a comparison of aggressive rejected (zakriski and coie 1996)

24
A Comparison of Aggressive-Rejected and Nonaggressive-Rejected Children's Interpretations of Self-Directed and Other-Directed Rejection Audrey L. Zakriski and John D. Coie Duke University ZAKRISKI, AUDREY L., and COIE, JOHN D. A Comparison of Aggressive-Rejected and Nonaggres- sive-Rejected Children's Interpretations of Self-Directed and Other-Directed Rejection. CHILD DEVELOPMENT, 1996,67,1048-1070. The hypothesis that aggressive-rejected children are unaware of their social status because they are self-protective when processing negative peer feedback was tested in 3 studies. In Study 1, fourth-grade girls and boys were asked to name peers they liked or disliked, as well as peers they thought liked or disliked them. Gomparisons of aggressive- rejected, nonaggressive-rejected, and average status groups revealed that aggressive-rejected children were more unrealistic in their assessments of their social status than were nonaggres- sive-rejected children. In Study 2, rejected and average boys identified in Study 1 were asked to name who they thought liked or disliked other children from their classroom. Gomparisons of perceived and actual nominations for peers revealed that aggressive-rejected children were able to assess the social status of others as well as did nonaggressive-rejected and average status children. Because the difficulties aggressive-rejected children demonstrated in Study 1 did not generalize to judging the status of others in Study 2, the self-protective hypothesis was supported. Study 3 provided a parallel test of this hypothesis under more controlled conditions. Subjects from Study 2 viewed other children receiving rejection feedback from peers in videotaped inter- actions and received similar feedback themselves from experimental confederates. Whiie all subjects rated self-directed feedback somewhat more positively than other-directed feedback, aggressive-rejected subjects had the largest self-favoring discrepancy between their judgments of self- and other-directed feedback. Thesefindingsalso suggest that aggressive-rejected children may make self-protective "errors" when judging other children's negative feelings about them. Ethnicity differences in evaluating peer feedback emerged in Studies 1 and 3, raising questions about the impact of minority status on children's evaluations of rejection feedback. The search for the causes of childhood aggressive-rejection (Cole, Lochman, Terry, peer rejection has led to an important dis- & Hyman, 1992). Nonaggressive-rejection is tinction between children who seem to be less stable (Cillessen et al., 1992) and may rejected because of excessive aggressive- be more closely linked to depression (Boi- ness and those who are not overly aggressive vin, Poulin, & Vitaro, 1994; Hymel, Rubin, (French, 1988). Some of this latter group are Rowden, & LeMare, 1990; Rubin et al., thought to be socially withdrawn and anx- 1990). ious (Rubin, LeMare, & Lollis, 1990), but it is likely that there are several other reasons It is also becoming increasingly clear for rejection in this group. There is evidence that these subgroups of rejected children do suggesting these two different subgroups are not experience their rejection in the same at risk for different negative outcomes: Ag- way. For example, nonaggressive-rejected gression-related rejection is more stable children, defined as either high on submis- (Cillessen, van IJzendoorn, van Lieshout, & siveness or high on social isolation and shy- Hartup, 1992) and is more closely linked to ness, report greater feelings of loneliness, delinquency and school drop-out than non- have lower self-esteem, and are more likely This work was supported.by grants (nos. 1 F31 MH10393-01 and 383-4745-7324) from the National Institute of Mental Health and the Lowenstein Genter for Disruptive Behavior Disor- ders awarded to the first author. We are grateful to Kristen Thompson for her help with data collection and to Jack G. Wright for his helpful comments on an earlier version of this manuscript. Portions of this research were presented at the biennial meeting of the Society for Research in Ghild Development, New Orleans. Gorrespondence should be addressed to Audrey L. Zakriski, Bradley Hospital, 1011 Veterans Memorial Parkway, East Providence, RI 02915. [Child Development, 1996,67,1048-1070. © 1996 by the Society for Research in Child Development, Inc. All rights reserved. 0009-3920/96/6703-0008$01.001

Upload: jennwangster

Post on 13-Apr-2015

48 views

Category:

Documents


1 download

DESCRIPTION

Empirical article

TRANSCRIPT

Page 1: A Comparison of Aggressive Rejected (Zakriski and Coie 1996)

A Comparison of Aggressive-Rejected andNonaggressive-Rejected Children'sInterpretations of Self-Directed andOther-Directed Rejection

Audrey L. Zakriski and John D. CoieDuke University

ZAKRISKI, AUDREY L., and COIE, JOHN D. A Comparison of Aggressive-Rejected and Nonaggres-sive-Rejected Children's Interpretations of Self-Directed and Other-Directed Rejection. CHILDDEVELOPMENT, 1996,67,1048-1070. The hypothesis that aggressive-rejected children are unawareof their social status because they are self-protective when processing negative peer feedbackwas tested in 3 studies. In Study 1, fourth-grade girls and boys were asked to name peers theyliked or disliked, as well as peers they thought liked or disliked them. Gomparisons of aggressive-rejected, nonaggressive-rejected, and average status groups revealed that aggressive-rejectedchildren were more unrealistic in their assessments of their social status than were nonaggres-sive-rejected children. In Study 2, rejected and average boys identified in Study 1 were askedto name who they thought liked or disliked other children from their classroom. Gomparisonsof perceived and actual nominations for peers revealed that aggressive-rejected children wereable to assess the social status of others as well as did nonaggressive-rejected and average statuschildren. Because the difficulties aggressive-rejected children demonstrated in Study 1 did notgeneralize to judging the status of others in Study 2, the self-protective hypothesis was supported.Study 3 provided a parallel test of this hypothesis under more controlled conditions. Subjectsfrom Study 2 viewed other children receiving rejection feedback from peers in videotaped inter-actions and received similar feedback themselves from experimental confederates. Whiie allsubjects rated self-directed feedback somewhat more positively than other-directed feedback,aggressive-rejected subjects had the largest self-favoring discrepancy between their judgmentsof self- and other-directed feedback. These findings also suggest that aggressive-rejected childrenmay make self-protective "errors" when judging other children's negative feelings about them.Ethnicity differences in evaluating peer feedback emerged in Studies 1 and 3, raising questionsabout the impact of minority status on children's evaluations of rejection feedback.

The search for the causes of childhood aggressive-rejection (Cole, Lochman, Terry,peer rejection has led to an important dis- & Hyman, 1992). Nonaggressive-rejection istinction between children who seem to be less stable (Cillessen et al., 1992) and mayrejected because of excessive aggressive- be more closely linked to depression (Boi-ness and those who are not overly aggressive vin, Poulin, & Vitaro, 1994; Hymel, Rubin,(French, 1988). Some of this latter group are Rowden, & LeMare, 1990; Rubin et al.,thought to be socially withdrawn and anx- 1990).ious (Rubin, LeMare, & Lollis, 1990), but itis likely that there are several other reasons It is also becoming increasingly clearfor rejection in this group. There is evidence that these subgroups of rejected children dosuggesting these two different subgroups are not experience their rejection in the sameat risk for different negative outcomes: Ag- way. For example, nonaggressive-rejectedgression-related rejection is more stable children, defined as either high on submis-(Cillessen, van IJzendoorn, van Lieshout, & siveness or high on social isolation and shy-Hartup, 1992) and is more closely linked to ness, report greater feelings of loneliness,delinquency and school drop-out than non- have lower self-esteem, and are more likely

This work was supported.by grants (nos. 1 F31 MH10393-01 and 383-4745-7324) from theNational Institute of Mental Health and the Lowenstein Genter for Disruptive Behavior Disor-ders awarded to the first author. We are grateful to Kristen Thompson for her help with datacollection and to Jack G. Wright for his helpful comments on an earlier version of this manuscript.Portions of this research were presented at the biennial meeting of the Society for Research inGhild Development, New Orleans. Gorrespondence should be addressed to Audrey L. Zakriski,Bradley Hospital, 1011 Veterans Memorial Parkway, East Providence, RI 02915.

[Child Development, 1996,67,1048-1070. © 1996 by the Society for Research in Child Development, Inc.All rights reserved. 0009-3920/96/6703-0008$01.001

Page 2: A Comparison of Aggressive Rejected (Zakriski and Coie 1996)

than average status children or aggressive-rejected children to refer themselves forhelp with their peer relations. Aggressive-rejected children, on the other hand, do notdiffer from average status children in self-reported loneliness. They report at least av-erage levels of self-esteem and are unlikelyto refer themselves for help with their peerrelations (Asher, Zelis, Parker, & Bruene,1991; Boivin, Thomassin, & Alain, 1989;Boivin, Vitaro, & Tremblay, 1989; Parkhurst& Asher, 1992). Aggressive-rejected chil-dren also rate themselves higher than aver-age status children on more self-concept di-mensions than do nonaggressive-rejectedchildren, defined either as high on passivewithdrawal and shyness or simply as nonag-gressive (Boivin, Thomassin, & Alain, 1989;Hymel, Bowker, & Woody, 1993; Patterson,Kupersmidt, & Griesler, 1990). This pattemof findings suggests that aggressive- andnonaggressive-rejected children might dif-fer in their awareness of their own rejec-tion. Aggressive-rejected children appearto be unaware of the extent to whichthey are rejected by their peers, whereasnonaggressive-rejected children appear tobe quite aware of the extent to which theyare rejected.

This proposed difference in statusawareness is important because it providesa potential link to the outcome differencesassociated with aggressive- and nonaggres-sive-rejection. One developmental model ofthe connection between nonaggressive peerrejection and internalizing problems in-volves the mediating role of self-concept andself-esteem (Rubin et al., 1990). This modelsuggests that awareness of negative peer sta-tus is a necessary condition for lowered self-esteem. The greater risk of extemalizingproblems among aggressive-rejected chil-dren and their greater tendency to be chroni-cally rejected than nonaggressive-rejectedchildren may be related to their proposedstatus awareness deficits. Lack of awarenessof their peer rejection would make thesechildren less likely to attempt to correcttheir own behavior, thus leading them fe re-peat the maladaptive interactions that origi-nally caused, them to be disliked by peers.This same social insensitivity may renderthem less likely to benefit from social skillsinterventions because they are not moti-vated to change.

Indirect evidence for rejected subgroupdifferences in social status awareness hasbeen obtained by researchers utilizing thesocial acceptance subscale of the Perceived

Zakriski and Coie 1049

Competence Scale (Harter, 1982) and thepeer relations subscale of the Self-Description Questionnaire (Marsh, Smith, &Barnes, 1983). The results of several of thesestudies suggest that aggressive-rejected chil-dren overestimate their social competencewhile nonaggressive-rejected children donot (Boivin, Cote, & Dion, 1991; Boivin, Vi-taro, & Tremblay, 1989; Hymel et al., 1993).Although it bears on the question of sub-group differences in self-perceptions of peerstatus, this evidence can only be consideredindirect for two reasons. First, these per-ceived social competence subscales con-found the assessment of classroom statusawareness with loneliness and subjects' gen-eral beliefs about their own likability. Also,they do not permit a precise comparison be-tween perceived social status and actual so-cial status, because sociometric measuresnot only have different content than thesesubscales, but utilize a different metric intheir response format. More direct evidenceof subgroup differences in social statusawareness comes from a study by Cillessenet al. (1992) in which subgroups of rejectedboys were asked to rate how much theythought they were liked by peers. Theyfound that as a group, aggressive-rejectedboys expected themselves to be less rejectedthan shy-rejected boys, when in fact theywere more rejected.

Taken together, these studies supportthe hypothesis that aggressive-rejected andnonaggressive-rejected children, particu-larly withdrawn-rejected children, differ intheir levels of rejection awareness, with non-aggressive-rejected children being moreself-aware than aggressive-rejected chil-dren. The three studies described in thisarticle conduct a more precise test of thishypothesis and explore several possible ex-planations for such group differences. In thefirst study, subgroups of rejected childrenare compared to each other and to averagestatus children on the accuracy of their self-perceptions of peer status. In contrast to thestudy by Cillessen et al. (1992), in whichcomparisons were assessed only on peer lik-ing measures, measures of peer liking anddisliking were included and accuracy wasassessed at two levels. The first level wasgroup-based expectations, the procedureemployed by Cillessen et al., and involvedcomparing subject groups' niean estimates ofthe number of peers who liked or dislikedthem. The second level compared individ-ual subjects' expected liking scores withtheir actual liking scores through the use of

Page 3: A Comparison of Aggressive Rejected (Zakriski and Coie 1996)

1050 Child Development

a difference score. This procedure providesan accuracy score for each individual subjectand addresses our hypothesis more directly.

In the second study, subgroups werecompared on the accuracy of their assess-ments of the social status of classmates. Asubset of specific peers was identified as thetarget of these judgments, and accuracy instatus attributions was contrasted across re-jected subgroups and average status chil-dren. If subgroup differences in the accuracyof perceptions of other children's peer statusmatched differences in the accuracy of self-perceptions, then a case could be made forgeneralized deficits in the reading of socialcues about liking or disliking.

The third study was designed to assesschildren's accuracy in reading actual peerfeedback about liking and disliking. Sub-jects first observed videotapes of childrenplaying a game together and then madejudgments as to how much each liked or dis-liked the other, based on comments theyheard and interactions they observed. Next,they played the same game with a confeder-ate who expressed feedback identical to thefeedback seen on the videotape. Subjectswere then asked to rate how much the con-federate liked them as a way of comparingsubjects' abilities to read cues directed atthemselves with their abilities to read thesame cues directed at other people.

Because it is important not to assumethat children of different ethnic backgroundsjudge information about their social worldsin the same way, both African-American andCaucasian subjects were included in this se-ries of studies. Previous research on loneli-ness, social goals and expectations, self-concept, and status awareness did notprovide us with any specific hypothesesabout ethnicity differences in classroom sta-tus awareness. Research on sociometricnominating patterns in mixed-ethnicity ele-mentary school settings suggests that chil-dren in the minority are less favorably evalu-ated (Singleton & Asher, 1977), and thatsocial preference ratings in these situationsmay be influenced by ethnic group prefer-ence as well as social behavior (Kupersmidt& Coie, 1990; Lochman & Wayland, 1994).This may result in ostensibly less accuratestatus awareness among minority children inthese settings, because the task of assessingwho likes them is a more complicated oneinvolving the assessment of both actual lik-ing/disliking and hidden prejudices.

Study 1: Assessment of SelfMETHOD

SubjectsParent permission to participate in

classroom sociometrics was solicited from826 children in the fourth grade at eight pub-lic elementary schools in a southeasternschool system serving urban and suburbanareas. Children of this age were selected forseveral reasons. Krantz and Burton (1985)found that perceptions of peers' status werefairly accurate by first grade and remainedstable through third grade, while accurateself-perceptions of status were delayed untilthird grade. Because of this developmentalpattern, we chose to study children in thesecond half of their fourth-grade year to en-sure that the majority of our subjects wouldbe able to perform the task. Second, by theage of 9 or 10, peer social status appears tobe relatively stable (Coie & Dodge, 1983).This is important because children wereasked to assess their own social status, andasking them to assess an unstable parametermight result in apparent inaccuracy that wasnot necessarily due to social judgmentdeficits.

Written parental consent was receivedfor 621 children (75% consent rate). Thesample was 58% Caucasian, 37% African-American, and 5% of other ethnic origin. Be-cause of their small numbers, we eliminatedchildren as subjects who were not classifiedas African-American or Caucasian. However,we included these children's nominationswhen calculating social status and aggres-sion scores because the children were partof the peer group. Fifty-two percent of thisremaining sample of 591 were female and48% were male. The average age ofthe sub-jects was 10.3 years (SD = .56). Data on pa-rental SES and occupation were not col-lected on this sample. However, datacollected on an older, but otherwise compa-rable, sample of 114 aggressive and nonag-gressive boys from the same school systemrevealed that subjects were primarily fromlower-middle- to middle-class homes withthe following distribution of parental occu-pations: professional or managerial jobs(29%); semiprofessional (29%); and skilled,semiskilled, or unskilled workers (42%)(Lochman & Wayland, 1994).

ProcedureSociometric questionnaires were group

administered in 33 classrooms late in thespring semester. Children received a packet

Page 4: A Comparison of Aggressive Rejected (Zakriski and Coie 1996)

Zakriski and Coie 1051

containing roster lists of all children in theirgrade. They also received a questionnaireincluding actual status, perceived status, andaggression items, as well as four other socialbehavior and social network items. Two re-search assistants administered the sociomet-ric measures. Afler an item was read to theclass, children were asked to check thenames of all students who fit this descriptionand then to circle the names of three chil-dren who best fit the description. Childrenvoted first on all peers in their classroom andthen on the rest of the peers in their grade.

MeasuresIn addition to the questions "Who do

you like most?" (actual liking) and "Who doyou like least?" (actual disliking) (Coie,Dodge, & Coppotelli, 1982), children wereasked to answer the questions "Who reallylikes you?" (perceived liking) and "Who re-ally does not like you?" (perceived dislik-ing) (Zakriski, Coie, & Wright, 1992). Tomeasure aggression, subjects were asked toname the children in their grade who "startfights, pick on other kids and tease them."Withdrawal was measured using the item"Who stays by themselves and away fromother kids?" Both limited and unlimitednominations were collected on all items.

Scoring of measures.—Unlimited nomi-nations for actual and perceived liking anddisliking within the classroom^ were used tocalculate the two measures of status aware-ness. Group-based expectation scores werecalculated by dividing the number of per-ceived liking or disliking nominations sub-jects expected to receive by the number ofvoters in each class to adjust for differencesin class size and student participation. Toexamine accuracy at the individual level, in-dividual accuracy scores were calculated bydividing the number of actual and perceivedliking or disliking nominations a subject re-ceived by the number of voters in each class.Then these actual liking or disliking scoreswere subtiacted from the perceived likingor disliking scores, respectively. Individualaccuracy scores therefore could be positiveor negative, with positive scores indicating

an overestimation of peer nominations, neg-ative scores indicating an underestimation,and scores near the zero point indicating ac-curate assessment.

Categorization of subjects.—Limitednominations were used to form social statusgroups as follows: Liking and disliking nom-inations were summed for each child overall voters in their grade. These sums werethen standardized within each school. Socialpreference was then calculated by sub-tracting standardized disliking scores fromstandardized liking scores, and social impactwas calculated by adding the liking and dis-liking scores. Both social preference and so-cial impact scores were then restandardizedwithin school (Coie et al., 1982). Childrenwho received a social preference z score lessthan —.8, a liking score less than 0, and adisliking score greater than 0 were labeledrejected. A .8 cutoff criterion was used toobtain adequate numbers of rejected chil-dren who could be subdivided according toaggressiveness in this and subsequent stud-ies. Our use of .8 SD rather than 1 SD asa rejection cutoff is not unprecedented (seeHymel et al., 1993, for a review). In fact, thedecision to use a more lenient rejection cut-off is supported by research on interventionselection criteria which suggests that the useof a more lenient rejection cutoff better iden-tified those rejected children who developadjustment problems in the long run (Terry& Coie, 1990). According to the procedureused by Kupersmidt and Coie (1990), chil-dren who did not meet criteria for inclusionin the other four status groups using .8 SDcutoffs were categorized as having averagestatus. Average subjects were included inthis study as a standard against which theaccuracy ofthe rejected subgroups could becompared (Hymel et al., 1993).

Aggression scores were calculated foreach child by first summing nominations forthe limited "starts fights" item over all vot-ers in that school and then standardizingwithin school (Patterson et al., 1990) andgender (Coie et al., 1992). To qualify as ag-gressive, subjects had to receive a z score of

' Although actual and perceived nominations for this study were collected from students inthe whole fourth grade, perceived nominations in Study 2 were only measured within class. Toallow for comparability across studies, only within-class nominations were used to calculateaccuracy. Replication of the Study 1 analyses using actual and perceived nominations over theentire grade revealed the same pattem of results. Our choice to focus on within-class nominationsis also justified by the following reasons. First, inspection of the data revealed that childrenvoted predominantly for classmates. Second, because elementary school children spend most oftheir time in school with their classmates, within-class perceptions were likely to be more stableand reliable than grade-wide nominations.

Page 5: A Comparison of Aggressive Rejected (Zakriski and Coie 1996)

1052 Child Development

at least -I- .5 on their standardized aggressionscore.^ To qualify as nonaggressive, subjectshad to receive an aggression z score of lessthan 0. When the withdrawal nominationswere tabulated, the totals yielded a badlyskewed distribution owing to the fact thatsubjects named very few of their peers asfitting this description. Because of this, wedid not use the withdrawal measure tosubdivide the rejected group. Younger,Schwartzman, and Ledingham (1985) havereported related problems in the use of with-drawn behavior nomination procedures withchildren of this age.

Sociometric classifications yielded 122rejected children (70 boys and 52 girls) and328 average status children (154 boys and174 girls). Although the names of all thesechildren were included on the nominationrosters, not all of these children were pres-ent when the nomination data were col-lected. A total of 80 rejected (46 boys and 34girls) and 253 average (115 boys and 138girls) children participated in the sociomet-ric administration and therefore had com-plete data on both perceived and actual so-cial status. The rejected subjects wereclassified into aggressive (AR) and nonag-gressive (NR) subgroups, yielding 26 aggres-sive-rejected subjects (14 boys and 12 girls)and 43 nonaggressive-rejected subjects (24boys and 19 girls). Eleven rejected childrenwere categorized as neither aggressive nornonaggressive and their data were not in-cluded in the analyses.

Data AnalysisSubjects' group-based expectations of

acceptance or rejection and individual accu-racy scores were compared in four separate3 (status group) X 2 (ethnicity) x 2 (gender)ANOVAs, using the Ceneral Linear Modelsframework. Because gender did not produceany significant main effects or interactions,we collapsed over this variable. Only whenthe overall model was significant were sig-nificant main effects and interactions exam-ined. Because comparisons of status groupswere planned and involved only threegroups. Fisher's LSD (Maxwell & Delaney,

1990) was used to test for differences be-tween group means.

RESULTS

Preliminary analyses revealed that thetwo rejected subgroups did not differ in ac-tual social acceptance or social rejection. A3 (status group) x 2 (ethnicity) ANOVA onsocial acceptance revealed the expectedmain effect for status group, F(2, 319) =19.28, p < .0001. However, comparisons ofstatus group means revealed no differencesin social acceptance between the rejectedsubgroups (MAR = .18, MMR = .21). Bothgroups received fewer "like most" nomina-tions than did average status children (MAV= .31). There was no difference between Af-rican-American and Caucasian subjects inthe number of "like most" nominations theyreceived. Similarly, an ANOVA on within-class social rejection revealed fhe expectedmain effect for stattis group, F(2, 319) =48.13, p < .0001, but comparisons of groupmeans revealed that both subgroups of re-jected children received significantly more"like least" nominations than did averagechildren (MAR = .42, M^R = .37, MAV = -21).There was no difference between African-American and Caucasian subjects in thenumber of "like least" nominations they re-ceived. The fact that rejected subgroups didnot differ on measures of actual social accep-tance or rejection indicates that potentialsubgroup differences in subjects' individualaccuracy scores are unlikely to be the resultof a confound with actual popularity. Sub-jects who are more disliked, for example,have a greater possibility of underestimatingtheir rejection, whereas those who are lessdisliked have more opportunity to overesti-mate their rejection. Thus, accuracy can beconfounded by base rates (Gage & Cron-bach, 1955; Zakriski et al., 1992). By this rea-soning, accuracy differences between eitherof the rejected subgroups and the averagegroup must be interpreted with caution,since it is possible that such a differencecould be explained by this base-rate con-found.

^ The aggression cutoff was set at .5 SD in order to be consistent across studies. In Studies2 and 3, we needed to use a more lenient aggression cutoff to obtain roughly equal numbers ofCaucasian and African-American subjects in the aggressive-rejected group. With a more stringentaggression cutoff in those studies, the aggressive-rejected group becomes unbalanced, with moreAfrican-American subjects than Caucasian subjects. This creates a confound of rejection subgroupand ethnicity, making interpretation of the results problematic. To test the consequences ofusing this more inclusive criterion for aggressiveness, we repeated all of the Study 1 analyseswith subject groups defined by a 1 SD aggression cutoff These analyses revealed the samepattern of findings as were obtained with the groups defined by the .5 SD criterion.

Page 6: A Comparison of Aggressive Rejected (Zakriski and Coie 1996)

Zakriski and Coie 1053

Group-based Comparisons of ExpectedAcceptance and Rejection

T'he ANOVA on group-based expecta-tions of social acceptance (liking) revealedonly a main effect for ethnicity, F(l, 316) =7.73, p < .01, with African-American sub-jects expecting more liking nominations (M= .28) than Caucasian subjects (M = .21).As shown in Table 1, the number of ex-pected "like most" nominations did not dif-fer across status groups. This means thatboth subgroups of rejected children ex-pected an unrealistic number of "like most"nominations, but did not differ from eachother in this respect.

The ANOVA on group-based expecta-tions of social rejection revealed a marginalmain effect for status group, F(2, 316) =2.72, p < .07. As shown in Table 1, compari-sons of status group means revealed thatnonaggressive-rejected children expectedmore "like least" nominations than did aver-age status children, F(l, 316) = 6.87, p <.01, whereas aggressive-rejected childrendid not differ from average children, F < 1.A main effect for ethnicity, F(l, 316) = 6.54,p < .01, indicated that African-Americansubjects (M = .15) expected fewer "likeleast" nominations than did Caucasian sub-jects (M = .21).

Individual AccuracyThe ANOVA on individual accuracy

scores for social acceptance revealed a maineffect for status group, F(2, 316) = 7.29, p <.001. As seen in Table 1, comparisons of sta-tus group means indicated that both aggres-sive-rejected, F(l, 316) = 6.33, p < .01, andnonaggressive-rejected children, F(l, 316)= 7.15, p < .01, overestimated their socialacceptance more than average status chil-dren. The two rejected subgroups, however,did not differ, F(l, 316) < 1. In absoluteterms, both rejected groups expected to re-

ceive as many "like most" nominations asthey actually received. This was refiected bythe fact that their accuracy scores were nearzero. Average status children, on the otherhand, expected fewer "like most" votes thanthey received, as refiected by their negativeaccuracy scores. There was also a main effectfor ethnicity, F(l, 316) = 3.91, p < .05, indi-cating that Caucasian subjects (M = — .07)were more likely to underestimate their so-cial acceptance than were African-Americansubjects (M = - .02). The ANOVA on indi-vidueJ accuracy scores for social rejectionalso revealed a main effect for status group,F(2, 316) = 14.11, p < .0001. The means inTable 1 refiect the fact that aggressive-rejected children underestimated how muchthey were disliked significantly more thanboth average children, F(l, 316) = 21.56, p< .0001, and nonaggressive-rejected chil-dren, F(l, 316) = 6.10, p < .01. Nonaggres-sive-rejected children underestimated theirsocial rejection only more than average chil-dren, F(l, 316) = 4.21, p < .04. In abso-lute terms, aggressive-rejected children ex-pected many fewer "like least" nominationsthan they received, nonaggressive-rejectedchildren expected slightly fewer "like least"nominations than they received, and averagestatus children expected to receive approxi-mately the same number of "like least" nom-inations as they received. A main effect forethnicity, F(l, 316) = 12.39, p < .001, indi-cated that African-American subjects (M =— .12) underestimated their social rejectionmore than Caucasian subjects (M = - .03).

Clarification ofEthniciiy EffectsBecause in mixed ethnicity school set-

tings, children have been shown to nomi-nate more same-ethnicity peers as "likedmost" than cross-ethnicity peers, and morecross-ethnicity peers as "liked least" thansame-ethnicity peers (Singleton & Asher,

TABLE 1

MEAN STATUS EXPECTATIONS AND ACCURACY AS A FUNCTION OF STATUS SUBGROUP

GROUP-BASED

EXPECTATIONS INDIVIDUAL ACCURACY

STATUS GROUP N Acceptance Rejeetion Acceptance Rejection

Aggressive-Rejected 26 .22a .17ab .04a -.24aNonaggressive-Rejected 43 .24a .25a .02a - . l i bAverage 253 .24a .18b -.08b -.04c

NOTE.—Means with different letters within each column are significantly different at p < .05. Positive individualaccuracy scores indicate overestimation, negative scores indicate underestimation, and scores near zero indicateaccurate assessments.

Page 7: A Comparison of Aggressive Rejected (Zakriski and Coie 1996)

1054 Child Development

1977), we were concerned that our African-American subjects might have been at a dis-advantage when judging their social statusbecause they were in the minority (at least60% Caucasian and 40% African-American)in 31 ofthe 33 classrooms in this study. Bi-ased cross-ethnicity voting patterns or diffi-culties in assessing cross-ethnicity accep-tance and rejection therefore would have amuch greater impact on African-Americansubjects' accuracy than on Caucasian sub-jects' accuracy. To assess the impact of same-and cross-ethnicity judgments on social sta-tus perceptions, we conducted a 2 (ethnicityof subject) X 2 (gender) x 2 (ethnicity ofvoter: same-ethnicity vs. cross-ethnicity)ANOVA on individual accuracy for both so-cial acceptance and social rejection with eth-nicity of voter as a repeated measure. Forsocial acceptance, we found a main effect forethnicity of subject, F(l, 588) = 6.16, p <.01, indicating that Caucasian subjects un-derestimated their social acceptance morethan did African-American subjects. We alsofound a significant ethnicity of subject xethnicity of voter interaction, F(l, 588) =22.74, p < .001, indicating that African-American (M = — .03) and Caucasian (M =.01) subjects were not different in their judg-ments of same-ethnicity social acceptance.That is, both ethnic groups were equally ac-curate in judging how much same-ethnicitypeers liked them. However, African-American and Caucasian subjects divergedin accuracy when judging their cross-ethnicity social acceptance: Caucasian sub-jects underestimated the number of "likemost" nominations they would receive fromAfrican-American peers (M = — .24), and Af-rican-American subjects overestimated thenumber of "like most" nominations theywould receive from Caucasian peers (M =.23). Similar results were found for social re-jection. We found the expected main effectfor ethnicity, F(l, 588) = 4.23, p < .04. Wealso found a significant ethnicity of subjectX ethnicity of voter interaction, F(l, 588) =12.52, p < .001. This interaction indicatedthat African-American (M = .01) and Cauca-sian (M = .01) subjects did not differ in accu-racy when judging their same-ethnicity so-cial rejection, but differed when judgingtheir cross-ethnicity social rejection. Cauca-sian subjects overestimated the number of"like least" nominations they would receivefrom African-American peers (M = .17), andAfrican-American subjects underestimatedthe nurriber of "like least" nominations theywould receive from Caucasian peers (M =-.22).

DISCUSSION

These results support previous researchfindings suggesting that aggressive-rejectedchildren overestimate their peer social sta-tus, whereas nonaggressive-rejected chil-dren more accurately assess their social sta-tus. Furthermore, the results suggest that itis social rejection that most clearly distin-guishes aggressive-rejected from nonag-gressive-rejected children in the statusperception domain. Nonaggressive-rejectedchildren appropriately expected more "likeleast" nominations than did average statuschildren, whereas aggressive-rejected chil-dren did not. Consequently, aggressive-rejected children underestimated theirsocial rejection more than nonaggres-sive-rejected and average status children.Aggressive-rejected children also underesti-mated their social rejection in absoluteterms, as evidenced by their large negativeaccuracy scores. In contrast, nonaggressive-rejected and average status children had ac-curacy scores closer to zero. That is, theywere highly accurate.

Both rejected groups unrealistically ex-pected the same number of "like most" nom-inations as average status children, and con-sequently both groups overestimated theirsocial acceptance more than average statuschildren. For social preference, both re-jected groups were accurate in absoluteterms, whereas average status children un-derestimated their social preference, as evi-denced by their negative accuracy score.

These rejected subgroup findings pro-vide one possible explanation for some ofthe differences reported in the literaturebetween aggressive and nonaggressive-rejected children. If aggressive-rejectedchildren believe that they are disliked lessthan they really are, they might very wellfeel less lonely (Parkhurst & Asher, 1992)and feel less need for help (Asher et al.,1991). This also may explain why aggres-sive-rejected children would be more apt toexpress a desire for continued social interac-tion with peers (Parkhurst & Asher, 1992;Rabiner & Gordon, 1993) and hold generallypositive beliefs about them (Rabiner &Keane, 1991), whereas nonaggressive-rejected children would not.

In addition to status subgroup differ-ences in children's abilities to judge theirsocial status, we found ethnicity differencesin children's judgments of their social status.When asked to assess classroom social sta-tus, African-American subjects expected

Page 8: A Comparison of Aggressive Rejected (Zakriski and Coie 1996)

more "like most" nominations than Cauca-sian subjects, and fewer "like least" nomina-tions than Caucasian subjects. Analyses ofactual liking and disliking by peers revealedno differences between ethnic groups forthese variables. Consequently, African-American subjects underestimated their so-cial acceptance less than Caucasian subjectsand underestimated their social rejectionmore than Caucasian subjects.

Interpretation of these ethnicity differ-ences was complicated by the fact that 88%of our African-American subjects were in aminority in their classrooms. It is possiblethat children in general have less access toinformation about cross-ethnicity peers' so-cial preferences than same-ethnicity peers'social preferences. If so, minority status chil-dren would be at a disadvantage when askedto judge who liked or did not like them. Ad-ditional analyses supported this viewpointand suggested that African-American sub-jects' inaccuracy in the general analyseswas infiuenced by their minority status.These analyses demonstrated that African-American and Caucasian children did notdiffer in their judgments of same-ethnicitypeers. Both were inaccurate, however, whenjudging cross-ethnicity peers. Caucasiansubjects were pessimistic about how wellthey were regarded by African-Americanpeers, whereas African-American subjectswere optimistic about how well they wereregarded by Caucasian peers. This cross-ethnicity inaccuracy translated into overallinaccuracy for the African-American subjectsbecause they had to estimate the likes anddislikes of many more cross-ethnicity peersthan same-ethnicity peers.

Study 2: Assessment of OthersThe results of Study 1 support the hy-

pothesis that aggressive-rejected childrengenerally are less accurate in their judg-ments of peer dislike than nonaggressive-rejected children. This raises questionsabout the reason for this difference. Onepossible explanation is that aggressive-rejected children inaccurately perceive theirsocial status due to a general social informa-tion processing deficit. There is a substantialbody of research describing social informa-tion processing difficulties in aggressive andaggressive-rejected children (Dodge, Pettit,McClaskey, & Brown, 1986; Lochman,Myer, Rabiner, & White, 1991). However,that literature does not appear to explain thephenomenon described in this article. An-other possible explanation is that aggressive-

Zakriski and Coie 1055

rejected children possess specific difficultiesin processing social rejection cues onlywhen the self is involved. That is, theirdeficits in status awareness might be moreheavily infiuenced by defensive and self-protective mechanisms than by general in-formation processing difficulties. This expla-nation is supported by findings in othersocial judgment tasks, such as children's per-ceptions of intent in ambiguous situations,that biased perceptions are more pro-nounced when the situation involves the self(Dodge & Somberg, 1987). It is also consis-tent with the "ego-defensiveness" hypothe-sis offered by Asher and colleagues to ex-plain why aggressive-rejected childrenmight report less loneliness (Asher, Park-hurst, Hymel, & Williams, 1990), and issupported by recent research linkingaggressive-rejection with narcissistic charac-teristics (Bukowski, Sippola, Verlan, & New-comb, 1993).

The purpose of Study 2 was to deter-mine whether the subgroup differences ob-served in Study 1 were specific only to self-perceptions of social status, or whether thesedifferences extend to perceptions of otherchildren's peer status. All of the rejectedchildren identified in Study 1, and an equalnumber of average status children, wereasked to participate in Study 2. Subjects inthis study were asked to report on the statusof several classroom peers. The goal was toassess whether aggressive-rejected chil-dren's status awareness difficulties are dueto general social information processingproblems or self-specific processing issues.If aggressive-rejected subjects are able to ac-curately report on how well other childrenare liked or disliked, it is likely that theirown poor status awareness is the result of amore self-specific bias in reading or re-porting on social status. If instead, aggres-sive-rejected children inaccurately report onthe status of their peers, their own poor sta-tus awareness may be the result of a moregeneral social information processingproblem.

METHOD

SubjectsThe subjects for this study were 55 boys

identified in Study 1, mean age = 10.35years (SD = .43). Parent permission to par-ticipate in the present study was solicitedfrom all children who could be classified associometrically rejected (n = 117) and anequal number of randomly selected averagestatus children whose aggression scores fell

Page 9: A Comparison of Aggressive Rejected (Zakriski and Coie 1996)

1056 Child Development

within .5 SD of their same-gender standardscore for aggression. Out ofthe 234 potentialsubjects, written parental consent was re-ceived for 155 children (66%), 46 parents de-clined permission (20%), and 33 parentsfailed to respond after several attempts atcontact (14%). Of the 155 children whoseparents consented, 111 children (78%) par-ticipated in this study. The 34 children withparental consent who did not participatewere unavailable during the data collectionperiod.

Ofthe 111 children who participated inthis study, 56 were girls. The data for girlsare not included in this report because therewas an insufficient number of aggressive-rejected girls to perform the analyses testingthe main hypothesis of this study. When ag-gression was standardized within gender,there were only 16 aggressive-rejected fe-males in our population of 826 children(1.9%), and we were able to recruit five ofthem for this study. These rates are consis-tent with other studies of aggressive-rejected children: In a sample of 663 chil-dren, Patterson et al. (1990) only identifiedsix aggressive-rejected girls, and in a sampleof 112, Kupersmidt and Coie (1990) onlyidentified one aggressive-rejected girl.

The recruitment procedures describedabove yielded the following groups of sub-jects: 18 aggressive-rejected males (8 Cauca-sian and 10 African-American); 16 non-aggressive-rejected males (9 Caucasianand 7 African-American); 22 average males(11 Caucasian and 11 African-American).Twelve of the recruited rejected males hadaggression scores that fell between our cut-offs for the aggressive and nonaggressivegroups and thus were excluded from theanalyses.

To test for differences between partici-pating and nonparticipating rejected boys,we performed ANOVAs on our two main in-dependent variables: aggression and socialpreference. Neither variable showed differ-ences between participating and nonpartici-pating rejected boys, F(l, 54) = 1.39, p <.24 and F(l, 54) = 1.73, p < .19.

MeasuresUsing a modified sociometric proce-

dure, subjects were interviewed about thesocial status of selected peers in their class-room to assess their knowledge of other chil-dren's peer relations. Subjects were asked toname who they think likes/does not likeeach of several peers in their classroom. Weattempted to randomly select two popular.

two average, and two rejected children fromeach subject's classroom to serve as targetchildren for this task. However, in 10 classesthere were fewer than two popular children,and in one class there were fewer than tworejected children. Therefore, subjects madejudgments on lists of four to six children.One African-American aggressive-rejectedsubject was excluded from the analyses be-cause he was the only rejected child in hisclass, and therefore could not assess the sta-tus of a rejected other.

Subjects made unlimited judgments ofliking and disliking about each of the targetchildren. These judgments were then usedto compute individual accuracy scores asdiscussed in Study 1. Because of time limita-tions, perceptions of other children's statuswere only assessed within class. For this rea-son, only within-class actual nominationswere used for accuracy comparisons. Whensubjects judged more than one child of agiven status group, their individual accuracyscores were averaged across the two same-status targets. As in Study 1, two scores werecomputed for individual accuracy: one forsocial rejection, and one for social accep-tance.

ProcedureIn the early summer following fourth

grade, subjects were interviewed individu-ally in our laboratory by one of four experi-menters who were blind to the status ofthesubject. Subjects were told that we were in-terested in learning about children's friend-ships and how much children know aboutwho likes whom in their classroom. Subjectswere reminded ofthe sociometric they com-pleted in the spring and were given tiie fol-lowing instructions: "I am going to read yousome names of children in your class [targetchildren] and I want you to tell me who youthink voted for these kids as someone theyliked the most or someone they liked theleast." After each target child's name wasread, the interviewer went through the listof classmates and had the subject indicatethe ones who would have named the targetas someone they liked most. A similar proce-dure was then followed to identify class-mates thought to have named the target assomeone they liked least. The order of pre-sentation of target children from differentstatus groups was fully counterbalanced.Each subject was told by the experimenterthat their answers would be kept confiden-tial. They were also asked not to discusstheir choices with other subjects or withpeers in their classroom.

tt IMMII .fi,w

Page 10: A Comparison of Aggressive Rejected (Zakriski and Coie 1996)

Zakriski and Coie 1057

RESULTS AND DISCUSSION

Accuracy Differences in the Perceptions ofOther Children's Social Status

Two 3 (social status of subject) x 2 (eth-nicity) X 3 (status of other) ANOVAs, withstatus of other as a repeated measure, wereconducted using the framework of GeneralLinear Models. The ANOVA on accuracy forsocial acceptance revealed a significant sta-tus of other main effect, F(2, 98) = 11.61, p< .0001, suggesting that subjects were gen-erally better able to assess the social accep-tance of rejected others than of average andpopular status others (see Table 2). How-ever, a status of subject X status of otherinteraction, F(4, 98) = 2.63, p < .04, sug-gested that this was not equally true for allsubjects. Tests of contrasts revealed that av-erage status subjects showed greater accu-racy when judging rejected others thanwhen judging average or popular others,compared to aggressive-rejected subjectswhose accuracy assessments of all others didnot differ. Inspection of the means in Table2 reveals that aggressive-rejected subjects,whose accuracy scores are near zero, werequite accurate in the assessment ofthe statusof all types of others. A significant status ofother X status of subject X ethnicity interac-tion revealed a more dramatic sbift in accu-racy scores over the target children for Afri-can-American average subjects than forCaucasian average subjects. Caucasian sub-jects had better accuracy scores when assess-ing the social acceptance of rejected others(M = —.02) than they did when assessingpopular (M = — .13) or average status others(M = -.19). Recall that negative scores re-flect an underestimation of the number ofnominations a target received, positivescores reflect an overestimation, and scoresnear the zero point refiect accurate assess-ment. African-American subjects had higheraccuracy scores when assessing rejected oth-ers (M = .16) than they did when assessingpopular (M = —.18) or average others (M =— .06), but these scores reflected a differenttype of inaccuracy. Whereas their negativescores indicated underestimation of popularand average children's social acceptance,their positive score indicated overestimationof rejected children's social acceptance.

In addition to these significant effects, amarginal overall status of subject main ef-fect, F(2, 49) = 2.63, p < .08, was exploredto better understand the accuracy differ-ences between our subject groups. Univari-ate tests for rejected, popular, and averageothers revealed that this group difference

was only significant in the assessment ofpopular others, F(2, 49) = 4.43, p < .02. Ag-gressive-rejected boys more accurately as-sessed the social acceptance of popular chil-dren than did either average boys, F(l, 49)= 7.08, p < .01, or nonaggressive-rejectedboys, F(l, 49) = 5.36, p < .02. There was asimilar trend in the assessment of averageothers, F(2, 49) = 2.42, p < .10, with aggres-sive-rejected boys being more likely to accu-rately assess the social acceptance of averageothers than were average subjects, F(l, 49)= 4.92, p < .03. For rejected others, theoverall model was not significant, F(5,49) =1.36, p < .26. Together these findings sug-gest that rather than being less accurate inthe assessment of others, aggressive-rejectedboys were more consistently accurate intheir assessments of other children's socialacceptance than were some other subjects,and were at times more accurate in their per-ceptions of other children's social accep-tance than were average and nonaggressive-rejected boys.

We next performed the same ANOVA onaccuracy of social rejection judgments.There was a status of other main effect, F(2,98) = 7.27, p < .002, suggesting that subjectswere better able to assess the social rejectionof average and popular children than that ofrejected children (see Table 2). There was amarginal status of other x status of subjectinteraction, F(4, 98) = 2.27, p < .07, whichqualified this main effect. Tests of contrastsrevealed that average children had morevariability in their accuracy scores, com-pared to aggressive-rejected boys whose ac-curacy scores were quite consistent acrossothers. All ofthe accuracy scores for the ag-gressive-rejected subjects were near zero,suggesting that they were quite accurate intheir rejection assessments. A status of otherX status of subject x ethnicity interaction,F(4, 98) = 4.06, p < .006, revealed that thiscontrast was most extreme between African-American average subjects and Caucasianaggressive-rejected subjects. African-American subjects made an extreme shift inaccuracy between their overestimation of so-cial rejection for average (M = .12) and pop-ular otiiers (M = .26), and their underestima-tion of rejection for rejected others (M =— .12), whereas Caucasian aggressive-rejected subjects were consistent and accu-rate in their assessments of average (M =— .04), popular (M = —.02), and rejectedothers (M = —.09). African-American ag-gressive-rejected subjects and Caucasian av-erage subjects demonstrated patterns that

Page 11: A Comparison of Aggressive Rejected (Zakriski and Coie 1996)

sfto

T3

ID

gCJ

'5?

CI

I0

'3?

§

'5?rt

r.H 00 t -O O -H

f CO CDO --H .-<

- J IO t -

o o oi 1

^ o cin! . (U

Si I§2 S(D a «

3 gaj c 3•S.SPg

tt, u 5

la ° -^ B o

Page 12: A Comparison of Aggressive Rejected (Zakriski and Coie 1996)

were less extreme but similar to their same-status counterparts. There was no status ofsubject main effect. In summary, insteadof being less accurate in their assessmentsofthe social rejection of others than averageor nonaggressive-rejected boys, aggressive-rejected boys appear to make more consis-tently accurate assessments of other chil-dren's rejection than do average status boys.

These findings lend some support to thehypothesis that aggressive-rejected boys arejust as accurate as nonaggressive-rejectedand average status boys in judging the socialstatus of other children. In absolute terms,there were some instances when they wereeven more accurate. Aggressive-rejectedboys made consistently more accurate as-sessments of other children's social accep-tance and social rejection than did averagestatus children. When judging the socialacceptance of popular children, they weremore accurate than either average status ornonaggressive-rejected boys. It is importantto note that often these absolute accuracieswere relative inaccuracies when aggressive-rejected children were compared with theother status groups who tended to underesti-mate other children's social acceptance andoverestimate their social rejection. Nonethe-less, their other-perceptions appear to be animprovement over their self-perceptions.Recall that when judging their own socialrejection, aggressive-rejected children werethe most inaccurate both in absolute termsand in relative comparisons between statusgroups.

These findings suggest that aggressive-rejected boys are capable of reading cuesabout rejection in others, even though theyunderestimate their own rejection. Thus, weare inclined to reject the hypothesis that ag-greSsive-rejected children are less aware oftheir social status than other children as aresult of a general deficiency in reading so-cial status cues from peers. In light of thisconclusion, we proceed to consider a morespeoific explanation for group differencesin s°elf-awareness, namely, that aggressive-rejected boys are less perceptive in readingnegative feedback directed at themselvesthan when it is directed at other children.

Study 3: Interpretations ofExperimentally ControlledSelf-directed and Other-directedRejection Feedback

The goal ofthe third study was to deter-mine whether failure to recognize one's re-

Zakriski and Coie 1059

jected status with peers might be traced to asimilar failure in recognizing that one hasreceived negative evaluation cues frompeers. If so, this would help to explain whythe aggressive-rejected boys were less accu-rate in estimating how much they were dis-liked by peers. There are, of course, otherexplanations for this phenomenon. One isthat peers are not as open or as direct in theirexpressions of dislike for aggressive-rejectedchildren as they are toward nonaggressive-rejected children. One reason for being lessopen with the aggressive children is thatpeers may worry about physical retaliationand intimidation if they make their feelingsknown. There is some evidence that al-though aggressive-rejected and nonaggres-sive- rejected children do not differ in actualrejection (Hymel et al., 1993), aggressive-rejected children may be less excluded bytheir peers (Boivin et al , 1991; Caims,Cairns, Neckerman, Gest, & Gariepy, 1988;Hymel et al., 1993) and suffer less active andpassive peer disregard than nonaggressive-rejected children (Boivin et al., 1991). Fur-thermore, aggressive-rejected boys more of-ten have their aggressive acts ignored bypeers than other boys and thus receive unin-tended reinforcement of their aggressive be-havior (Coie, Dodge, Terry, & Wright, 1991).In this and other ways, they may not receivedirect indications of peer dislike.

Because there was no way of determin-ing what kind of negative feedback thegroups in Study 1 actually may have re-ceived from peers, we designed an experi-ment to provide a more controlled test ofthehypothesis that aggressive-rejected childrenare self-protective when processing peer re-jection cues, rather than simply failing to re-ceive them. In Study 3, we exposed aggres-sive-rejected and nonaggressive-rejectedboys to experimentally manipulated rejec-tion feedback from peer-like confederates todetermine whether there were subgroup dif-ferences in the recognition of negative feed-back. We hoped to replicate the pattern ofresults from the first two studies with socialfeedback that was equally rejecting of selfand other. We expected nonaggressive-rejected and average status boys to judgeself- and other-directed feedback compara-bly, and for aggressive-rejected boys to rateself-directed feedback more positively. Re-sults such as these would not rule out differ-ences in the quality of actual rejection feed-back from peers as a potential influence onsocial status assessments subjects may makein their everyday interactions, but they

Page 13: A Comparison of Aggressive Rejected (Zakriski and Coie 1996)

1060 Child Development

might point to a role for defensive informa-tion processing. In everyday interactionsit might be the case that both these fac-tors work to influence the inaccuracy ofaggressive-rejected children's status self-perceptions.

Subjects from the second study partici-pated in this third study, which includedtwo major manipulations. First, subjectsjudged comparable self-directed and other-directed rejection feedback. We did this toreplicate the self-other differences found inStudies 1 and 2, without the confound of dif-ferences in rejection feedback. Second, sub-jects judged two different levels of socialfeedback; ambiguous (void of direct positiveand direct negative feedback) and negative.We did this to further test the hypothesisthat aggressive-rejected children are moredefensive when processing direct rejectionfeedback. If defensiveness is involved inthis process, we would expect subjects tomake greater distortions when the feedbackis more negative. Although participation inthe first two studies may have sensitizedchildren to affiliation cues in their class-rooms, the sensitization was no greater forany one of our subject groups since all par-ticipated equally.

METHOD

SubjectsThe subject groups were the same as

those involved in Study 2: 18 aggressive-rejected males'' (8 Caucasian and 10 African-American); 16 nonaggressive-rejected males(9 Caucasian and 7 African-American); 22 av-erage males (11 Caucasian and 11 African-American). As with Study 2, only males wereincluded in this study because ofthe scarcityof aggressive-rejected females.

ProcedureOverview.—Subjects participated in

two experimental tasks. The first involved

viewing videotapes of other children receiv-ing social feedback, referred to as the other-directed feedback condition. The secondinvolved receiving feedback from experi-mental confederates, referred to as the self-directed feedback condition. The other-directed feedback condition was alwayspresented first, followed by a 5-min dis-tractor task in which the subjects played acard game with their interviewers, and thenfollowed by the self-directed feedback con-dition. The order ofthe negative and ambig-uous vignettes was counterbalanced. Smallnumbers of subjects made it impossible tocounterbalance both the feedback target(self vs. other) and the feedback valence(negative vs. ambiguous). We decided tocounterbalance valence because pretestingindicated that children's judgments of theambiguous feedback were heavily influ-enced by whether it came before or after thenegative feedback. Specifically, if it cameafter the negative feedback, it was seen asquite positive by comparison. Because ourmain hypothesis involved an interaction ofstatus of subject with target of feedback,rather than a main effect for target of feed-back, counterbalancing for target of feed-back did not seem as critical. We also feltthat presenting the other-directed conditionfirst would serve to familiarize all subjectsequally with the game that they were to playwhen they themselves participated in theself-directed condition. Because of this de-sign feature, target-of-feedback main effectsshould be interpreted with caution. Study 3was conducted during the same laboratoryvisit as Study 2. Half of the subjects partici-pated in Study 2 before participating in theself-directed condition of Study 3, and halfof the subjects participated in the self-directed condition of Study 3 before partici-pating in Study 2.

Other-directed feedback.—In the other-directed feedback condition, each subject

^ The aggression cutoff for forming this group was .5 as discussed in Study 1. Analyses werealso conducted using the more traditional 1 SD cutoff. The results duplicated the results pre-sented in this paper. The distrihution of subjects in these more traditional groups—11 aggressive-rejected (3 Caucasian and 8 African-American), and 16 nonaggressive-rejected (9 Caucasian and7 African-American)—however, made it difficult to detennine whether status subgroup or eth-nicity was responsible for the critical findings. Because the results are not different when moreCaucasian subjects are added to the aggressive-rejected group through the use of a less stringentaggression cutoff, we believe that ethnicity cannot be the only variable accounting for the effectswe observe. We realize that our use of a more ienient aggression cutoff may appear to makeour results less generalizable. However, we would like to underscore that even with a moreheterogeneous aggressive-rejected group we were able to find the differences we predictedbased on our review ofthe literature. Also, our results do not appear to be driven by the additionof subjects that others might not consider aggressive-rejected. Instead, we find the same resultswhen these children are eliminated.

Page 14: A Comparison of Aggressive Rejected (Zakriski and Coie 1996)

Zakriski and Coie 1061

viewed two 5-min videotaped vignettes. Ineach videotaped vignette, two childrenplayed a board game in which they tookturns asking each other to do somethingspecified by cards that they drew as theymoved about the game board. In the contextof this game, one of the children providedcues about how much he liked or dislikedthe other child. In one of the vignettes thefeedback was negative, and in the other thefeedback was ambiguous. The child givingthe feedback (the protagonist) varied acrossthe two vignettes, while the child receiv-ing the feedback (the target child) remainedconstant. Two sets of vignettes were pro-duced so that children could view and makejudgments about children of their own eth-nicity. The aetors for the vignettes were se-lected from another school district so as tominimize the chances of subjects knowingthem. None of the subjects reported recog-nizing any of the actors they viewed on thevideotapes.

Subjects viewed the videotapes in indi-vidual sessions with one of four experiment-ers present. Half of the subjects of each eth-nicity worked with an African-Americaninterviewer and half with a Caucasian inter-viewer. Before viewing each vignette, sub-jects were told that we were testing a newgame that was designed to help children getto know each other. We told them weneeded them to help us decide if the gamewas fun for children to play, and if it helpedchildren to get to know each other. Afterviewing each tape, subjects were asked toindicate on a seven-point scale, ranging from"did not like at all" to "liked a lot," howmuch they thought the protagonist liked thetarget child, referred to as attributed liking,and how much they thought the target childliked the protagonist, referred to as liking.To support this cover story, children alsowere asked to rate how much they thoughteach of the children liked the game theywere playing. The latter responses are notincluded in the data analyses.

Self-directed feedback.—In the self-directed feedback condition, subjects inter-acted for 5 min each with two experimentalconfederates who were scripted to providedifferent types of feedback to the subject.The 5-min duration was chosen for both theother-directed and the self-directed feed-back conditions to provide an interaction ofsuiEcient length for children to be able tomake judgments about liking, but briefenough for confederates to be able to main-tain their roles. The two types of feedback

were virtually identical to the negative andambiguous feedback given in the other-directed feedback condition. As part of theirtraining, confederates studied the video-tapes produced for the other-directed condi-tion to make their style of feedback deliveryas similar as possible. The subjects and con-federates played the same board gameviewed in the other-directed feedback con-dition. The negative confederate was differ-ent from the ambiguous confederate for eachsubject, but the role assignments were heldconstant across subjects to insure consis-tency of feedback within valence condition.All subjects interacted with confederates oftheir same ethnicity to minimize the influ-ence of cross-ethnicity biases on children'sjudgments. Thus, there were four confeder-ates: African-American ambiguous, African-American negative, Caucasian ambiguous,and Caucasian negative. The same persontrained the African-American and Caucasianconfederates to reduce interindividual vari-ability. Also, confederates had at least tworefresher rehearsals on the days they wereto interact with subjects to reduce intra-individual variability. As in the other-directed condition, confederates were se-lected from a different school district tominimize the chances of subjects knowingthem. Unfortunately, four subjects knew oneof the confederates they were to interactwith. In these cases, we spoke with the con-federate in private and told him to play thegame as he would if we had not given himinstructions. As a result, one aggressive-rejected subject, two average subjects, andone nonaggressive-rejected subject are notincluded in this report.

Subjects were led through the proce-dure by the same person who interviewedthem in the other-directed condition. Beforesubjects were introduced to the confederate,they were told that they were going to playthe same game they just saw on video to seeif they enjoyed playing it themselves. Wetold them they would be playing it with twodifferent children because "sometimes howmuch you like a game has a lot to do withwho you are playing with, and we want tomake sure we find out how much you likethe game itself." After the subject and con-federate were introduced and given instruc-tions, the interviewer told the children thatthe subject should go first. This insured thatthe confederates would receive the cards forwhich they had rehearsed answers, and in-sured consistency of the interaction acrossall subjects. After each 5-min interaction.

Page 15: A Comparison of Aggressive Rejected (Zakriski and Coie 1996)

1062 Child Development

subjects were asked to rate on the same scalehow much the confederate liked them (at-tributed liking). The subjects were also askedhow much they liked the confederate (liking).The children were also asked to rate howmuch they liked the game and how muchthey thought the confederate liked the game.These latter responses were not analyzed.

The attributed liking measure providedthe main assessment of subjects' interpreta-tions ofthe feedback they saw or were given.However, it was also expected that the likingmeasure would serve as an indirect measureof how positively subjects judged the protag-onists' behavior toward themselves and thevideo target child. According to the rein-forcement-affect model of interpersonal at-traction (Clore & Byrne, 1974), if subjectsreally thought a protagonist acted positivelytoward them they would like the protago-nist, because people like others who act inways that are reinforcing. Likewise, theywould probably not like the protagonist ifthey felt the protagonist had acted nega-tively toward them, because people tend notto like others who make them feel bad. Itis also a well-supported social psychologicalfinding that we are attracted to those we be-lieve are attracted to us, and not attractedto those we believe are not attracted to us(Backman & Secord, 1959; Curtis & Miller,1986).

To prevent subjects from feeling dis-tressed by their interaction with the nega-tive confederate, they were debriefed fol-lowing the second confederate interactionby their individual interviewer. They weregiven an opportunity to ask questions andthen taken to a play session where theycould interact naturally with the confeder-ates and other subjects. Because play ses-sions took place after everyone present hadparticipated in both tasks, and because theexperiment took place in the summer, it isunlikely that subjects who had completedthe procedure contaminated subjects whohad not yet participated. All post-experimentplay sessions were supervised by adults toinsure that subjects had positive interactionswith the confederates.

MaterialsThe dimensions of rejecting behavior

manipulated in the scripts for this studywere drawn from Furman and Bierman's(1984) study of children's conceptions offriendship as well as our own informal ob-servations of children's peer interactions.We inverted the dimensions of friendship

they derived from interviews with childrento create dimensions of behavior one mightexpect from someone who was unfriendly,and included examples of each in ourscripts. Pilot testing suggested that childrenrecognized the rejection behaviors we se-lected and found them to be credible.

In the negative feedback condition, theprotagonist begrudgingly played the gamesuggested by the subject (in the self-directedcondition) or the video target child (in theother-directed condition) and complainedabout being there with the subject/video tar-get child. The protagonist refused to answersome of the questions, displayed negativeaffect, and was competitive with the subjector the video target child. Perhaps the strong-est liking cue came when the subject/videotarget child read the game caid "Ask yourpartner to tell you something he likes aboutyou." The negative protagonist refused toanswer. In the ambiguous feedback condi-tion the protagonist agreed to play the gamesuggested by the subject/video target childbut did so without enthusiasm. He cooper-ated, but answered questions briefly,vaguely, and offered nothing beyond whatwas asked. He displayed neutral affect andwas neither supportive nor competitive withthe subject/video target child. In responseto the question asking what he liked aboutthe subject/video target child, the ambigu-ous protagonist said, after a short pause,"Hmmm . . . Well . . . I guess your shirt isnice? "

The scripts for the self- and other-directed conditions were developed throughextensive field testing. It was essential to de-velop scripts that were realistic and thatcould be performed in both the video andconfederate contexts. For example, becausethe self-directed feedback condition in-volved child subjects receiving feedback, itwas essential that the negative feedback notbe blatant or hurtful. A second reason for notusing blatant feedback was to enable us toobserve a range of accuracy in children'sjudgments of liking. The results of pilot test-ing demonstrated that when asked howmuch they thought the protagonist likedthem on a 7-point scale, the target childrencould clearly distinguish between the twodifferent feedback conditions, F(2, 64) =92.04, p < .0001 (Ma, b = 4.95, SD = 1.93;M^ g = 1.82, SD = 1.45).

Data AnalysisThe main analyses were two four-way

repeated-measures ANOVAs: 3 (social sta-

Page 16: A Comparison of Aggressive Rejected (Zakriski and Coie 1996)

Zakriski and Coie 1063

tus: aggressive-rejected, nonaggressive-rejected, average) x 2 (ethnicity: African-American, Caucasian) X 2 (target offeedback: other, self) X 2 (valence of feed-back: ambiguous, negative) with target offeedback and valence of feedback as re-peated measures. One ANOVA was con-ducted on attributed liking data, and theother on liking data. Both analyses utilizedthe General Linear Models framework. Onlywhen the overall model was significant weresignificant main effects and interactions ex-amined. Because there were only three lev-els of status group and because status groupcomparisons were planned, all status groupmain effects and interactions were followedby contrast tests using Fisher's LSD.

RESULTS

Judgments of Feedback Given to Self andOthers

The following analyses reflect the ex-tent to which subjects interpreted the socialcues in the videotaped vignettes and thecues they received from their experimentalplay partners as indicators that the partnerliked them (or the video target child).

The ANOVA on attributed liking scoresrevealed a significant main effect for valenceof feedback, F(l, 46) = 129.49, p < .0001.Overall, subjects attributed more liking tothe ambiguous feedback condition (M =5.02) than to the negative condition (M =

3.17). This suggests that the manipulation offeedback valence was successful. There wasa main effect for ethnicity, F(l, 46) = 15.13,p < .001. African-American subjects (M =4.43) attributed more positive feelings to theprotagonists than did Caucasian subjects (M= 3.62). There also was a main effect fortarget of feedback, F(l, 46) = 42.82, p <.0001, indicating that all subjects attributedmore liking to the protagonists in the self-directed condition (M = 4.62) than to theprotagonists in the other-directed condition(M = 3.57). However, because these condi-tions were not counterbalanced, this effectshould be interpreted with caution. Therewas no main effect for social status.

The ANOVA on attributed liking scoresalso revealed four significant interactionsthat qualified the main effects outlinedabove. The first interaction speaks to the ma-jor question being addressed in this study.There was a significant social status x targetof feedback interaction, F(2, 46) = 4.12, p <.02. As seen in Figure 1, tests of contrastsindicated that, as predicted, aggressive-rejected subjects reported a larger differencein attributed liking between the self-directed and other-directed conditions thandid the nonaggressive-rejected subjects, F(l,47) = 7.98, p < .01, and reported a margin-ally larger difference than did the averagestatus subjects, F(l, 47) = 3.70, p < .06.Nonaggressive-rejected and average statuschildren did not differ from each other, F(l,

self-directed

other-directed

agg-rej nonagg-rej average

FIG. 1.—Mean attributed liking by status across target of feedback

Page 17: A Comparison of Aggressive Rejected (Zakriski and Coie 1996)

1064 Child Development

47) = 1.18, p < .28. In terms of absoluteratings, aggressive-rejected boys rated other-directed feedback more negatively than non-aggressive-rejected boys, F(l, 46) = 8.08, p< .01, and average status boys, F(l, 46) =5.00, p < .03. Average and nonaggressive-rejected boys did not differ in their ratingsof other-directed feedback, and none of thegroups differed in their ratings of self-directed feedback.

The second significant interaction was asocial status x valence of feedback interac-tion, F(l, 46) = 3.12, p < .05. Tests of con-trasts indicated that average status subjectsreported a greater difference in attributedliking between the ambiguous and negativeconditions (Majt, = 5.0, M eg = 2.7) than didthe nonaggressive-rejected subjects (M^i, =5.0, M eg = 3.7). Third, there was a signifi-cant ethnicity x target of feedback interac-tion, F(l, 46) = 7.26, p < .01. Inspectionofthe means revealed that African-Americansubjects reported a greater difference in at-tributed liking between the self-directedand other-directed conditions (M gif = 5.3,Mother = 3.6) than did Caucasian subjects(Mseif = 4.0, Mother = 3-3)' The two ethnicgroups did not differ in their ratings of other-directed feedback, F(l, 46) = 2.76, p < .10,but did differ in their ratings of self-directedfeedback, F(l, 46) = 21.97, p < .001.

Finally, as can be seen in Figure 2,there was a significant ethnicity x target offeedback x valence of feedback interaction,F(l, 46) = 5.42, p < .02. Inspection of themeans revealed that African-American sub-jects reported a greater difference in attrib-uted liking between the self-directed andother-directed conditions than did Cauca-sian subjects especially when the feedbackwas negative. The effect for social status xtarget of feedback X valence of feedbackwas not significant (F < 1).

Liking of Protagonists by Self and OthersThese analyses reflect more indirectly

on subjects' reading of social cues. They re-flect the subjects' liking of their play partneror their judgments ofthe video target child'sliking of his play partner. As predicted, chil-dren's judgments of how much the protago-nists liked them and the video target childwere related to their judgments of how muchthey liked the protagonists and thought thevideo target child liked the protagonists.The correlations between liking and attrib-uted liking were as follows: self-directed,r(52) = .76, p < .0001; other-directed, r(55)= .40, p < .002.

The ANOVA on liking scores revealeda significant valence of feedback effect, F(l,46) = 65.50, p < .0001. Subjects liked, andthought the video targets liked, the ambigu-

3

1

self-directed

other-directed

Caucasianamb neg

African Americanamb neg

FIG. 2.—Mean attributed liking by ethnicity across target of feedback and valence of feedback

Page 18: A Comparison of Aggressive Rejected (Zakriski and Coie 1996)

ous feedback protagonists (M = 5.62) betterthan the negative feedback protagonists (M= 4.39). There was an ethnicity main effect,F(l, 46) = 6.95, p < .01. African-Americansubjects both .liked, and thought othersliked, the protagonists (M = 5.31) more thandid the Caucasian subjects (M = 4.60).There also was a target of feedback main ef'-fect, F(l, 46) = 9.40, p < .004, indicatingthat all subjects said they liked the protago-nists (M = 5.27) more than they thought thevideo targets liked the protagonists (M =4.74). However, due to a lack of counterbal-ancing, this effect should be interpretedwith caution. There was no social statusmain effect.

There was a significant social status xtarget of feedback interaction, F(2, 46) =3.61, p < .04, which appeared to qualify thetarget of feedback main effect. As can beseen in Figure 3, tests of contrasts indicatedthe disparity in liking for the protagonistsbetween the self-directed condition and theother-directed condition was significantiylarger for the aggressive-rejected subjectsthan the nonaggressive-rejected subjects,F(l, 47) = 5.84, p < .02, or the average statussubjects, F(l, 47) = 5.14, p < .03. Nonag-gressive-rejected and average status subjectsdid not differ from each other, F < 1. Subjectgroups did not differ in their absolute ratingsof other-directed feedback, or self-directed

Zakriski and Coie 1065

feedback. No other interactions were sig-nificant.

General DiscussionAggression, Social Status, and Judgmentsof Peer Rejection

The results of Study 3 show that whenaggressive-rejected and nonaggressive-rejected boys are given identical social feed-back, aggressive-rejected boys infiate theirratings of self-directed feedback relative totheir own ratings of comparable other-directed feedback. This finding is consistentwith the results of our first two studies, aswell as other research suggesting that ag-gressive-rejected children underestimatetheir social rejection more than nonaggres-sive-rejected children (Cillessen et al.,1992). This experimental finding also makesit unlikely that our field study findings weredriven solely by differences in the qualityand intensity of rejection feedback thataggressive-rejected and nonaggressive-rejected boys receive in their everyday peerinteractions. It also provides one possibleexplanation for reported differences in lone-liness and self-concept between aggressive-rejected and nonaggressive-rejected chil-dren (Boivin, Thomassin, & Alain, 1989;,Boivin, Vitaro, & Tremblay, 1989; Parkhurst& Asher, 1992; Patterson et al., 1990): Ag-gressive-rejected boys may defend them-

self-<UrBcted

other-directed

agg-rej nonagg-rej average

FIG. 3.—Mean liking by status across target of feedback

Page 19: A Comparison of Aggressive Rejected (Zakriski and Coie 1996)

1066 Child Development

selves against evidence of peer dislike bydistorting it in a self-protective way. Thus,they may report relatively low levels of lone-liness and high levels of perceived social ac-ceptance because they convince themselvesthat they are more liked than they really are.

It is important to note that in Study 3aggressive-rejected boys did not rate self-directed feedback more positively than theother two groups in absolute terms. Instead,the self-enhancing distortions made by ag-gressive-rejected boys were relative to theirown ratings of the same feedback given toothers. This difference appears to be the crit-ical one, because determining one's socialposition in a group often involves makingcomparisons between the way people treatoneself and the way others are treated (Fes-tinger, 1954). The results of Study 2 suggestthat ratings of other people are the appro-priate anchor point for interpreting these dif-ferences, since the three groups of boys allhad comparable accuracy in estimating thesocial status of a representative group oftheir classmates. This would seem to suggestthat the relative comparisons of Study 3 mayinvolve less of an exaggeration ofthe signifi-cances of negative feedback to others andmore of a hyposensitivity when this feed-back is applied to the self. This pattern ofsensitivity to something negative in othersand relative insensitivity to that same eventor attribute in oneself supports Asher etal.'s (1990) suggestion that aggressive-rejected children have overly positive self-perceptions due to ego defensiveness anddenial strategies.

Because we predicted that in Study 3aggressive-rejected boys would positivelydistort self-directed feedback self-protec-tively, we expected distortion to be strongerwhen the feedback was negative. We did notfind evidence of this. Aggressive-rejectedchildren attributed more liking to the self-directed condition regardless of whether thefeedback was ambiguous or negative. Thissuggests that the phenomenon is more gen-eral than we expected. It is likely that theself versus other discrepancy for aggressive-rejected children would disappear if theywere given positive feedback. However, fur-ther research is needed to determine this.

One question that remains is whetheraggressive-rejected boys inaccurately readthe negative feedback or whether they readit correctly but reported it inaccurately to theexaminer. Our second set of findings pro-vides some evidence which bears on this

distinction. In addition to reporting that theprotagonist liked them much more than heliked the video target child, aggressive-rejected boys also reported liking the protag-onist much more than the video target childliked the protagonist. As with attributedliking, this second finding was true of allsubjects; however, it was more true foraggressive-rejected boys than it was for non-aggressive-rejected or average boys. The factthat aggressive-rejected boys positively dis-torted both sides of the interaction suggeststhat they actually thought the feedback theyreceived was more positive than the feed-back the video target child received, and isconsistent with the reinforcement-affectmodel of attraction (Clore & Byrne, 1974).They seemed to be saying, "He liked mebetter than the other guy," and "I liked himbetter than the other guy did." Their inter-pretations ofthe self-directed feedback weremore positive than what normal social desir-ability might explain, and they also wereconsistent across judgments with varyinglevels of social sensitivity. That is, it couldbe argued that it would be easier to tell anadult that you did not like someone whomyou thought treated you badly than it wouldbe to tell an adult that someone who treatedyou badly did not like you. Aggressive-rejected boys apparentiy had difficulty withboth tasks.

Before leaving the topic of status groupdifferences in the judgment of negativefeedback, a point should be made about non-aggressive-rejected boys. Although the maineffect for target of feedback suggests that allsubjects rated self-directed feedback higherthan other-directed feedback, this was onlymarginally true for nonaggressive-rejectedboys. Comparisons of nonaggressive-rejected boys' ratings of both types of feed-back revealed that they felt the feedbackthey received was comparable to the feed-back they viewed on videotape. Thus, incomparison to the other two groups, theywere remarkably accurate and nondefensivewhen judging self-directed feedback. Thisfinding supports the link that has been sug-gested between nonaggressive-rejection anddepression. Studies with adults have foundthat under some circumstances, depressedadults can be more accurate and less defen-sive than nondepressed adults in readingsocial feedback, a phenomenon psychopa-thology researchers refer to as depressive re-alism (Alloy & Abrahamson, 1988; Margo,Creenberg, Fisher, & Dewan, 1993). Thereis some indication also that nonaggressive-

Page 20: A Comparison of Aggressive Rejected (Zakriski and Coie 1996)

Zakriski and Coie 1067

rejected children show more signs of depres-sion than other children (Boivin et al., 1994;Rubin et al., 1990). Thus, it is consistentwith these findings that nonaggressive-rejected children would be less self-protective about peer feedback. It is possi-ble that accurate assessments of peer dislikein some way serve as an early indication ofrisk for depression. Caution must be taken inassuming that the inverse of this speculationholds true, namely, that insensitivity to peerdislike may be a buffer against depression,since there is longitudinal evidence that ag-gressive-rejected youth are at risk for bothexternalizing and internalizing problems inadolescence (Coie, Terry, Lenox, Lochman,& Hyman, in press).

Ethnicity, Gender, and Judgments of PeerRejection

In addition to the predicted status sub-group findings, we also found interestingethnicity differences in judgments of rejec-tion feedback. Similar to the findings for ag-gressive-rejected boys, African-Americansubjects positively distorted self-directedfeedback in comparison to their judgmentsof other-directed feedback. Both African-American and Caucasian subjects demon-strated this tendency, but African-Americansubjects distorted more than Caucasian sub-jects. Unlike aggressive-rejected boys, how-ever, African-American subjects distortedmuch more when the feedback was nega-tive. This finding suggests a self-protectivemechanism: Under conditions where threatto self was greatest (self-directed, negativefeedback), African-American subjects weremore self-enhancing. However, African-American subjects did not report that theyliked the protagonists more than theythought the video targets liked the protago-nists. In sum, they appeared to be saying,"He liked me better than he liked the otherguy, but I did not like him very much." Thisis different from our findings for aggressive-rejected children. It suggests that African-American subjects experienced the self-directed feedback as negative and hadreciprocal feelings for their play partners,but simply distorted their reports of howmuch the negative confederate liked themwhen speaking with the examiner. Thus, wewould conclude that the ethnicity differ-ences are less a matter of perceptual distor-tion—the process we think occurs for ag-gressive-rejected boys—and more a matterof face-saving behavior with the interview-ers. In keeping with this explanation is thefinding that the African-American subjects

made more positive ratings, overall, on bothdependent measures, indicating that theywere displaying more positive attitudes tothe interviewers. This face-saving behaviorobserved in our minority subjects (23 out of25 African-American subjects were in a ra-cial minority at their schools) is consistentwith other self-esteem maintaining behav-iors observed in minority groups (Crocker &Major, 1989).

Further research on the social percep-tions of rejected girls is needed, since sam-ple limitations made it impossible to con-duct Studies 2 and 3 with females. It ispossible that the distinction between aggres-sive- and nonaggressive-rejection is not asmeaningful for girls, or that it is different insome ways for girls than it is for boys. Forexample, French's (1990) analysis of re-jected girls revealed two clusters. However,aggression did not differentiate between thetwo clusters, and the more deviant groupwas characterized by higher levels of with-drawal and anxiety. Another possibility toconsider in future research on rejected girlsis that aggression is displayed differently bygirls, and thus is not adequately assessed bysuch commonly used aggression measures as"Who starts fights, picks on other kids, andteases them?" Preliminary research by Crickand Grotpeter (1995) suggests that girls dis-play a more subtle form of aggression that ismore verbal, focusing on social exclusionand relationship destruction rather thanphysical dominance and control.

Future DirectionsWhile this series of studies points to im-

portant differences in the way subgroups ofrejected children interpret their social expe-riences, it also will be important to betterunderstand the actiial social experiences ofthese two groups of rejected children. Al-though we have used the tei-m "distortion"to describe the discrepancies in interpretingsocial fisedback among our subjects, our un-derstanding of this interpretive process isthat it is nested within a history of previousinteractions with peers and others. In thepresent study, children only interacted withthe confederates for 5 min and were thenasked to make judgments about whether thechild they played with liked them or not. Itis possible that given the short interactiontime, children based their judgments moreon their own prior relationship history withother children than on the brief interactionwith the confederate. Since research sug-gests that aggressive-rejected children mayreceive less clear feedback about peer dis-

Page 21: A Comparison of Aggressive Rejected (Zakriski and Coie 1996)

1068 Child Development

like than nonaggressive-rejected children(Boivin et al., 1991; Coie et al., 1991), it ispossible that they rated the self-directedfeedback more positively than nonaggres-sive-rejected boys, relative to their ratings ofother-directed feedback, because their priorexperiences with peers were inconsistentwith this negative feedback. That is, aggres-sive-rejected boys and average status boysmay have similar histories of peer feedback,in general, even though peers actually feeldifferently about them. In the case of theaggressive-rejected boys, peers may holdback their resentment and dislike lest theyprovoke retaliation and abuse from theseboys (Coie et al., 1991). Developmentally, itwill be important to understand at whatpoint rejection experiences differ for thesetwo groups of rejected children, and howthese differences might give rise to the dif-ferences we observed in the reading of so-cial feedback. A longitudinal study of peerrejection experiences and status awarenessis an important next step. In such a studythe possibility that an inability to read peerfeedback leads to aggressive-rejection mightalso be examined.

Longitudinal research can also informus about the point at which these differencesin status awareness emerge and whetherthey might decrease with age and advancingsocial-cognitive abilities. Preliminary, cross-sectional research by Rabiner and Keane(1991) suggests that aggressive-rejected andnonaggressive-rejected children's interpre-tations of rejection are similar until lateelementary school. In second grade, bothaggressive- and nonaggressive-rejected chil-dren reported poor peer treatment. How-ever, in fifth grade, nonaggressive-rejectedchildren reported significantly better peertreatment than others. It is unclear from thisstudy whether this shift in rejection percep-tions results from an improvement in actualpeer treatment of aggressive-rejected chil-dren, or whether it demonstrates an emerg-ing self-protective bias which may have de-veloped as a method of coping with chronicrejection. Either way, these data suggest thatthe difference we observed would not bepresent in younger rejected children, andmay persist even in older rejected childrenwithout intervention. Research on rejectedsubgroup differences in loneliness and in-terpersonal concerns in middle school (Park-hurst & Asher, 1992) supports this hypothe-sis by suggesting that aggressive-rejectedand submissive-rejected children remaindifferent in their perceptions of rejection

even at this later point in development. Fu-ture research on this topic should includechildren both younger and older than thoseincluded in the present study.

Fuicher research on how other factorsthat distinguish aggressive-rejected andnonaggressive-rejected children may infiu-ence their interpretation of rejection feed-back is also needed. For example, socialintelligence factors may differentiate aggres-sive-rejected and nonaggressive-rejectedchildren and may have infiuenced how wellthe two groups were able to process the in-formation presented to them in the brief, 5-min interaction. Aggressive-nonrejectedchildren should also be assessed to deter-mine what role aggression, independent ofrejection, plays in children's interpretationsof rejection feedback.

A final avenue for future research in-volves testing the limits ofthe distortions weobserved in this study to help us better un-derstand their source. Although we have hy-pothesized that aggressive-rejected boystruly distorted the interactions they experi-enced, whereas African-American boys sim-ply chose to report that they were liked morethan they were, further research is neededto support these hypotheses. One possiblestrategy for such a study would be to offer aprize to subjects for accurate assessment ofthe peer feedback and inform subjects thatthe confederate told the experimenter howmuch he actually liked them. If distortionswere found to diminish under such condi-tions, it could be concluded that the feed-back was accurately read and that the prob-lem rests in reporting self-derogatoryinformation.

Understanding the source of the distor-tions we have observed in this series of stud-ies will help us to understand their functionand their role in status maintenance andtreatment resistance. For example, Crockerand Major (1989) suggest that the types ofself-protective mechanisms employed byoppressed minority groups are adaptive.However, it seems that the types of distor-tions employed by aggressive-rejected boysmay contribute to their peer difficulties andserve to sustain their low social status. Re-jection has been demonstrated to be morestable for aggressive-rejecte'd than for nonag-gressive-rejected boys (Cillessen et al.,1992). Because self-awareness of one's prob-lems may be a necessary precursor for be-havior change, poor self-awareness maycontribute to these problems of status conti-

Page 22: A Comparison of Aggressive Rejected (Zakriski and Coie 1996)

Zakriski and Coie 1069

nuity. However, as suggested earlier, poorstatus awareness also may act as a temporarybuffer against depression and low self"-esteem for aggressive-rejected children.Finding ways to help aggressive-rejectedchildren incorporate accurate assessments ofpeer feelings toward them in a way thatallows them to benefit from this feedbackand be motivated to make changes in theirbehavior will be a challenge for future inter-vention research on this topic.

ReferencesAlloy, L. B., & Abrahamson, L. Y. (1988). De-

pressive realism: Four theoretical perspec-tives. In L. B. Alloy (Ed.), Cognitive pro-cesses in depression (pp. 223-265). NewYork: Guilford.

Asher, S. R., Parkhurst, J. T., Hymel, S., & Wil-liams, G. A. (1990). Peer rejection and loneli-ness in childhood. In S. R. Asher & J. D. Coie(Eds.), Peer rejection in childhood (pp. 253-273). New York: Cambridge University Press.

Asher, S. R., Zelis, K. M., Parker, J. G., & Bruene,C. M. (1991). Self-referral for peer relationsproblems among aggressive and withdrawnlow-accepted children. Paper presented atthe biennial meeting of the Society for Re-search in Child Development, Seattle.

Backman, C. W., & Secord, P. F. (1959). The effectof perceived liking on interpersonal at-traction. Human Relations, 12, 379-384.

Boivin, M., Gote, L., & Dion, M. (1991). The self-perceptions and peer experiences of aggres-sive-rejected and withdrawn-rejected chil-dren. Paper presented at the eleventhbiennial meeting of the International Societyfor the Study of Behavioral Development,Minneapolis, MN.

Boivin, M., Poulin, F., & Vitaro, F. (1994). De-pressed mood and peer rejection in child-hood. Development and Psychopathology, 6,483-498.

Boivin, M., Thomassin, L., & Alain, M. (1989).Peer rejection and self-perceptions amongearly elementary school children: Aggressiverejectees vs. withdrawn rejectees. In B. H.Schneider, G. Attili, J. Nadel, & R. P. Weiss-berg (Eds.), Socidl competence in develop-m.ental perspective (pp. 392-393). Boston:Kluwer.

Boivin, M., Vitaro, F., & Tremblay, R. (1989). Be-havior prohlems, competencies and self-perceptions of withdrawn-rejected and ag-gressive-rejected children. Paper presented atthe biennial meeting of the Society for Re-search in Child Development, Kansas City,MO.

Bukowski, W. M., Sippola, L. K., Verlan, P., &

Newcomb, A. F. (1993). The association he-tween aggression, popularity, and narcissismand children's feelings of social well-being.Paper presented at the biennial meeting ofthe Society for Research in Ghild Develop-ment, New Orleans, LA.

Caims, R. B., Cairns, B. D., Neckerman, H. J.,Gest, S. D., & Gariepy, J. L. (1988). Socialnetworks and aggressive behavior: Peer sup-port or peer rejection? Developmental Psy-chology, 24, 815-823.

Gillessen, A. H. N., van IJzendoom, H. W., vanLieshout, G. F. M., & Hartup, W. W. (1992).Heterogeneity among peer-rejected boys:Subtypes and stabilities. Child Development,63, 893-905.

Glore, G. L., & Byme, D. (1974). A reinforcement-affect model of attraction. In T. L. Huston(Ed.), Foundations of interpersonal at-traction (pp. 143-170). New York: AcademicPress.

Coie, J. D., & Dodge, K. A. (1983). Continuitiesand changes in children's sociometric status:A five-year longitudinal study. Merrill-Palmer Quarterly, 29, 261-282.

Coie, J. D., Dodge, K. A., & Goppotelli, H. (1982).Dimensions and types of social status: Across-age perspective. Developmental Psy-chology, 18, 557-570.

Coie, J. D., Dodge, K. A., Terry, R., & Wright, V.(1991). The role of aggression in peer rela-tions: An analysis of aggression episodes inboys' play groups. Child Development, 62,812-826.

Coie, J. D., Lochman, J. E., Terry, R., & Hyman,C. (1992). Predicting adolescent disorderfrom childhood aggression and peer rejection.journal of Consulting and Clinical Psychol-ogy, 60, 783-792.

Coie, J. D., Terry, R., Lenox, K., Lochman, J., &Hyman, G. (in press). Ghildhood peer rejec-tion and aggression as predictors of stable pat-terns of adolescent disorder. Developmentand Psychopathology.

Grick, N., & Grotpeter, J. (1995). Relational ag-gression, gender, and social-psychological ad-justment. Child Development, 66, 710-722.

Grocker, J., & Major, B. (1989). Social stigma andself-esteem: The self-protective properties ofstigma. Psychological Bulletin, 96, 608-630.

Gurtis, R. G., & Miller, K. (1986). Believing an-other dislikes you: Behaviors making the be-liefs eome true. Journoi of Personality and So-cial Psychology, 51, 284-290.

Dodge, K. A., Pettit, G. S., McClaskey, G. L., &Brown, M. (1986). Social competence in chil-dren. Monographs ofthe Society for Researchin Child Development, 51(2, Serial No. 213).

Dodge, K. A., & Somberg, D. R. (1987). Hostileattributional biases among aggressive boys

Page 23: A Comparison of Aggressive Rejected (Zakriski and Coie 1996)

1070 Child Development

are exacerbated under conditions of threats tothe self. Child Development, 58, 213-224.

Festinger, L. (1954). A theory of social comparisonprocesses. Human Relations, 7, 117-140.

French, D. C. (1988). Heterogeneity of peer re-jected boys: Aggressive and nonaggressivesubtypes. Child Development, 59, 976-985.

French, D. C. (1990). Heterogeneity of peer re-jected girls. Child Development, 61, 2028-2031.

Furman, W., & Bierman, K. (1984). Children'sconceptions of friendship: A multimethodstudy of developmental changes. Develop-mental Psychology, 20, 925-931.

Gage, N. L., & Cronbach, L. J. (1955). Conceptualand methodological problems in interper-sonal perception. Psychological Review, 62,411-422.

Harter, S. (1982). The Perceived CompetenceScale for Children. Child Development, 53,87-97.

Hymel, S., Bovi'ker, A., & Woody, E. (1993). Ag-gressive versus withdrawn unpopular chil-dren: Variations in peer and self-perceptionsin multiple domains. Child Development, 64,879-896.

Hymel, S., Rubin, K. H., Rowden, L., & LeMare,L. (1990). Children's peer relationships: Lon-gitudinal prediction of internalizing and ex-ternalizing problems from middle to latechildhood. Child Development, 52, 171-178.

Krantz, M., & Burton, C. (1985). The developmentof the social cognition of social status./ourna/of Genetic Psychology, 147, 89-95.

Kupersmidt, J. B., & Coie, J. D. (1990). Preadoles-cent peer status, aggression, and school ad-justment as predictors of externalizing prob-lems in adolescence. Child Development, 61,1350-1362.

Lochman, J. E., Meyer, B. L., Rabiner, D. L., &White, K. J. (1991). Parameters influencingproblem-solving of aggressive children. InR. J. Prinz (Ed.), Advances in behavioral as-sessment of children and families (pp. 3 1 -63). New York: Jessica Kingsley.

Lochman, J. E., & Wayland, K. (1994). Aggression,social acceptance and race as predictors ofnegative adolescent outcomes. Journal of theAmerican Academy of Child and AdolescentPsychiatry, 33, 1026-1035.

Margo, G. U., Greenberg, R. P., Fisher, S., & De-wan, M. (1993). A direct comparison of thedefense mechanisms of nondepressed peopleand depressed psychiatric inpatients. Com-prehensive Psychiatry, 34, 65-69.

Marsh, H. W., Smith, I. D., & Barnes, J. (1983).Multitrait-multimethod analyses of the Self-

Description Questionnaire: Student-teacheragreement on multidimensional ratings ofstudent self-concept. American EducationalResearch Journal, 20, 333-357.

Maxwell, S. E., & Delaney, H. D. (1990). Design-ing experiments and analyzing data: A modelcomparisons perspective (pp. 170—206). Bel-mont, CA: Wads worth.

Parkhurst, J. T., & Asher, S. R. (1992). Peer rejec-tion in middle school: Subgroup differencesin behavior, loneliness, and interpersonalconcerns. Developmental Psychology, 28,231-241.

Patterson, C. J., Kupersmidt, J. B., & Griesler,P. C. (1990). Children's perceptions of selfand of relationships with others as a functionof sociometric status. Child Development, 61,1335-1349.

Rabiner, D. L., & Gordon L. (1993). Relations be-tween children's self-concepts and social in-teraction strategies; differences betweenrejected and accepted boys. Social Develop-ment, 2, 83-94.

Rabiner, D. L., & Keane, S. P. (1991). Children'sbeliefs about peers in relation to their socialstatus. Paper presented at the biennial meet-ing of the Society for Research in Ghild De-velopment, Seattle.

Rubin, K. H., LeMare, L. J., & Lollis, S. (1990).Social withdrawal in childhood: Develop-mental pathways to peer rejection. In S. R.Asher & J. D. Coie (Eds.), Peer rejection inchildhood (pp. 217-249). New York: Cam-bridge University Press.

Singleton, L. C , & Asher, S. R. (1977). Peer pref-erences and social interaction among thirdgrade children in an integrated school dis-trict. Journal of Educational Psychology, 69,330-336.

Terry, R., & Coie, J. D. (1990). Comparative mod-els for estimating the utility of childhood pre-dictors of adolescent disorder. Paper pre-sented at the University of MontrealConference on Predicting Children's Disrup-tive Behavior, Baie-St. Paul, Quebec.

Younger, A. J., Schwartzman, A. E., & Ledingham,J. E. (1985). Age-related changes in children'sperceptions of aggression and withdrawal intheir peers. Developmental Psychology, 21,70-75.

Zakriski, A. L., Coie, J. D., & Wright, J. C. (1992).The accuracy of children's sociometric self-perceptions. In J. B. Kupersmidt (Chair), Mul-tiple sources of information about children'speer relationships. Symposium conducted atthe Conference on Human Development, At-lanta.

Page 24: A Comparison of Aggressive Rejected (Zakriski and Coie 1996)