Lectura 07-Crick Dodge (1994)

Social Information-Processing Mechanisms inReactive and Proactive Aggression

Nicki R. CrickUniversity of Illinois at Urbana-Champaign

Kenneth A. DodgeVanderbilt University

CRICK, NICKI R., and DODGE, KENNETH A. Social Information-Processing Mechanisms in Reactiveand Proactive Aggression, GHILD DEVELOPMENT, 1996, 67, 993-1002. Theories of aggressive be-havior and ethological obsei-vations in animals and children suggest the existence of distinctforms of reactive (hostile) and proactive (instrumental) aggression. Toward the validation of thisdistinction, groups of reactive aggressive, proactive aggressive, and nonaggressive children wereidentified (n = 624 9—12-year-olcls). Social infonnation-processing patterns were assessed inthese groups by presenting hypothetical vignettes to subjects. 3 hypotheses were tested: (1) onlythe reactive-aggressive children would demonstrate hostile biases in their attributions of peers'intentions in provocation situations (because such biases are known to lead to reactive anger);(2) only proactive-aggressive children would evaluate aggression and its consequences in rela-tively positive ways (because proactive aggression is motivated hy its expected external out-comes); and (3) proactive-aggressive children would select instrumental social goals rather thanrelational goals more often than nonaggressive children. All 3 hypotheses were at least partiallysupported.

Theories of aggressive behavior (Ban- groups of proactive-and reactive-aggressivedura, 1973; Berkowitz, 1993) and ethological children (Dodge & Coie, 1987). Evenobservations in animals (Lorenz, 1966) and though these two types of behavior are posi-children (Price & Dodge, 1989) suggest the tively correlated, distinct types of childrenexistence of distinct forms of reactive (hos- can be identified with a sufficiently largetile) and proactive (instrumental) aggression sample.(see Hartup, 1974). Reactive aggression,which has its theoretical roots in the frustra- The goal ofthe present research was totion-aggression model (Berkowitz, 1993; test hypotheses regarding distinctive socialDollard, Doob, Miller, Mowrer, & Sears, information-processing patterns that have1939), is an angry, defensive response to been implicated in the development andfrustration or provocation. Proactive aggres- maintenance of reactive and proactive ag-sion, which is described in social learning gression in children (Dodge, 1990; Dodge &theory formulations of aggression (Bandura, Coie, 1987). According to social information-1973), is a deliberate behavior that is con- processing models, children's social behav-trolled by external reinforcements (i.e., it is ior is a function of sequential steps of pro-a means for obtaining a desired goal). Re- cessing, including encoding of social cues,search on these two forms of aggression in interpretation of social cues, clarification ofchildhood has recently been made possible goals, response access or construction, re-by the development of a teacher-rating in- sponse decision, and behavioral enactmentstrument that reliably and validly identifies (Crick & Dodge, 1994; see Dodge, 1986;

Portions of this paper were presented in S. P. Hinshaw (Ghair), Externalizing Disorders ofGhildhood: Recent Research Developments, at the annual meeting ofthe American Psychologi-cal Association, August 1990, Boston. This research was supported in part by a Vanderbilt Univer-sity Fellowship to the first author and research grants to the second author hy the NationalScience Foundation and the National Institute of Ghild Health and Human Development. Wewould like to thank the Nashville, Tennessee, Metropolitan Public School teachers and studentswho participated in this project. Special thanks also to Zvi Strassberg, Tracy Arnold, GynthiaZiegler, and Beth Harris for their assistance with data collection and data coding. Please sendcorrespondence regarding this paper to the first author at Human Development and FamilyStudies, 1105 West Nevada Street, University of Ilhnois, Urhana, IL 61801.

[Child Development, 1996,67,993-1002. © 1996 by the Society for Research in Child Development, Inc.All rights reserved. 0009-3920/96/6703-0030$01.00]

994 Child Development

Dodge & Crick, 1990; Dodge, Pettit, Mc-Claskey, & Brown, 1986). Skillful process-ing at each step is hypothesized to lead tocompetent performance within a situation,whereas biased or deficient processing is hy-pothesized to lead to deviant social behavior(e.g., aggression). Research on individualdifferences in social information processinghas consistently shown that aggressive chil-dren perceive, interpret, and make decisionsabout social stimuli in ways that increase thelikelihood of their engaging in aggressiveacts (see Dodge & Crick, 1990, for a review).

At least two general social information-processing patterns have been found to becharacteristic of aggressive children. Thefirst pattern involves processing at the inter-pretation step of the previously describedmodel. Studies of individual differences inchildren's attempts to interpret social stim-uli have established that aggressive childrenexhibit hostile attributional biases in re-sponse to ambiguous provocation situations(i.e., they attribute malicious intent to thepeer provocateur more often than do otherchildren) (e.g.. Dell Fitzgerald & Asher,1987; Dodge, 1980; Dodge & Frame, 1982;Guerra & Slaby, 1989; Nasby, Hayden, &DePaulo, 1980). For children who interpreta peer's behavior as intentionally harmful tothe self, aggression may serve as a retaliationor defense against the peer. This type of ag-gressive act is reactive aggression (i.e., a re-action to a perceived threat).

A second social information-processingpattern that is characteristic of aggressivechildren involves the response decision stepof processing. During this step, childrenevaluate possible behavioral responses to aparticular social situation according to sev-eral criteria, such as the type of outcomeslikely to accrue for each response and thedegree of confidence that they feel abouttheir ability to perform each response(Dodge & Crick, 1990). The result of thisevaluation is a response decision in whichchildren select one response for enactment(possibly the selection of the response thatwas evaluated most favorably). Research onthe response decision step of processing hasdemonstrated that aggressive children eval-uate aggressive acts in ways that are likelyto lead to the enactment of aggressive behav-ior, that is, they expect relatively positiveoutcomes to accrue for aggressing and feelmore confident about their ability to performaggressive acts than do their nonaggressivepeers (e.g.. Crick & Ladd, 1990; Feldman &Dodge, 1987; Hart, Ladd, & Burleson, 1990;

Perry, Perry, & Rasmussen, 1986; Slaby &Guerra, 1988; Waas, 1988). Thus, for chil-dren who exhibit this second social informa-tion-processing pattern, aggression mayfunction as a viable means for obtaining de-sired goals. This type of aggression is proac-tive in nature.

In the present research, our major objec-tive was to test two hypotheses: (1) that thefirst social information-processing pattern(hostile attributional bias) is characteristic ofreactive-aggressive children, and (2) that thesecond processing pattern (positive evalua-tions of aggression) is characteristic of proac-tive-aggressive children. Initial support forthe first hypothesis has been demonstratedby Dodge and Coie (1987); fhe second hy-pothesis, however, has never been tested.Testing these hypotheses is important bothto validate the proactive-reactive distinctionand to provide empirical support for the the-oretical formulations posited here.

An additional goal of this research wasto contrast the social goals (processingstep 3) of reactive-aggressive, proactive-aggressive, and nonaggressive children. Be-cause of the hypothesized goal-oriented na-ture of proactive aggression, the social goalsof the proactive group were particularly ofinterest. Goals are focused arousal states thatfunction as orientations toward producingparticular outcomes (Crick & Dodge, 1994).Past research has demonstrated that, in gen-eral, social maladjustment is related to theformulation of social goals that are likely tobe relationship-damaging (Renshaw &Asher, 1983; Slaby & Guerra, 1988). Giventhis evidence, and the hypothesized instru-mental nature of proactive aggression, wehypothesized that proactive-aggressive chil-dren would endorse goals that are moreinstrumental, and less relational, in focusthan those endorsed by their nonaggressivepeer-s.

To address these objectives, teacher rat-ings (Dodge & Coie, 1987) were used toidentify groups of reactive-aggressive, pro-active-aggressive, and combined reactive-proactive aggressive children and a nonag-gressive comparison group in a sample ofthird- through sixth-grade children. Socialinformation-processing patterns were as-sessed with hypothetical-situation instru-ments that were adapted from prior research(Crick & Ladd, 1990; Dell Fitzgerald &Asher, 1987; Dodge, 1980; Perry, Perry, &Rasmussen, 1986; Wheeler & Ladd, 1982).Children's intent attributions, response de-

Crick and Dodge 995

cision processes (i.e., outcome expectationsand feelings of efficacy for enacting socialbehaviors), and social goals were assessedfor conflict and peer group entry situations.These two situation types were chosen be-cause previous research has shown themto be problematic for socially maladjustedchildren (Dodge, McClaskey, & Feldman,1985).

MethodSubjects

A total of 624 third- through sixth-gradeboys and girls (9-12 years of age) from fourpublic schools in a large, metropolitan cityserved as subjects.^ The sample included173 third graders, 169 fourth graders, 146fifth graders, and 136 sixth graders. About40% of the sample was African-Americanand the remainder was European-American.Each subject had parental consent to partici-pate in the project (the parental consent ratewas 79%).

Teacher Ratings of AggressionTeachers rated each of their students

with respect to 13 descriptive items using a1 (Never true of this child) to 5 (Always trueof this child) scale for each item. Thisteacher rating instrument, developed byDodge and Coie (1987), consists of a reac-tive-aggressive scale with three items (e.g.,"When this child has been teased or threat-ened, he or she gets angry easily and strikesback"); a proactive-aggressive scale withthree items (e.g., "This child uses physicalforce in order to dominate other kids"); andfiller items (e.g., "This child cooperates wellwith peers"). The internal consistency ofeach aggression scale was found to be high(alphas = .90 and .90, respectively).

Children's reactive- and proactive-aggression scale scores were used to iden-tify aggressive groups. Children were classi-fied as reactive-aggressive if they receiveda reactive-aggression scale score that wasgreater than 1 SD above the sample mean.All other children were classified as nonre-active-aggressive. Similarly, children wereclassified as proactive-aggressive if they re-ceived a proactive-aggression scale scorethat was greater than 1 SD above the samplemean. All other children were classified as

nonproactive-aggressive. This classificationprocedure resulted in the identification of508 nonaggressive children (261 third/fourthgraders, 247 fiflh/sixth graders; 285 girls, 223boys), 22 reactive-aggressive children (14third/fourth graders, 8 fifth/sixth graders; 9girls, 13 boys), 34 reactive-aggressive chil-dren (23 third/fourth graders, 11 fiflh/sixthgraders; 19 girls, 15 boys), and 60 reactive-plus proactive-aggressive children (44 third/fburth graders, 16 fifth/sixth graders; 20girls, 40 boys).

Social Information-Processing MeasuresIntent attribution instrument.—An in-

strument developed by Dell Fitzgerald andAsher (1987) was used to assess children'sintent attributions (adapted from Dodge,1980; Dodge & Frame, 1982). This instru-ment consists of six stories, each of whichdescribes a provocation situation in whichthe intent of the provocateur is ambiguous(e.g., a peer breaks the subject's new radiowhile the subject is out of the room). Forevery story, children answered two ques-tions, each of which assessed their attribu-tions ofthe provocateur's intent. For the firstquestion, children were asked to circle oneof four presented reasons for the provoca-tion. Two of the presented reasons refiectedhostile intent (e.g.. The kid was mad at me)and two refiected benign intent (e.g.. Theradio wasn't made well). For the secondquestion, children were asked to tellwhether the provocateur's behavior was in-tentional (i.e., hostile intent) or accidental(i.e., benign intent). Following the proce-dures described by Dell Fitzgerald andAsher (1987), children's responses to the twotypes of attribution questions were summedwithin and across the six stories (children'sscores for this 12-item measure could rangefrom 0 to 12). Cronbach's alpha was .90 forthe intent attribution scale.

Response decision instrument.—Thisinstrument was designed to assess children'sresponse decision processes for peer groupentry and peer conflict situations. Specifi-cally, the measure assessed children's out-come expectations (adapted from Crick &Ladd, 1990) and feelings of self-efficacy(adapted from Perry et al., 1986; Wheeler &Ladd, 1982) for aggression (see Appendix forexample items).^ Two exemplars of each sit-

' A few children (n = 16) moved prior to the administration of Session 2, and thus the cellsizes are slightly smaller for some analyses.

^ Ghildren also evaluated avoidant and assertive strategies for each situation. Analyses ofchildren's evaluations of these two strategy types (i.e., self-efScacy and outcome expectations)were not presented here because they were not relevant to our hypotheses. These items were


uation type (conflict and group entry) wereincluded in the instrument, resulting in a to-tal of four situations (i.e., two conflict situa-tions and two group entry situations). Chil-dren evaluated two aggressive strategies(one depicted verbal aggression and theother depicted physical aggression) for eachof the four situations (i.e., they evaluated atotal of eight strategies). As will be de-scribed in more detail below, children'sevaluations of physical and verbal aggres-sion were summed for each situation to yieldtotal aggression evaluation scores.

Children evaluated two types of out-comes for each ofthe eight strategy-situationstimuli: (1) instrumental (i.e., the strategywill or will not lead to a desired instrumen-tal goal), and (2) relational (i.e., the strategywill or will not result in the peer liking you).These types of outcomes have been shownin prior research to be salient to children andto distinguish between socially adjusted andmaladjusted groups of children (Crick &Ladd, 1990).

For each of the outcome expectationitems, children evaluated which of two op-posite outcomes would occur (one positiveand one negative) if they enacted the pre-sented strategy in the given situation. Thenthey decided whether that outcome wouldhappen "most of the time" or "sometimes"(format was based on Crick & Ladd, 1990).The order of presentation ofthe positive ver-sus negative outcomes was counterbalancedacross items and outcome types. Within eachsituation type (conflict and entry), children'sresponses to the outcome expectation itemswere summed across the two outcome types(relational and instrumental) and across thetwo aggressive strategies (physical and ver-bal aggression) to yield a total outcome ex-pectation score that ranged from 8 to 32.Cronbach's alpha for the outcome expecta-tion scale was .65 for the conflict situationsand .65 for the peer group entry situations.

For each of the eight strategy-situationstimuli, children also evaluated how easy orhard it would be for them to enact the pre-sented strategy (self-efficacy). For each item,the self-efficacy evaluations ranged from 1(very hard) to 4 (very easy) (see Appendixfor an example item). As with the outcomeexpectation items, children's responses to

the efficacy items were summed across theoutcome types (relational and instrumental)and across the two aggressive strategies(physical and verbal aggression) for each sit-uation type (conflict and entry) to yield twototal self-efficacy scores that each rangedfrom 4 to 16. Cronbach's alpha for the self-efficacy scale was .74 for the conflict situa-tions and .84 for the peer group entry situa-tions.

Social goal instrument.—For each strat-egy-situation stimulus presented in the re-sponse decision instrument, children's so-cial goals were also assessed. This was doneby asking children whether they would pre-fer to have the presented positive instru-mental or the positive relational outcome oc-cur in the given situation (see Appendix forexample item). Instrumental responses werescored as 1 and relational responses werescored as 0. Total goal scores were obtainedby summing over children's responses to theeight strategy-situation stimuli per situationtype (i.e., total goal scores ranged from 0 to8, with higher scores indicative of more in-strumental, rather than relational, goal pref-erences). Cronbach's alpha was .90 for theconflict goal scale and .90 for the peer groupentry goal scale.

Administration procedures.—The child-report instruments were administered duringtwo 40-min group assessment sessions con-ducted within each classroom. Children com-pleted the intent attribution measure duringSession 1. Approximately 1 month later, chil-dren completed the response decision and so-cial goal instrument (Session 2). These ses-sions were conducted by the first author, amale doctoral student, and two female under-graduate research assistants.

Results

To assess group differences in chil-dren's social information processing, sevenanalyses of variance were conducted inwhich reactive-aggression group (two levels:aggressive and nonaggressive), proactive-aggression group (two levels: aggressive andnonaggressive), and grade (two levels: third/fourth and fifth/sixth) served as the indepen-dent variables and children's social informa-tion-processing scores (i.e., intent attribu-tions, self-efficacy and outcome expectations

included in the measure to avoid response biases (i.e., to prevent response sets that might resultfrom evaluating negative strategies only) and to provide a balanced assessment of children'ssocial goals (i.e., to prevent children from thinking about their goals in the context of aggressivestrategies only).

Crick and Dodge 997

TABLE 1

CHILDREN'S INTENT ATTRIBUTION SCORES BY AGGRESSION GROUP AND GRADE

Third/fourth

Fifth/sixth

Nonagg.

5.9(3.8)4.9

(3.6)

Reactive

5.1(3.3)8.5

(2.5)

GROUP

Proactive

5.8(3.8)6.7

(3.9)

Reactive/Proactive

6.6(3.9)5.7

(4.4)

NOTE.—Groups: Nonagg, = Nonaggressive; Reactive = Reactive Aggressive; Proactive =Proactive Aggressive; Reactive/Proactive = Reactive and Proactive Aggressive. Higher scoresare indicative of more hostile attributions. Standard deviations are in parentheses.

for aggression enacted in peer conflict situa-tions, self-efficacy and outcome expectationsfor aggression enacted in peer group entrysituations, social goals for conflict situations,and social goals for group entry situations)served as the dependent variables.'^

Intent AttributionsSee Table 1 for cell means and standard

deviations of the intent attribution score xaggression group and grade. Analysis of vari-ance yielded a significant main effect ofgrade, F(l, 616) = 6.3, p < .05, and a sig-nificant proactive group x reactive group Xgrade interaction, F(l, 616) = 6.6, p < .05.Cell means showed that younger children(M = 6.0, SD = 3.8) attributed hostile intentto peer provocateurs significantly more oftenthan did older children (M = 5.1, SD = 3.7).To further evaluate the significant interac-tion effect, a 2 (proactive-aggression group)X 2 (reactive-aggression group) simple ef-fects analysis of variance was conducted foreach level of grade. The analysis of youngerchildren's attribution scores yielded no sig-nificant effects. However, the analysis ofolder children's attributions yielded a sig-nificant proactive group X reactive group in-teraction, F(l, 278) = 5.91, p < .05. A post-

hoc test (LSD, p < .05) of the olderchildren's group means showed that the re-active-aggressive group made significantlymore hostile attributions than did the nonag-gressive group. Although higher, the reac-tive-aggressive group's scores did not differsignificantly from those of the proactive-aggressive group and the reactive- plus pro-active-aggressive group.

Response Decision ProcessesSee Table 2 for cell means and standard

deviations by proactive-aggression group.^

Conflict situations.—Analyses of chil-dren's evaluations of aggression yielded asignificant main effect of proactive aggres-sion for children's outcome expectationscores, F(l, 598) = 3.8, p < .05, and self-efficacy scores, F(l, 598) = 8.7, p < .01. Cellmeans indicated that proactive-aggressivechildren reported significantly more positiveoutcome expectations and greater efficacyfor enacting aggression than did nonproac-tive-aggressive children.

Peer group entry situations.—Analysesof children's evaluations of aggressionyielded a significant main effect of proactive

^ Cell sizes were too small to permit the inclusion of sex as an additional independentvariable in these analyses. To insure that sex did not moderate any of the obtained efFects, asecond set of analyses was conducted in which sex, rather than grade, was included as an inde-pendent variable. The only unique effects obtained in these analyses involved three-way interac-tions among proactive aggression, reactive aggression, and sex for children's outcome expecta-tions for enacting aggression in peer group entry situations and children's social goals for entrysituations. Inspection of cell means for these situations indicated that aggressive girls (reactive-only and proactive-only groups) evsJuated aggression more favorably and held more instrumentalgoals than did their aggressive male counterparts. All other effects involving sex (except for maineffects, which were not the focus of this research) were nonsignificant.

^ A multivariate approach to these analyses was also considered (i.e., MANOVAs in whichoutcome expectation scores and efficacy scores for a particular strategy type within a particularsituation type would serve as the variates). However, correlational analyses showed that outcomeexpectations and self-efficacy scores were not strongly correlated (i.e., r = .15 for the conflictsituations and r = .35 for the peer group entry situations). Thus, a univariate approach seemedmost appropriate for the present data.


TABLE 2

CHILDREN'S OUTCOME EXPECTATION AND SELF-EFFICACY SCORES BY PROACTIVEAGGRESSION GKOUP

CROUP

Proactive Aggressive Nonproactive Aggressive

Conflict situations;Outcome expectations

Self-efficacy

Social goals

Peer group entry situations;Outcome expectations

Self-efficacy

Social goals

17.5*(3.9)13.3**(2.7)5.7*(2.1)

17.5(4.4)10.6**(3.8)4.2(2.8)

16.7(4.3)12.1(3.1)4.7(2.7)

16.4(3.8)8.6

(3.5)3.5

(2.8)

NOTE.—Higher scores are indicative of more positive outcome expectations, greater self-efficacy, and more instrumental (rather than relational) social goals. Standard deviations are inparentheses.

* p < .05.**p < .01.

aggression for children's self-efficacy scores,F(l, 598) = 7.9, p < .01, and significant maineffects of grade for children's self-efficacyscores, F(l, 598) = 6.3, p < .05, and outcomeexpectation scores, F(l, 598) = 9.4, p < .01.Cell means indicated that the proactive-aggressive children reported significantlygreater efficacy for enacting aggression thandid nonproactive-aggressive children. Fur-ther, younger children (M = 9.4, SD = 3.8)reported significantly greater efficacy andmore positive outcome expectations (M =17.0, SD = 3.8) for aggressive acts than didolder children (Ms = 8.4, 16.0, SDs = 3.4,3.9, respectively).

Social GoalsAnalyses of children's goals for conflict

situations yielded a significant main effectof proactive-aggression group, F(l, 590) =4.0, p < .05. Specifically, proactive-aggressive children selected instrumentalrather than relational goals significantlymore often than did their nonproactive-aggressive peers (see Table 2 for cell meansand standard deviations by proactive-aggression group). Analyses of children'sgoals for peer group entry situations yieldeda significant main effect of grade, F(l, 590)= 8.2, p < .01. Cell means showed that olderchildren (M = 3.2, SD = 2.7) chose rela-tional goals rather than instrumental goals

significantly more often than did youngerchildren (M = 3.9, SD = 2.8).

Discussion

This study provides evidence favoringthe validity of the distinction between reac-tive and proactive types of aggression. As hy-pothesized, subgroups of aggressive chil-dren were identified and found to processsocial information in distinctive ways. Thesefindings contribute to our understanding ofthe social-cognitive mechanisms relating toeach form of aggression.

As hypothesized, proactive-aggressivechildren evaluated verbally and physicallyaggressive acts in significantly more positiveways than did children who were not proac-tively aggressive. In terms of the describedsocial information-processing model, the re-sponse decision processes of proactive-aggressive children are likely to result in theenactment of instrumental aggressive behav-ior. These findings are consistent with thehypothesis that proactive aggression is con-trolled (and motivated) by the expectation ofexternal rewards (Dodge & Coie, 1987). Thatis, proactive-aggressive children are likely toview aggression as an effective and viablemeans for obtaining social goals. Becausepeers are likely to submit to aggressive over-tures, the positive view of aggression held

by proactive-aggressive children may be-come stronger over time (i.e., because theyhave increased opportunities to observe theeffectiveness of aggression and to build con-fidence in their ability to enact aggression).

Comparison ofthe social goals of proac-tive-aggressive children and their peers in-dicates that proactive-aggressive childrenare less likely than other children to endorserelationship-enhancing goals during socialinteraction. Rather, they are more likely toprefer goals that are instrumental in natureand relatively self-enhancing (e.g., obtaininga toy from a peer rather than becomingfriends with that peer). Aggressive acts typi-cally have great potential for damaging rela-tionships with others. However, becauseproactive-aggressive children seem less ori-ented toward relational goals than theirpeers, their social objectives may be rela-tively less likely to inhibit their use of ag-gression. Further, their greater preferencefor instrumental goals may support or evenmotivate the use of aggression (i.e., becauseaggression is typically an effective tool forachieving instrumental goals). In terms ofthe described social information-processingmodel, the present results provide evidencethat proactive-aggressive children exhibitcognitive patterns at step 3 of processing(goal clai-ification) that are likely to contrib-ute to the use of aggressive behavior duringpeer interaction.

Also as hypothesized, in response to am-biguous provocation situations, reactive-aggressive children attributed hostile intentto peer provocateurs more frequently thandid nonaggressive peers, at least among thefifth and sixth graders. This finding repli-cates the result obtained by Dodge and Coie(1987) and extends it by showing that thehostile attributional biases of reactive-aggressive children still hold when girls, aswell as boys, are included in the subgroup(i.e., unlike the present sample, the Dodgeand Coie sample consisted of boys only).Findings from both studies are consistentwith the hypothesis that, for reactive-aggressive children, aggressive acts are mo-tivated by attributions of hostility towardpeers, who are perceived as mean andthreatening to the self. In some cases, thishostile interpretation of a peer's intent maybe accurate, while in other cases it is likelyto be inaccurate. Reactive-aggressive chil-dren are not likely to give peers the "benefitofthe doubt," however, as they tend to per-ceive hostility on the part of the peer evenwhen none is necessarily intended.

Crick and Dodge 999

Perceptions of hostile peer intent arepredictive of retaliatory aggressive re-sponses (Dodge, 1980), and hostile attribu-tional biases are known to predict later ag-gressive acts by the child (Dodge, Bates, &Pettit, 1990). These aggressive acts are likelyto lead to increases in peers' actual hostil-ity and rejection toward the child. Thus,reactive aggression may be maintained by anegative cycle in which: (1) a reactive-aggressive child attributes hostile intent topeers (whether it is intended by the peersor not) and retaliates aggressively, (2) thechild's peers then respond with increasedhostility toward the child, and (3) the childinterprets the peers' hostility as confirmationof the earlier interpretation (i.e., that peersare mean). In this way, the biased social in-formation processing of reactive-aggressivechildren may become a self-fulfilling proph-ecy (i.e., with time, peers may actuallybecome more hostile toward reactive-aggressive children). Of course, longitudinaldesigns are needed to test adequately thishypothesis.

One finding that fails to support this for-mulation was the lack of hostile attributionalbiases among reactive-aggressive third andfourth graders (i.e., the youngest group).This finding is in contrast to that obtainedby Dodge and Coie (1987), who found youngreactive-aggressive (first- and third-grade)children to exhibit attributional biases. Sev-eral reasons for the present results seemplausible. First, an analysis of children's ag-gression scores revealed that reactive-aggressive third and fourth graders were sig-nificantly more proactive-aggressive thanreactive-aggressive fifth and sixth graders (asindicated by a significant proactive group byreactive group by grade interaction for chil-dren's proactive-aggression scores). Thus,the youngest reactive-aggressive group inthe present sample may have been relativelyless likely to exhibit the processing patternhypothesized to be characteristic of reactive,but not proactive, aggression. Second, anumber of methodological differences existbetween the Dodge and Coie study and thepresent research. Specifically, the Dodgeand Coie study involved: (1) the use of peerrejection as an additional classification crite-ria for reactive-aggressive children, (2) theuse of video-recorded stimuli and individualinterviews to assess attributions rather thana group-administered questionnaire, and(3) the use of an African-American, male sam-ple rather than a multiracial sample of bothboys and girls. It is possible, for example.


that the questionnaire method used in thepresent research may be less effective foreliciting attributional biases of young reac-tive-aggressive children than the more en-gaging, video-record method employed byDodge and Coie (this seems likely given thatthe video-record method is more similar topeer interaction). Clearly, additional re-search is needed in order to clarify the roleof age and other factors in the hostile attribu-tional biases of reactive-aggressive children.

Results ofthe present study also demon-strate that reactive-aggressive children donot evaluate the instrumental use of aggres-sion in relatively positive ways and do nothold more instrumental social goals than doother children. This finding is consistentwith the view that reactive aggression is notmotivated by the hope (or expectation) ofachieving instrumental goals. Because reac-tive aggression is typically accompanied byfeelings of intense anger, it seems likely thatreactive-aggressive children strike out to-ward peers in an out-of-control manner,without consideration of the consequencesof their behavior. In terms ofthe social infor-mation-processing model described earlier(Crick & Dodge, 1994), it is possible that,once reactive-aggressive children make ahostile interpretation of a particular situa-tion, this interpretation (and accompanyingemotion) preempts any further processing inthe situation. A similar finding was demon-strated for aggressive children as a group un-der conditions of emotional arousal byDodge and Somberg (1987). If preemptionholds, reactive-aggressive children wouldnot engage in response evaluation for thesetypes of situations (making these processesirrelevant or unrelated to reactive-aggressive children's behavior for these par-ticular situations).

These findings have important implica-tions for intervention with aggressive chil-dren. First, they support the existence of dis-tinct forms of aggression (reactive andproactive) and demonstrate that the socialinformation-processing mechanisms associ-ated with each form are also distinctive. Thispattern of findings suggests the need for in-tervention programs that are also unique foreach type of aggression (i.e., that are tailoredto the specific problems and strengths exhib-ited by each group). In a recent review, Coieand KoeppI (1990) describe two types of in-tervention programs, and these types may beappropriate for treating reactive-aggressiveversus proactive-aggressive children. Thefirst type of program is oriented toward

anger control (e.g., Lochman, 1985; Loch-man, Burch, Curry, & Lampron, 1984). Typi-cally, this kind of intervention programteaches children to recognize body cues thatsignal anger, to use self-control techniquesduring the course of a conflict, and to prob-lem solve. The second type of program fo-cuses on changing the reinforcement contin-gencies in the child's environment so thataggression does not lead to positive conse-quences for the child and to subsequentfeelings of efficacy for enacting aggression(Patterson, 1974, 1982). Both of these typesof intervention programs have been usedwith aggressive children; however, as Coieand KoeppI (1990) point out, the first pro-gram seems best suited for the problemsexperienced specifically by reactive-aggressive children, whereas the secondprogram seems well suited for the treatmentof proactive-aggressive children. Based onthe present findings, these intervention pro-grams might also be more effective if, in ad-dition to targeting a more specific subpopu-lation of aggressive children, they alsoaddressed the unique social information-processing biases and deficits of each sub-group. Most important, intervention shouldbe targeted to changing the interpretationprocesses of reactive-aggressive childrenand the response decision and social goalprocesses of proactive-aggressive children.

The grade differences found in socialgoals indicate that older children prefer rela-tional goals to instrumental goals more fre-quently than do younger children. Thisfinding is consistent with prior studies thathave shown older children to be more awareof and more concerned about relational as-pects of social interaction relative to theiryounger peers (Crick & Ladd, 1990, 1993).Longitudinal x'esearch is needed to assessthe relative roles of instrumental versus rela-tional goals with advancing age (e.g., to de-termine whether relational goals replace in-strumental goals as children grow older orwhether children find ways to balance thetwo types of goals).

Present findings also indicate that olderchildren evaluate aggression more nega-tively than do their younger peers. Theseresults are consistent with developmentalhypotheses derived from social information-processing theory. In general, this modelposits that developmental changes in behav-ior are related to developmental changes inprocessing (Crick & Dodge, 1994). In thisinstance, declining use of aggression withadvancing age should be associated with de-

Crick and Dodge 1001

velopmental changes in children's pro-cessing of information about aggressive be-havior. Specifically, processing changesshould occur that are less likely to supportthe use of aggression as children grow older(e.g., through the development of more neg-ative evaluations of aggression). The presentfindings provide support for this hypothesis;the cross-sectional nature ofthe study, how-ever, prevents firm conclusions.

AppendixThings That Happen to MeBall SituationOn the playground one day, you have nothing todo. You see some kids in your class playing a gamewith a ball. You walk over to where they are play-ing.

1. What would happen if you grabbed the ballaway from one of the kids?

a. The kids would not like you.This would happen most of the time.This would happen sometimes.

OR

The kids would like you.This would happen most of the time.This would happen sometimes.

b. The kids would let you have the ball.This would happen most of the time.This would happen sometimes.

OR

The kids would not let you have the ball.This would happen most of the time.This would happen sometimes.

2. How hard or easy would it be for you to grabthe ball away from one ofthe kids?

very hardhardeasyvery easy

3. Which of these two things would you most liketo have happen in this situation?

The kids like you.The kids let you have the hall.

ReferencesBandura, A. (1973). Aggression: A social learning

theory analysis. New York: Prentice-Hall.Berkowitz, L. (1993). Aggression: Its causes, con-

sequences and control. New York: AcademicPress.

Coie, J. D., & KoeppI, G. K. (1990). Adapting in-tervention to the problems of aggressive and

disruptive rejected children. In S. R. Asher &J. D. Coie (Eds.), Peer rejection in childhood.New York: Cambridge University Press.

Crick, N. R., & Dodge, K. A. (1994). A reviewand reformulation of social information-processing mechanisms in children's socialadjustment. Psychological Bulletin, 115,74-101.

Crick, N. R., & Ladd, G. W. (1990). Children'sperceptions of the outcomes of social strate-gies: Do the ends justify being mean? Devel-opmental Psychology, 26, 612-620.

Crick, N. R., & Ladd, C. W. (1993). Children'sperceptions of their peer experiences: Attri-butions, loneliness, social anxiety, and socialavoidsmce. Developmental Psychology, 29,244-254.

Dell Fitzgerald, P., & Asher, S. R. (1987, August).Aggressive-rejected children s attributionalbiases about liked and disliked peers. Paperpresented at the annual meeting ofthe Ameri-can Psychological Association, New York.

Dodge, K. A. (1980). Soeial cognition and chil-dren's aggressive behavior. Child Develop-ment, 51, 162-170.

Dodge, K. A. (1986). A social information pro-cessing model of social competence in chil-dren. In M. Perlmutter (Ed.), Minnesota Sym-posium on Child Psychology (Vol. 18, pp.77-125). Hillsdale, NJ: Eribaum.

Dodge, K. A. (1990). The structure and functionof reactive and proactive aggression. In D.Pepler & K. H. Rubin (Eds.), The develop-ment and treatment of childhood aggression(pp. 201-218). Hillsdale, NJ: Eribaum.

Dodge, K. A., Bates, J. E., & Pettit, G. S. (1990).Mechanisms in the cycle of violence. Science,250, 1678-1683.

Dodge, K. A., & Coie, J. D. (1987). Social informa-tion-processing factors in reactive and proac-tive aggression in children's playgroups. Jour-nal of Personality and Social Psychology, 53,1146-1158.

Dodge, K. A., & Crick, N. R. (1990). Social infor-mation processing bases of aggressive behav-ior in children. Personality and Social Psy-chology Bulletin, 53, 1146-1158.

Dodge, K. A., & Frame, C. L. (1982). Social cogni-tive biases and deficits in aggressive boys.Child Development, 55, 163-173.

Dodge, K. A., McClaskey, C. L., & Feldman, E.(1985). A situational approach to the assess-ment of social competence in children. Jour-nal of Consulting and Clinical Psychology,53, 344-353.

Dodge, K. A., Pettit, G. S., McGlaskey, C. L., &Brown, M. M. (1986). Social competence inchildren. Monographs of the Society for Re-search in Child Development, 51(2, Serial No.213).


Dodge, K. A., & Somberg, D. R. (1987). Hostileattributional biases among aggressive boysare exacerbated under conditions of threats tothe self. Child Development, 58, 213-224.

Dollard, J., Doob, C. W., Miller, N. E., Mowrer,O. H., & Sears, R. R. (1939). Frustration andaggression. New Haven, CT: Yale UniversityPress.

Feldman, E., & Dodge, K. A. (1987). Social infor-mation processing and sociometric status:Sex, age, and situational effects. Journal ofAbnormal Child Psychology, 15, 211-227.

Guerra, N. G., & Slaby, R. G. (1989). Evaluativefactors in social problem solving by aggres-sive boys. Journal of Abnormal Child Psy-chology, 17, 277-289.

Hart, C. H., Ladd, G. W., & Burleson, B. (1990).Ghildren's expectations ofthe outcomes of so-cial strategies: Relations with sociometric sta-tus and matemal disciplinary styles. ChildDevelopment, 61, 127-137.

Hartup, W. W. (1974). Aggression in childhood:Developmental perspectives. American Psy-chologist, 29, 336-341.

Lochman, J. E. (1985). Effects of different treat-ment lengths in cognitive behavioral inter-ventions with aggressive boys. Child Psychia-try and Human Development, 16, 45-56.

Lochman, J. E., Burch, P. R., Gurry, J. F., & Lam-pron, L. B. (1984). Treatment and generaliza-tion effects of cognitive behavioral and goalsetting interventions with aggressive boys.Journal of Consulting and Clinical Psychol-ogy, 52, 915-916.

Lorenz, K. (1966). On aggression. New York: Har-court. Brace, and World.

Nasby, W., Hayden, B., & DePaulo, B. M. (1980).Attributional bias among aggressive boys tointerpret unambiguous social stimuli as dis-plays of hostility. Journal of Abnormal Psy-chology, 89, 459-468.

Patterson, G. R. (1974). Interventions for boyswith conduct problems: Multiple settings,treatments, and criteria. Journal of Con-sulting and Clinical Psychology, 42,471-481.

Patterson, G. R. (1982). Coercive family processes.Eugene, OR: Castalia.

Perry, D. G., Perry, L. G., & Rasmussen, P. (1986).Gognitive social learning mediators of aggres-sion. Child Development, 57, 700-711.

Price, J. M., & Dodge, K. A. (1989). Reactive andproactive aggression in childhood: Relationsto peer status and social context dimensions.Journal of Abnormal Child Psychology, 17,455-471.

Renshaw, P. D., & Asher, S. R. (1983). Ghildren'sgoals and strategies for social interaction.Merrill-Palmer Quarterly, 29, 353-374.

Slahy, R. G., & Guerra, N. G. (1988). Gognitivemediators of aggression in adolescent offend-ers: 1. Assessment. Developmental Psychol-ogy, 24, 580-588.

Waas, G. A. (1988). Social attributional biases ofpeer-rejected and aggressive children. ChildDevelopment, 59, 969-992.

Wheeler, V., & Ladd, G. W. (1982). Assessmentof children's self-efficacy for social interactionwith peers. Developmental Psychology, 18,795-805.

Lectura 07-Crick Dodge (1994)

Documents