada: Holt, Rinehart, & Win-ston of Canada, 1971.

Levine, R., Chein, I., & Murphy, G. The relation ofthe intensity of a need to the amount of perceptualdistortion: A preliminary report. Journal of Psy-chology, 1942, 13, 283-293.

Messick, S. The standard problem—meaning andvalues in measurement and evaluation. AmericanPsychologist, 1975, 30, 955-966.

Neisser, U. Cognitive psychology. New York: Apple-ton-Century-Crofts, 1966.

Pepitone, A. Motivational effects in social perception.Human Relations, 1950, 3, 57-76.

Shaver, K. G. An introduction to attribution pro-cesses. Cambridge, Mass.: Winthrop Publishers,1975.

Snyder, M. The self-monitoring of expressive be-havior. Journal of Personality and Social Psychol-ogy, 1974, 30, 526-537.

Stephan, W., Berscheid, E., & Walster, E. Sexualarousal and heterosexual perception. Journal ofPersonality and Social Psychology, 1971, 20, 93-101.

Thibaut, J., & Kelley, H. The social psychology ofgroups. New York: Wiley, 1959.

(Received March 29, 1976)

Outcome Dependency: Attention, Attribution, and Attraction

Documents

Transcript of Outcome Dependency: Attention, Attribution, and Attraction