Ostro & Crick, 2007

School Psychology Review,2007, Volume 36, No. 1, pp. 22-43

Forms and Functions of Aggression During FarlyChildhood: A Short-Term Longitudinal Study

Jamie M. OstrovUniversity at Buffalo, The State University of New York

Nicki R. CrickUniversity of Minnesota, Twin Cities Campus

Abstract. The school classroom and playground provide an important context forlearning about young children's social interactions. A multimethod, multiinfor-mant, short-term longitu.dinal study was conducted to investigate the utility ofincluding school-based observational assessments of both form (i.e., physical andrelational) and function (i.e., proactive and reactive) of aggressive behavior atschool with a young sample during early childhood (132 children; M = 44.37months; SD = 9.88). The study revealed low intercorrelations between observedproactive and reactive functions of aggression and low to moderate levels ofstability. Based on 160 min of observation per child for an academic year, thefindings revealed that boys are more physically aggressive to peers than are girls,whereas girls are more relationally aggressive than are boys. The results provideevidence for the differential association between aggression categories and futuresocial-psychological adjustment constructs with particular relevancy for schoolcontexts (i.e., peer rejection and student-teacher conflict).

Recently, school psychologists have (Cullerton-Sen & Crick, 2005; Leff & Lakin,called for more extensive research of young 2005; Young, Boye, & Nelson, 2006). Aggres-children's peer relationships and, in particular, sive behavior is defined as any behavior in-relationally aggressive behavior at school tended to hurt, harm, or injure another person

This study was supported by grants from the National Institute of Mental Health (NIMH; MH63684) andthe National Science Foundation (BCS-0126521) to Nicki R. Crick and by NIMH traineeship (MH-15755)to the Institute of Child Development in support of the first author. We thank Drs. W. Andrew Collins,Anthony D. Pellegrini, and Richard A. Weinberg for comments on an earlier draft. Special thanks to JeanBurr, Crystal Cullerton-Sen, Alison Eudeikis, Elizabeth Jansen, Peter Ralston, and the entire PreschoolAggression Longitudinal Study (PALS) Project team for assisting in the collection of the present data.Thank you to the directors, teachers, parents, and children for their participation. Portions of this studywere presented at the XVII World Meeting of the International Society for Research on Aggression,Minneapolis, Minnesota, and the 18th Biennial Conference on Human Development, Washington, DC.This research served as partial fulfillment for the degree of doctor of philosophy at the University ofMinnesota for the first author, under the direction of the second author.

Correspondence regarding this article should be addressed to Jamie M. Ostrov, Department of Psychology,University at Buffalo, The State University of New York, 214 Park Hall, Buffalo, NY 14260-4110; E-mail:[email protected]

Copyright 2007 by the National Association of School Psychologists, ISSN 0279-6015

22

Forms and Functions of Aggression

(Coie & Dodge, 1998). More specifically, re-lational aggression, in which the relationshipor friendship serves as the means of harm, isobserved when malicious gossip, lies, or se-crets are spread, as well as by intentionallyignoring (i.e., "silent treatment") or mali-ciously excluding a peer from an activity orgroup interaction (e.g., "You can't come to mybirthday party" or "You are not my friendanymore"; Crick & Grotpeter, 1995). Recentreviews emphasize that school counselors andpsychologists are better equipped to addressincreasingly complex peer relationship and ad-justment problems when they understand andassess for relationally aggressive behaviors(Young et al., 2006). In addition, to facilitatethis assessment, a call for the use of ecologi-cally valid observational systems for use in theclassroom and on the playground has alsobeen raised by school psychologists (Leff &Lakin, 2005). To date, though, school anddevelopmental psychologists have neglectedthe important early school years. Theoristshave postulated that subtypes of aggressivebehavior (i.e., physical and relational aggres-sion) are present and salient during all devel-opmental periods, including early childhood(for a review, see Crick & Zahn-Waxler,2003). The majority of past developmentaltheory, methods, and research designs haveconcerned physical (i.e., using physical forceor the threat of physical force as the means ofharm) or verbal (i.e., verbal insults and meannames) aggression, and these empirical tradi-tions are not necessarily applicable to thestudy of relational aggression in school set-tings (Crick & Zahn-Waxler, 2003).

The study of aggressive behavior duringearly childhood is an important endeavor giventhe salience of emerging peer relationships andfriendships for the academic, social, emotional,and cognitive development of young children(Crick & Zahn-Waxler, 2003). The interest inearly onset and the developmental course of ag-gression results in part from concems about thestability of aggression across development (see(National Institute of Child Health and HumanDevelopment Early Child Care Research Net-work, 2004) and the developmental significanceof early onset aggressive behavior (Moffitt,

1993). In addition, physical aggression is com-mon during early childhood when emotion reg-ulation and language capacities are not fullydeveloped (Coie & Dodge, 1998). Currently agreat deal is known about physical aggression(e.g., links with marital discord, coercive pun-ishments, and reactive temperaments; see Coie& Dodge, 1998); however, despite majorprogress in recent years, there is still much todiscover concerning the development of rela-tional aggression at school during earlychildhood.

During the last few years researchershave established the salient role of relationalaggression during early childhood (e.g., Os-trov & Keating, 2004). During early child-hood, school-based research using teacher,peer, and occasionally observational methodshave documented that boys are significantlymore likely to display and receive frequentlevels of physical aggression than girls, espe-cially in the company of male peers; girls, onthe other hand, are significantly more likelythan their male peers to be identified as rela-tionally aggressive, especially in the companyof female peers (e.g., Bonica, Yeshova, Ar-nold, Fisher, & Zeljo, 2003; Crick, Casas, &Mosher, 1997; Hawley, 2003; McNeilly-Cho-que. Hart, Robinson, Nelson, & Olsen, 1996;Ostrov, 2006; Ostrov & Keating, 2004; cf.Hart, Nelson, Robinson, Olsen, & McNeilly-Choque, 1998; Kupersmidt, Bryant, & Wil-loughby, 2000; McEvoy, Estrem, Rodriguez,& Olson, 2003). Comparatively fewer studieshave occurred in early childhood relative tomiddle childhood and adolescence, perhapsbecause of the relative difficulty in assessingyoung children's relationally aggressive be-havior. For example, younger children oftenhave limited language skills and may not al-ways be valid and reliable informants, render-ing self and peer assessments challenging andoften of questionable utility (Crick, Ostrov,Appleyard, Jansen, & Casas, 2004).

Proactive and Reactive Aggression

One of the principal areas of inquirywithin the field of aggression is based on thefunctions of aggressive behaviors in the con-

23

School Psychology Review, 2007, Volume 36, No. 1

text of peer relationships. Psychologists havedistinguished aggressive behaviors displayedto serve a goal-directed end (i.e., proactiveaggression) and those displayed in response toa perceived threat and motivated by hostilityor anger (i.e., reactive aggression; Dodge,1991; Vitaro, Gendreau, Tremblay, & Oligny,1998). Recently, there has been a call for theelimination of a dichotomous view of func-tional aggression types because of the concep-tual and empirical overlap between the con-structs and because a reliance on these con-structs precludes an assessment of the multiplemotives of aggression (Bushman & Anderson,2001). The authors argued that a dimensionalapproach to the understanding of proactiveand reactive aggression may be more usefulthan a categorical approach. Bushman andAnderson (2001) recognized that dichotomiesoften generate many conceptual and empiricalissues and called for a more complex andgeneral conceptualization of aggressive be-havior. However, the authors do recognize thebenefit of studying particular aggression func-tion types, which is in keeping with a numberof studies in schools across North Americathat have documented the divergent nature ofthe proactive and reactive aggression con-structs. Past studies have revealed both dis-crete factor loadings (see Dodge & Coie,1987; Poulin & Boivin, 2000) and discrimi-nant validity (Crick & Dodge, 1996; Price &Dodge, 1989; Dodge, Lochman, Hamish,Bates, & Pettit, 1997; Hubbard, Dodge, Cil-lessen, Coie, & Schwartz, 2001; Hubbard etal., 2002; Vitaro et al., 1998; Waschbusch,Willoughby, & Pelham, 1998). Finally, duringearly childhood aggressive behaviors oftenmay be classified as proactive in nature (i.e., toget a toy or object; Cummings, Iannotti, &Zahn-Waxler, 1989), and reactive aggressionmay be less common.

In a further attempt to test the utility ofthe proactive and reactive function types ofaggression, researchers have argued that it isimportant to take into account both the func-tions (i.e., proactive and reactive) and theforms (i.e., physical and relational) of aggres-sion (Little, Jones, Henrich, & Hawley, 2003;Little, Brauner, Jones, Nock, & Hawley, 2003;

Prinstein & Cillessen, 2003). Proactive rela-tional aggression would occur during earlychildhood peer contexts, for example, when achild tells a peer "I will not be your friendanymore unless you give me that toy." Reac-tive relational aggression would occur duringearly childhood peer contexts, for example,when a child retaliates with "You can't cometo my birthday party," in response to beingexcluding from sitting at the snack table withother girls in her class or perhaps in responseto being hit from a peer. With a more complexdelineation it may be possible to see furtherbenefits of these constructs, which would pro-vide additional support for continuing to studyproactive and reactive aggression as relativelyindependent constructs. These recent studies(Little, Jones et al., 2003; Prinstein & Cil-lessen, 2003) with early and late adolescentsdemonstrate some utility for exploring the roleof both forms and functions of aggression forpredicting social-psychological adjustmentoutcomes. However, to date these associationshave not been explored in early childhood.

Despite the recent advances in the physi-cal and relational aggression literature, collec-tively past proactive and reactive aggressionstudies are limited in several ways (see Little,Jones et al., 2003.; Little, Brauner et al., 2003).First, all of the past studies on proactive andreactive functions of physical (and even rela-tional aggression) have been conducted withmiddle childhood or adolescent samples (e.g.,Coie & Dodge, 1998). Second, most studies,even those that include relational aggression as-sessments, are based on potentially biased self-report, parent, or teacher ratings and have notused observational methods (see Leff &Lakin, 2005; cf. Price & Dodge, 1989) or amultimethod, multiinformant approach (cf.Hubbard et al., 2001). Third, the past literaturehas neglected the role of the gender of both theperpetrator and victim in the development ofthese behaviors. Fourth, the range of potentialoutcome variables (i.e., social-psychologicaladjustment problems) has been limited in paststudies and has not assessed adjustment prob-lems that may be particularly problematic foracademic performance (i.e., student-teacherrelationship) in children.

24


Social Psychological Adjustment andAggression Subtypes

Hundreds of studies have demonstratedthe significant association between physicalaggression and future adjustment problems(for a review see Coie & Dodge, 1998). Todate only a few studies have explored possiblelinks between relational aggression and so-cial-psychological adjustment problems overtime (see Crick & Zahn-Waxler, 2003). Ingeneral, work in the field has been limited bya theoretical orientation that has been calledthe "benign childhood" hypothesis, statingthat girls are particularly resilient to disruptionand distress during childhood, which is ob-"served in a lower prevalence of externalizingproblems (Crick & Zahn-Waxler, 2003, p.729). The second component of current theo-ries of early developmental psychopathologyis that when girls do experience disorder, it isin the form of only internalizing problems(e.g., Keenan & Shaw, 1997). Recently, re-searchers have tested these theoretical asser-tions but in most case these adjustment assess-ments (i.e., peer rejection by classmates) in theCrick (1996) and Cillessen and Mayeux(2004) longitudinal studies were limited tomiddle childhood and early adolescence, £mimportant but narrowly focused developmen-tal period. These initial studies do show thattesting the utility of relational aggressionabove and beyond the role of physical aggres-sion is theoretically important for refuting the"myth of the benign childhood of girls" (Zahn-Waxler, 1993). Therefore, in the present earlychildhood investigation, peer rejection was in-cluded as an outcome to be replicated andextended to a younger school context.

The link between relationally aggressivebehaviors and disruptions in teacher-child re-lationships has also been examined. Past re-search has documented that these close rela-tionships serve as a protective factor leading toresilient outcomes even among aggressiveyouth (see Meehan, Hughes, & Cavell, 2003).Specifically, children who have close relation-ships with their teachers have higher rates ofprosocial behavior and social competence dur-ing early childhood (Copeland-Mitchell, Den-

ham, & DeMulder, 1997). Alternatively, chil-dren who have conflictual relationships withearly childhood teachers often have problemsin multiple domains (Pianta, Steinberg, & Rol-lins, 1995), including academic problems inkindergarten (Pianta & Stuhlman, 2004) andacademic and achievement problems througheighth grade (Hamre & Pianta, 2001). Chil-dren who have a relationship with their pri-mary teacher marked by conflict are also muchmore likely to receive a referral for specialeducation services than their peers who havean open, accepting, and close relationship withtheir teacher (Pianta, 1997). These conflictualrelationships are likely to stay somewhat sta-ble across time (Hughes, Cavell, & Jackson,1999) and are predictive of future social com-petence and peer preference (Hughes, Cavell,& Willson, 2001). For these reasons, chil-dren's peer behavior (i.e., functions of rela-tional aggression) was tested as a predictor offuture conflict in the student-teacher relation-ship, even when controlling for the role ofrespective physical aggression functions. Inaddition, the inclusion of these two outcomevariables (i.e., peer rejection and student-teacher conflict) will permit tests of the hy-pothesis that proactive and reactive functionsof relational aggression have unique statisticalassociations with indices of social-psycholog-ical adjustment.

To extend past studies, the current studyhad four main goals. The first was to assess thepsychometric properties of a new school-based observational classification system thatdifferentiates subtypes of physical and rela-tional aggression by two common functions ofaggression (i.e., proactive and reactive) duringearly childhood. In addition, to expand thestudy of relational aggression to younger ages,children as young as 30 months were included.The second goal was to test the short-termstability of these four categories during earlychildhood. The third major goal was to ex-plore gender differences for these behaviors.The fourth goal was to test the unique associ-ations between these aggression categoriesand future peer rejection and conflict withinthe student-teacher relationship.

25


In the present research, in keeping withpast theory on physical and relational aggres-sion (Crick & Grotpeter, 1995) in which chil-dren use gender-linked aggressive strategies tofulfill gender-specific social goals, it is hy-pothesized that similar gender effects will beseen when functions of aggression are as-sessed. That is, proactive and reactive rela-tional and physical aggression categories willreflect gender differences documented in theaggression literature (see Coie & Dodge,1998), but not previously tested with childrenin early childhood school contexts. Specifi-cally, across the study, it is predicted thatyoung girls will display more relational ag-gression than their male peers and young boyswill deliver more physical aggression thantheir female peers regardless of the proactiveor reactive category. In addition, it is predictedthat girls will be more relationally aggressiveto female peers than male peers, and boys willbe more physically aggressive to male versusfemale peers during early childhood (Ostrov &Keating, 2004). It is predicted that proactivefunctions will be more frequent than reactivefunctions (Cummings et al., 1989). Consistentwith the past physical aggression literatureduring early childhood (see Coie & Dodge,1998) and the past relational aggression stud-ies during middle childhood (e.g.. Crick, 1996;Crick, Ostrov, & Werner, 2006), it is hypoth-esized that stability will be demonstrated forall four aggression categories.

It is hypothesized that proactive and re-active relational and physical aggression willbe associated with future adjustment problems(Vitaro et al., 1998; cf. Little, Jones et al.,2003; Little, Brauner et al., 2003). Based onpast research, it is expected that reactive andproactive functions of relational and physicalaggression will be associated with unique out-comes (Crick & Dodge, 1996; Hubbard et al.,2001; Poulin & Boivin, 2000; Waschbusch etal., 1998); Given that prior research has notexplored these questions during early child-hood, these final questions are exploratory innature. However, the two social-psychologi-cal adjustment variables that are included inthe present study (i.e., peer rejection and stu-dent-teacher confiict) are included because

each of these outcomes has been associated orhypothesized to be associated with relationalaggression in the past (Crick & Zahn-Waxler,2003; Cullerton-Sen & Crick, 2005). Based onresearch with older children, it is predictedthat reactive functions of relationally aggres-sive behavior will predict adjustment prob-lems above and beyond proactive aggressivebehavior (e.g., Waschbusch et al., 1998). Thepresent study will not explore the unique as-sociations between physical aggression andadjustment outcomes to reduce the number ofmodels evaluated and because the focus of thepresent study was to test if relational aggres-sion had utility in the prediction of adjustmentproblems above and beyond functions of phys-ical aggression. Thus, based on past findingswith older children (e.g.. Crick, 1996, Crick,Ostrov, & Werner, 2006), it is predicted thatrelational aggression function types will pre-dict future social-psychological adjustmentproblems above and beyond the role of phys-ical aggression. This finding could be impor-tant because it would imply that relationalaggression may uniquely account for variancein the prediction of these adjustment problemsand, therefore, intervention and prevention ef-forts to improve young children's social com-petence would presumably obtain higher ef-fects by including assessments of both formsand functions of aggressive behavior.

Method

Participants

A total of 132 children (69 girls) attend-ing two preschool centers in a large midwest-em city participated in this study. The ongoinglongitudinal study (see Crick, Ostrov, Burr etal., 2006) consisted of two time points sepa-rated by approximately 4 to 5 months andbegan during the fall (data were collected afterchildren had at least 1 month to become fa-miliar with the school and their peers). AtTime 2 (spring of the same year), the sampleconsisted of 116 children (62 girls). Thus,Time 1 began in the fall (2002) and was 2 to 3months in length, followed by a 4- to 5-monthperiod in which no measures were collected,and then followed by another 2- to 3-month

26


Table 1Suinmaty of Sample Size and Measures at Each Time

Validity SampleMeasure Time 1 Time 2 (Time 3)

Sample sizeObservations of proactive or reactive aggressionTeacher reports of physical and relational aggressionTeacher reports of proactive or reactive aggressionTeacher reports of peer rejectionStudent-teacher conflict

Note. The check mark indicates when each measure was collected. The number of girls is presented in parentheses.

132 (69)yy

116(62)yy

yy

66 (33)y

y

assessment period in the spring (2003). A sub-group of the children (n = 29; 18 girls) con-tinued participating in the study into the fall(2003) of a second year (Time 3) and thesedata are used only for assessing the validity ofthe observational methods. An additional 37children (15 girls) were also added during thefall of the second year for the validity assess-ment. Thus, for validity purposes, observa-tions of aggression were conducted and teach-ers completed reports on proactive and reac-tive aggression for 66 children (33 girls) atTime 3. The attrition rate from Time 1 toTime 2 (12%) was because of children chang-ing schools or moving from the area. Therewere no differences on any key study variable(i.e., observations of relational or physical ag-gression, peer rejection or student-teacherconflict) between those 16 children who leftthe study and those who stayed in the study.

At the start of this study, the mean ageof the children was 44.37 months {SD = 9.88,range = 30-48 months). The sample was 8%African American, 8% Asian American, 75%European American, 4% Latino, and 5% otherethnic backgrounds. Based on yearly incomereported by parents and education level, chil-dren were from primarily middle-class back-grounds, with a majority of the families con-sisting of two adult caregivers. However, 29%of children's families were known to be livingbelow the federal poverty level. Twenty-threepercent of children were learning English as asecond language. The overall parental consent

rate was 73% and active written parental con-sent was obtained from all participating par-ents. This study was approved by the Univer-sity of Minnesota's Institutional ReviewBoard. Both schools were nationally accred-ited with similar contexts and demographics.That is, classrooms (n = 9) at both schoolswere multiage classrooms, with similar curric-ulum, head teacher (n = 9) experience oreducation (BA or MA as highest degree), classsize (there were between 16 and 20 childrenper class), and child-centered educational phi-losophy. See Table 1 for a timeline of wheneach measure was collected. Head teacherpackets were distributed in the middle of eachassessment period and each teacher individu-ally retumed his or her forms in a sealedenvelope to project staff. Teachers received asmall honorarium (i.e., gift card) to compen-sate them for their participation and schoolsreceived a small yearly donation to facilitatecontinuing education efforts.

The present study is unique from paststudies published with related data sets (Crick,Ostrov, Burr et al., 2006) in that the presentstudy consisted of a secondary recoding ofobservational data to include proactive andreactive functions of both physical and rela-tional aggression observations. The presentstudy sample does not entirely overlap withthe past published sample given that only 1year of the research was used in the presentstudy to maximize statistical power and to

17


avoid attrition issues that occurred across theentire study.

Measures and Procedure

Observations of aggression and vic-timization. A revised focal child observa-tional procedure (Ostrov & Keating, 2004)was adopted to record the physical and rela-tional aggression of the participants. Specifi-cally, at each time point, eight independent10-min ohservations during a 2- to 3-monthperiod were collected for each child for a totalof 160 min of observation per child and a totalof 18,560 min (309 hr) of observations overall(plus five, 280-min of observations for validitypurposes) for the present study. There weretypically two observers present for classroomobservations; occasionally three or four ob-servers were present during playground obser-vations. A total of 10 male and female observ-ers were used for Times 1 and 2 (two observ-ers were new at Time 2). Nine male andfemale observers, four of whom were new tothe study, were used for Time 3. All observa-tions were conducted live and no videotapeswere used in the present study. During each10-min assessment interval, which always oc-curred during free play, observers randomlyselected a focal child and recorded in detail thefocal child's engagement in the following mu-tually exclusive behaviors: (a) Physical ag-gression, defined as the use of physical forceor threat of physical force to hurt, harm, orinjure another child (e.g., hitting, shoving,kicking, pinching, pulling, punching, and tak-ing objects with the intent to hurt or harmanother person etc.); and (b) Relational ag-gression, defined as the removal, or threat ofthe removal, of the relationship as the meansof hurting, harming, or injuring another child(e.g., verbally or nonverbally excluding froman activity; using friendship withdrawal;spreading rumors, gossip, or malicious lies;and ignoring by covering ears or silenttreatment).

Consistent with past procedure, eachtime a separate and discrete behavior was ob-served the observer would record the behaviorin full written detail (e.g., "Focal child hit a

male peer in the stomach to get a teddy bear"or "Focal child kicked female peer on leg inresponse to being hit on head") under therepresentative category (e.g., physical or rela-tional aggression; Ostrov & Keating, 2004). Abreak in time was needed to indicate one be-havior from two (see Ostrov & Keating,2004). For example, a child observed as say-ing, "You are not my friend and can't come tomy birthday party," would only be counted asone event. A child who said to a peer, "Youare not my friend," and 5 min later, after aclear break in time and in interaction, told thesame or a different child, "You can't come tomy birthday party," would be counted as twoevents. Observers also recorded the responseand delineated the order of behaviors duringthe session. Observations were made withenough detail so that i;eliability checks couldbe conducted and behaviors could be com-pared. Observers were trained to recognizeand rule out rough-and-tumble play (Pelle-grini, 1989), which was identified by playful,friendly, and positive interactions without theintention to hurt (e.g., stayed together after theinteraction, did not protest, and did not displaynegative affect). No more than one observa-tion was conducted per child per day.

Observers were trained (observationmanual and forms are available upon requestto the first author) with videotapes from paststudies (Ostrov & Keating, 2004; Ostrov,Woods, Jansen, Casas, & Crick, 2004) andwith several days of live practice coding. Toavoid observer drift, observers were assessedfor reliability for the duration of each obser-vation assessment period with different ob-servers (Pellegrini, 2004). This study was con-ducted over a relatively long period of timeand in a variety of settings to minimize someof the context specificity problems of obser-vational research and to increase sampling op-portunities (Pellegrini, 2004). In addition, ob-servers spent a great deal of time in the schoolsetting with a "minimally responsive stance,"in which they controlled their nonverbal be-haviors, movements around the room or play-ground, position in the room or playground,and interactions with the children. These pro-cedures allowed the children to habituate to

28


the observers' presence (Pellegrini, 2004).Once reactivity levels were low, observationswere initiated and observers stayed within ear-shot of the focal child while controlling theireye contact, body movements, and interactionswith the child. When possible (i.e., observerscould clearly hear and see the focal child),classroom observations were conducted froma visually shielded observation booth in theclassroom to reduce reactivity levels. Reactiv-ity levels (e.g., looks, comments, questions, orbehaviors directed at the observer) were rela-tively low across the present study (e.g.,range = 0-3, M = 0.50 per session; see Atlas& Pepler, 1998).

Coded behaviors were checked by thetrainers for accuracy at the conclusion of thestudy and were summed within category.Behavior frequencies were first summed andthen converted to reflect the number of com-pleted sessions by dividing the total fre-quency counts by the number of sessions(4-8) that observations were collected foreach child. Children with less than half ofthe possible observation sessions per timepoint (i.e., <4) were dropped from that par-ticular time period of the study (i.e., thisonly affected two children in the study). AtTime 1, 99% of the focal children had 8sessions, at Time 2, 98% of the focal chil-dren had 8 completed observational sessionsand at Time 3, 94% of the participants had 8sessions (6% had 6 or 7 sessions). Thus achild with an original frequency count of 4for physical aggression, indicating that 4independent cases of physical aggressionwere observed during the 80 min of obser-vations, would receive a converted scoreof 0.50 (i.e., 4 behaviors/8 sessions).

The observational measure has demon-strated acceptable psychometric properties inthe past (see Ostrov, 2006; Ostrov & Keating,2004). This observational system has been re-viewed in the recent school psychology liter-ature, judged to meet all psychometric stan-dards of the field, and deemed appropriate foruse in multiple contexts during early child-hood (Leff & Latkin, 2005). Specifically, ev-idence for favorable interrater reliability hasbeen demonstrated in past research for physi-

cal aggression (intraclass correlation coeffi-cient [ICC] = .95), relational aggression(ICC = .93), and prosocial behavior (ICC =.90; Ostrov & Keating, 2004). Evidence forvalidity has also been established. For exam-ple, the association between observationallybased relational aggression scores and thoseobtained via teacher reports on the PreschoolSocial Behavior Scale—Teacher Fomi wasr = .50, p < .01, in the Ostrov and Keating(2004) study, and r = .40, p < .01, in anindependent sample (Ostrov, 2006). In addi-tion, observations of relational aggression dur-ing free play has tended to correspond to in-dependent observations of relational aggres-sion conducted during a semistructured task(i.e., r = .48, p < .05; Ostrov & Keating,2004).

Reliability of physical and relationalaggression observations. In the presentstudy, two raters observed the same focal childfor 15% of the time for reliability purposesacross each time. Scores are reported acrossthe year of the project for ease of communi-cation. To assess interobserver reliability,ICCs were computed between the two raters(McGraw & Wong, 1996). Reliability coeffi-cients were acceptable: physical aggression,ICC = .85, and relational aggression, ICC =.77. The manner in which the observationswere collected was not amenable to kappacoefficients because the observers did not spe-cifically record intervals when the aggressivebehaviors were absent. The use of ICCs hasbeen suggested in similar situations (seeMcGraw & Wong, 1996) and used in the past(e.g.. National Institute of Child Health andHuman Development Early Child Care Re-search Network, 2004; Ostrov, 2006).

Coding of proactive and reactive re-lational and physical aggression. Based onthe past definitions (Dodge & Coie, 1987) andrecent teacher, peer, and parent report mea-sures (e.g.. Crick & Dodge, 1996; Little, Joneset al., 2003; Poulin & Boivin, 2000), eachbehavior originally coded as physical or rela-tional aggression was recoded, using the fullwritten details originally recorded during the

29


observations, for the presence of four mutuallyexclusive categories (i.e., physical proactiveaggression, physical reactive aggression, rela-tional proactive aggression, and relational re-active aggression). Each behavior was alsoclassified according to the gender of the victim(e.g., physical proactive aggression to a male).

For proactive aggression, reference to aspecific goal, toy, object, resource, or socialposition was needed, and the behavior was notin response to a threat or retaliation for beingvictimized. Reactive aggression was based onan assessment of hostility, a threat or per-ceived threat based on the presence of specificvictimization behaviors, or a hostile exchangebased on previously recorded response codes.The order of hehaviors was used to documentthat retaliation had taken place. Observerswere trained to record the function of theaggressive behavior and to record the se-quence of events when conducting live obser-vations. To avoid errors, the observers did notlabel these behaviors as proactive or reactiveand independent highly trained coders con-ducted a secondary coding to determinewhether each original aggression behavior wasproactive or reactive.

Reliability of proactive and reactiveaggression. For 30% of the observations, atleast two observers independently coded thesame behaviors. Kappa (K) coefficients werecalculated for each of the behaviors. Kappacoefficients could be calculated for thesecodes because these data were from a second-ary coding procedure in which the coders in-dicated a precise code for each observed be-havior. Analyses were conducted by time pe-riod and at the level of the target gender (e.g.,proactive physical aggression to a male), butfor ease of communication they are only re-ported for the total categories. For proactivephysical aggression, K was >.69. For reactivephysical aggression, K was >.63. For proac-tive relational aggression, K was >.64. Forreactive relational aggression, K was >.68.Kappas above .60 are considered acceptable(Pellegrini, 2004) and therefore all reliabilitycoefficients were acceptable.

Teacher Report of Relational andPhysical Aggression

A teacher-rating measure of children'ssocial behavior (i.e.. Preschool Social Be-havior Scale—Teacher Form; Crick et al.,1997) was used to assess the validity of theobservation measures and was completed bythe head teachers in each classroom. Thisinstrument consists of 16 items, 6 of whichassess relational aggression (e.g., "Thischild tries to get others to dislike a peer"); 6of which assess physical aggression (e.g.,"This child kicks or hits others"); and 4 ofwhich measure prosocial behavior, whichserved as positive-toned filler items. Teach-ers respond to the items by rating on a5-point scale how true each item is for eachof their students. Evidence supports the fa-vorable psychometric properties (i.e., inter-nal consistency, factor analyses, and concur-rent validity) of the Preschool Social Behav-ior Scale—Teacher Form (e.g., Bonica et al.,2003; Crick et al., 1997; Hart et al., 1998;Ostrov & Keating, 2004). In the presentstudy, for relational aggression Cronbach'salpha was acceptable: .87 (Time 1) and .89(Time 2). For physical aggression Cron-bach's alpha was acceptable: .93 (Time 1)and .89 (Time 2).

Teacher Reports of Proactive andReactive Aggression

To assess the validity of the observa-tional codes, a new teacher measure (Pre-school Proactive and Reactive Aggression—Teacher Report) was developed based on a36-item adolescent self-report measure (Formsand Functions of Aggression Measure) ofphysical and relational aggression that in-cludes reactive and proactive functions (Little,Jones et al., 2003). The present measure in-cluded 12 items, 3 items for each type ofaggressive behavior: physical proactive ag-gression (e.g., "This child often hits, kicks, orpushes to get what s/he wants"), physical re-active aggression (e.g., "If other children an-ger this child, s/he will often hit, kick, orpunch them"), relational proactive aggression(e.g., "To get what s/he wants, this child will

30


often tell others that s/he won't be their friendanymore"), and relational reactive aggression(e.g., "When this child is upset with others,s/he will often ignore or stop talking tothem"). Teachers responded on a 5-point scalefrom (1) Never or Almost Never True to (5)Always or Almost Always True. Teachers com-pleted reports on proactive and reactive ag-gression for 66 children at Time 3. For thisstudy, Cronbach's alpha was acceptable foreach behavior: physical proactive aggression,a = .88; physical reactive aggression, a =.92; relational proactive aggression, a = .88;relational reactive aggression, a = .82.

Teacher Reports of Social-PsychologicalAdjustment

Teacher reports of children's social ad-justment were obtained with the peer rejectionscale of the Preschool Social BehaviorScale—Teacher Form instrument (Crick et al.,1997). The peer rejection scale consists of 2items and assesses rejection by same- andother-gender peers. Teachers respond to theseitems using the same response format de-scribed previously for the behavioral items.Values for these items are summed to indicatethe level of rejection. Favorable psychometricproperties for this scale have been demon-strated in prior research (Ostrov et al., 2004)and in the current study, with Cronbach's al-pha = .90.

As a further index of adjustment outcomesuniquely related to the school context, the focalchild's relationship with his or her teacher wasmeasured with the Student-Teacher Relation-ship Scale (Pianta et al., 1995) at Time 2. Thiswidely-used 28-item measure with acceptablepsychometric properties is designed to be aquantitative measure of teachers' perceptions oftheir relationships with students (e.g., "Thischild and I always seem to be struggling witheach other"). Only the conflict subseale was usedin the present analyses to reduce the number ofanalyses conducted. Teachers respond to theitems on this measure on a scale that rangesfrom 1 {Definitely does not apply) to 5 (Defi-nitely applies). For this study, Cronbach's alphawas .85.

Results

Descriptive statistics were calculated forkey study variables and are presented in Table2. To examine the various research questionsof the study, first validity correlations wereconducted both concurrently (Time 3) andprospectively (Time 1 to 3) to assess predic-tive validity. Second, bivariate correlationswere conducted at each time to assess theintercorrelations of proactive and reactivefunctions for each subtype of aggression (i.e.,physical and relational) as well as the stabilityof the aggression types. To investigate genderand time differences, a series of repeated mea-sure analyses of variance (physical and rela-tional aggression models were run separately)were run with follow-up tests conducted asneeded. Finally, a sedes of hierarchical regres-sion models were computed to test for theunique effects of relational aggression catego-ries (e.g., proactive relational aggression) inpredicting future social-psychological adjust-ment (i.e., peer rejection and student-teacherconflict), above and beyond physical aggres-sive behaviors (e.g., proactive physicalaggression).

Validity

Observations of relational aggressionwere correlated with teacher reports at bothtime points (r = .34, p < .001; r = .42, p <.001). For physical aggression, observationswere moderately correlated with teacher re-ports at both time points (r = .35, p < .001;r = .63, p < .001).

The concurrent validity flndings fromthe third time point indicates that there weremoderate levels of association betweenteacher and observer reports for proactive andreactive physical aggression and in part forproactive relational aggression. Specifically,teachers and observers demonstrated moderateagreement for proactive physical aggression(r = .40, p < .001) and for reactive physicalaggression (r = .43, p < .001). Teacher a'ndobserver informants did not agree regardingproactive relational aggression (r = .15, notsignificant) or reactive relational aggression(r = - .06, not significant). In an effort to

31


Table 2Descriptive Statistics for Key Variables by Gender and Time

Behavior

Proactive physical aggression(O)T1

Reactive physical aggression(O)T1

Proactive relational aggression(O)T1

Reactive relational aggression(O)T1

Proactive physical aggression{0)T2

Reactive physical aggression(0)T2

Proactive relational aggression(0)T2

Reactive relational aggression(0)T2

Peer rejection(TR) T2

Student-teacher conflict(TR) T2

Proactive physical aggression(0)T3

Reactive physical aggression(0)T3

Proactive relational aggression(0)T3

Reactive relational aggression(0)T3

Proactive physical aggression(TR) T3

Reactive physical aggression(TR) T3

Proactive relational aggression(TR) T3

Reactive relational aggression(TR) T3

Mean

0.16

0.09

0.05

0.02

0.13

0.05

0.04

0.01

3.84

23.15

0.15

0.11

0.03

0.004

4.71

7.29

5.43

7.00

Focal Boys

SD

0.20

0.15

0.10

0.07

0.15

0.13

0.07

0.04

1.76

7.74

0.11

0.21

0.07

0.02

2.51

3.84

2.01

2.28

Range

0-1.00

0-0.75

0-O.50

0-0.38

0-0.50

0-0.75

0-0.25

0-0.25

2-8

15^7

0-0.63

0-1.13

0-0.25

0-0.13

3-12

3-15

3-9

3-9

Mean

0.13

0.02

0.09

0.03

0.09

0.01

0.09

0.02

3.19

21.95

0.07

0.03

0.10

0.03

3.10

4.17

5.07

6.27

Focal Girls

SD

0.18

0.05

0.18

0.08

0.14

0.04

0.13

0.07

1.34

7.08

0.10

0.07

0.17

0.09

0.41

1.67

2.01

2.08

Range

0-0.75

0-0.25

0-1.13

0-0.38

0-0.50

0-0.25

0-0.50

0-0.50

2-8

14-47

0-0.38

0-0.25

0-0.63

0-0.38

3-5

3-10

3-9

3-9

Note. Observational descriptive statistics are based on an average of the specific behavior types across the number ofcompleted sessions (i.e., total frequency divided by the number of sessions out of eight possible sessions in whichobservational data was collected). O = observation; TR = teacher report; TI = Time 1; T2 = Time 2; T3 = Time 3.At Time 1, n = 132; at Time 2, n = 116; at Time 3, n = 66.

explore the tiature of these correlatiotis, addi- otily, teacher reports and observed proactivetiotial post hoc exploratory correlatiotis were relatiotial aggressioti were sigtiificatitly asso-conducted separately by getider. For girls ciated (r = .36, p < .05), but correlations for

32


Table 3Stability Correlations and Intercorrelations for Aggression Types

1. Proactive physicalTime I — .32***

2. Reactive physicalTime 1 —

3. Proactiverelational Time 1

4. Reactiverelational Time 1

5. Proactive physicalTime 2

6. Reactive physicalTime 2

7. Proactiverelational Time 2

8. Reactiverelational Time 2

Note. For Time 1 correlations, n = 132, For correlations at Time 2 and between Times 1 and 2. n = 116*p < .05 .**p < . 0 1 .***p < . 0 0 1 .

05

04

.05

.03

.33***

—

.38***

.28**

.04

.07

—

.21*

.21*

.17

.10

.13

—

.10

-.08

.13

.13

.03

.15

- .04

- .02

.11

.56***

-.07

.08

.15

boys (r = — .04) were not significant. For theassociation between observed and teacher-re-ported reactive relational aggression, post hoccorrelations by gender did not reveal signifi-cant correlations for girls (r = .22) or boys{r = - . 2 5 ) .

The present study revealed evidence ofpredictive validity for all of the aggression cat-egories (total observation scores were used forthe four categories of aggression). Observed re-active relational aggression at Titne 1 predictedteacher-reported reactive relational aggression atTime 3 (r = .27, p < .05). Observed reactivephysical aggression at Time 1 predicted teacher-reported reactive physical aggression at Time 3(r = .62, p < .001). Observed proactive rela-tional aggression at Time 1 tended to predictteacher-reported relational aggression at Time 3(r = .24, not significant). Given the nonsignifi-cant correlation, exploratory post hoc analyseswere conducted by gender. For girls only, ob-served proactive relational aggression at Time 1predicted teacher-reported proactive relational

aggression at Time 3 (r = .43, p < .05), butanalyses for boys (r = - .07) were not signifi-cant. Observed proactive physical aggression atTime 1 was not significantly associated withteacher-reported proactive physical aggression atTime 3 (r = .20). Exploratory post hoc analyseswere conducted by gender. For girls, observedproactive physical aggression at Time 1 wasassociated with teacher-reported proactive phys-ical aggression at Time 3 (r = .58, /? < .001), butthe correlation was not significant for boys (r =.16).

Intercorrelations

Bivariate correlations were conductedbetween observed proactive and reactive ag-gression for both physical and relational ag-gression. These correlations were computed atthe level of total behaviors (e.g., proactivephysical aggression) and were conductedwithin time points (see Table 3). Intercorrela-tions between proactive relational and proac-

33


tive physical, as well as reactive relationaland reactive physical, are also presented (seeTable 3).

Analyses were also conducted to assessthe degree of intercorrelation for proactive andreactive functions of relational and physicalaggression for the teacher-report measure ad-ministered only at Time 3. Consistent with thepast literature that almost exclusively reliedupon the teacher-report method (see Poulin &Boivin, 2000), reactive and proactive physicalaggression were highly correlated (r = .79,p < .001). In addition, reactive and proactiverelational aggression were highly correlated(r = .70, p < .001). The findings were some-what different when analyzed separately bygender. For boys, reactive and proactive phys-ical aggression were highly correlated (r =.81, /? < .001), whereas for girls the correla-tion was not significant (r = .27). For rela-tional aggression and boys, the intercorrela-tion of proactive and reactive functions washigh (r = .15, p < .001) and was similar forgirls (r = .64, p < .001).

Stability of Aggression Categories

Stability coefficients were conducted foreach of the four observed aggression behav-iors. Findings were generally similar whencontrolling for age and thus only the unad-justed coefficients are reported (see Table 3).In sum, it appears that three of the four ag-gression types are significantly stable (rangefrom r = .21, p < .05 to r = .56, p < .001).Although only proactive relational aggressionwas found not to be significantly stable, it isimportant to highlight that some of the stabil-ity correlations were small in magnitude (e.g.,reactive physical aggression).

Gender and Time Differences

Physical aggression. To test for genderand time differences for functions of physicalaggression, a 2 (focal child gender: boys andgirls) X 2 (target child gender: male and fe-male peers) X 2 (time) X 2 (function type:proactive physical or reactive physical) re-peated measures analysis of variance wasconducted with aggression frequencies as the

dependent variable. Given the number of com-parisons, a Bonferroni correction was per-formed to control for Type I error rate. How-ever, the analyses were identical with andwithout the correction, so the uncorrectedfindings are presented.

A main effect for focal child genderemerged [F(l, 113) = 6.89, p < M, t]^ =.06], which indicated that focal boys(M = 0.05; SE = 0.01) were more physicallyaggressive overall than focal girls {M = 0.03;SE = 0.01). A main effect for function typewas found [F(l, 113) = 48.03, p < .001, TI =.30], with proactive physical aggression(M = 0.06; SE = 0.01) being more frequentthan reactive physical aggression (M = 0.02;SD = 0.01). A main effect of target childgender was revealed [F(l, 113) = 8.99, p <.01, -x]^ = .07], which indicated that males{M = 0.05; SE = 0.01) received more physicalaggression than female peers (M = 0.03;SE = 0.01). A target child gender by focalchild gender interaction was revealed [F(\,113) = 17.94,/? < .001, -x]^ = .14]. For focalboys, a main effect of target child gender wasfound [E(l, 52) = 11.31, p< .001, J]^ = .25],indicating that focal boys directed more phys-ical aggression to males {M = 0.07;SE = 0.01) than to female peers (M = 0.03;SE = 0.01). The effect was not significant forgirls. It was also revealed that focal boys(M = 0.30; SD = 0.29) directed more physicalaggression to male peers than did focal girls[M = 0.10; SD = 0.19; F{1, 114) = 19.57,p < .001, -n = .15].

Relational aggression. To test forgender and time differences for functions ofrelational aggression, a 2 (focal child gen-der: boys and girls) X 2 (target child gender:male and female peers) X 2 (time) X 2(function type: proactive relational or reac-tive relational) repeated measures analysisof variance was run with aggression fre-quencies as the dependent variable. Giventhe number of comparisons, a Bonferronicorrection was adopted to control for Type Ierror rate. However, the analyses were iden-tical with and without the correction, so theuncorrected findings are presented.

34


Table 4Correlations Between Aggression Types and Outcome Variables

Aggression Type

Proactive physical aggression (O) TIReactive physical aggression (O) TIProactive relational aggression (O) TIReactive relational aggression (O) TIProactive physical aggression (O) T2Reactive physical aggression (O) T2Proactive relational aggression (O) T2Reactive relational aggression (O) T2Peer rejection (TR) T2

Note, n = 116; O = observations; TR = teacher report; STRS = Student-Teacher Relationship Scale; TI = Time1;T2 = Time 2.*p < .05.**p < .01.***p < .001.

Peer rejection(TR)Time2

.36***

.30***

.22**

.08

.19*

.12

.04

.05

STRS conflict(TR) Time 2

.47***

.43***

.30**

.20*

.43***

.19*- .10

.23**49***

A main effect for focal child gender wasfound [F(l, 114) = 5.56, p< .05, T]^ = .05],which indicated that focal girls (M = 0.03;SE = 0.01) were more relationally aggressiveoverall than were focal boys {M = 0.01;SE = 0.004). A main effect for function typewas also found [F(l, 114) = 27.19, p < .001,r\^ = .19], which revealed that proactive rela-tional aggression {M = 0.03; SE = 0.004) wasmore frequent than reactive relational aggres-sion {M = 0.01; SE = 0.004). Finally, a func-tion type by focal child gender interaction wasfound [F(l, 114) = 5.28, p< .05, T]^ = .04].For focal boys, a main effect of function typeemerged [F(l, 53) = 7.29, p < .01, T]^ = .12],indicating that boys displayed more proactive{M = 0.05; SE = 0.01) than reactive(M = 0.01; SE = 0.004) relational aggression.For focal girls, a main effect of function typealso was found [F(l, 61) = 21.72, p < .001,T| = .26], revealing that girls displayed moreproactive (M = 0.02; SE = 0.01) than reactive(M = 0.0\; SE = 0.004) relational aggression.Finally, focal girls {M = 0.19; SD = 0.24)were more proactive relationally aggressivethan were focal boys \M = 0.09; SD = 0.14),F(\, 114) = 7.57, p < .01, Ti = .06].

Regression Models: Predicting FutureSocial-Psychological Adjustment

To test the bivariate association betweenaggression types and the two outcome mea-sures (i.e., peer rejection and student-teacherconflict), a sedes of bivariate correlationswere conducted (see Table 4). Most of theTime 1 aggression types were significantlyassociated with Time 2 adjustment outcomes.Next, to test for the unique role of aggressiontypes in the prediction of these outcome vari-ables, a sedes of hierarchical multiple regres-sion equations were conducted in whichTime 2 adjustment scores served as the depen-dent vadable. In the first set of analyses (Mod-els 1 and 2), observed proactive functions ofrelational aggression were entered first andobserved reactive relational aggression wasentered second to test the hypothesis that re-active relational aggression would add vad-ance above and beyond the role of proactiverelational aggression in predicting future ad-justment problems. The third and fourth mod-els test for the unique effects of proactiverelational aggression when controlling for re-active relational aggression. In the second setof analyses (Models 5-8), observed function

35


types of physical aggression were entered atStep 1 and observed relational aggressionfunction types were entered at Step 2 to testthe prediction that proactive and reactive rela-tional aggression (in separate models) wouldadd significant information in the prediction offuture adjustment problems (i.e., peer rejec-tion and student-teacher conflict). Interactionterms with gender and predictors were enteredin the regression models and were not signif-icant, so they were dropped from further dis-cussion. In all models, however, gender wasentered at Step 1 and served as a covariate (seeTable 5 for all eight regression models andfindings).

In summary, it was found that proactiverelational aggression predicted both futurepeer rejection and student-teacher confiictwhen controlling for gender and concurrentreactive relational aggression. Proactive rela-tional aggression also uniquely predicted bothfuture peer rejection and student-teacher con-fiict above and beyond the role of proactivephysical aggression and gender. Reactive re-lational aggression uniquely predicted stu-dent-teacher confiict above and beyond therole of reactive physical aggression.

Discussion

The current study had four main goalsfor expanding the school and developmentalpsychology literatures. The first was to assessthe reliability and validity of a new school-based observational classification system thatdifferentiates subtypes of physical and rela-tional aggression by two common functions ofaggression (i.e., proactive and reactive) duringearly childhood. The second goal was to in-vestigate the stability of these four categories,across an academic year, using observationalmethods during early childhood. The third ma-jor goal was to investigate gender differencesacross time for these behaviors with a morefunction-specific classification system. Thefourth goal was to test the association betweenthese categories and future adjustment prob-lems, which included an assessment of bothpeer rejection and conflict within the student-teacher relationship.

It was documented that the observa-tional coding methods were reliable, as indi-cated by two methods of interrater reliability.Given that the findings were consistent acrosstime and across multiple coders, it is con-tended that the training and codes were suffi-cient to reliably capture these behaviors. Inaddition, the assessments of validity, using anewly adapted teacher-report measure that in-cluded reliable subscales of proactive and re-active relational and physical aggression, sug-gested that both teachers and observers signif-icantly agreed in many cases. Despite theevidence for concurrent validity (i.e., signifi-cant and moderate agreement between teach-ers and observers) for global physical andrelational aggression variables, the correla-tions for function types of aggression withinthe smaller validity sample (at Time 3) werenot always significant. For example, proactiverelational aggression was only significant forgirls in the concurrent and predictive validityanalyses, and reactive relational aggressionwas only significant in the predictive validityanalyses. These findings await replication witha larger validity sample; however, they sug-gest that some caution may be warranted. Inparticular, proactive relational aggression maybe challenging to observe among boys. De-spite these cases, there were sufficient corre-lations to indicate that the overall observa-tional scheme demonstrated preliminary evi-dence of validity. The present methods werebased on theory (see Dodge, 1991) and priorempirical work that demonstrated acceptablefactor loadings, internal consistency, and dis-criminate validity for the proactive and reac-tive aggression constructs (Dodge & Coie,1987; Hubbard et al., 2002; Little, Jones et al.,2003; Little, Brauner et al., 2003; Poulin &Boivin, 2000). Thus, the observational meth-ods and teacher reports appear to have accept-able psychometric properties, as suggested bya recent review in the school psychology lit-erature (Leff & Lakin, 2005), and the newobservational codes of aggressive functionsappear to be reliable and to demonstrate pre-liminary evidence of vahdity.

The intercorrelations of observed proac-tive and reactive aggression provide further

36


Table 5Hierarchical Regression Models: Unique Associations Between Aggression

Types Predicting Future Adjustment Problems

Outcome, Step, Predictors p F, AF /f A/f

I. Student-Teacher ConflictTime 21. Proactive Ragg Time 1 .32*** F(2, 112) = 6.34, p < .01 .10

gender —.132. Reactive Ragg Time 1 .11 AF(1, 109) = 1.39, ns .01

IT. Peer rejection Time 21. Proactive Ragg Time 1 .26** F(2, 114) = 6.71, p < .01 .11

gender -.24**2. Reactive Ragg Time 1 .00 AF(1, 113) = 0.00, ns .00

m. Student-teacher conflictTime 21. Reactive Ragg Time 1 .21* F(2, 110) = 2.85, p < .06 .05

gender — .092. Proactive Ragg Time 1 .28** AF(I, 109) = 8.05,p < .01 .07

IV. Peer rejection Time 21. Reactive Ragg Time I .09 F(2, 114) = 2.94, p < .06 .05

gender - . 2 1 *2. Proactive Ragg Time 1 .26** AF(1, 113) = 7.10,p < .01 .06

V. Student-teacher conflictTime 21. Proactive Pagg Time 1 .46*** F(2, 112) = 15.32, p < .001 .22

gender - .052. Proactive Ragg Time 1 .28*** AF(1, 109) = 11.64, p < .001 .08

VI. Peer rejection Time 21. Proactive Pagg Time 1 .34*** F(2, 114) = 10.60, p < .001 .16

gender - .172. Proactive Ragg Time 1 .23** AF(1, 113) = 7.33,p < .01 .05

VII. Student-teacher conflictTime 21. Reactive Pagg Time 1 44*** F(2, 110) = 12.53, p < .001 .19

gender .052. Reactive Ragg Time 1 .18* AF(1, 109) = 4.38, p < .05 .03

VTII. Peer rejection Time 21. Reactive Pagg Time 1 .26** F(2, 114) = 6.57, p < .01 .10

gender - .132. Reactive Ragg Time 1 .07 AF(1, 113) = 0.61), ns . .01

Note. Pagg = physical aggression; Ragg = relational aggression; ns = not significant. Gender: Boys = 1, Girls = 2.*p < .05.** p < .01.***p < .001.

evidence for the utility of these constructs, as These findings were in contrast to past physi-they were found to be relatively independent cal aggression research, which has docu-with only slight to moderate correlations be- mented quite substantial overlap between thetween the various subscales across the study, proactive and reactive physical aggression

37


constructs (see Dodge & Coie, 1987; Vitaro etal., 1998). The present findings are the first todelineate the influence of both physical andrelational aggression from these constructs(i.e., proactive and reactive functions of ag-gression) using observational data, which mayexplain the substantially lower correlations. Infact, the intercorrelations among the teacherreport measure, in the validity subsample atTime 3, indicated much higher (as expected)effects. Interestingly, the finding that the pro-active and reactive functions were either notcorrelated or demonstrated low correlations isconsistent with the other known observationalstudies of proactive and reactive physicalaggression (see Price & Dodge, 1989). Thus,the high level of intercorrelation between pro-active and reactive functions of aggressionmay be an artifact of the measurement tool andshared method variance, and future observa-tional methods are needed to replicate thiseffect. Future research is needed to replicatethese findings with myriad informants andmeasurement techniques during earlychildhood.

In general, proactive behaviors weremore common than reactive, consistent withpast research, which has documented the sa-lience of preferred objects and resources dur-ing early childhood play (see Cummings et al.,1989). In alignment with predictions, the as-sessments of gender differences indicated thatgirls were more relationally aggressive, re-gardless of the proactive or reactive functionof the behavior. Similarly, boys were morephysically aggressive, across both types ofaggressive functions. Not surprisingly, thesefindings suggest that gender differences existat the specific level, of the recipient of theaggression (i.e., target child gender: male andfemale peers).

Knowing to whom the behavior was di-rected in this study was meaningful informa-tion. It was revealed that focal boys displayedmore physical aggression to male peers thanthey did to female peers and focal boys deliv-ered more physical aggression to male peersthan did focal girls. In general, these findingsprovide further evidence of the importance ofstudying aggressive behavior at the level of

the target child's gender (see also Ostrov &Keating, 2004; Pellegrini & Long, 2003). Thisfurther specificity in behavioral interactions isimportant because children during early child-hood may leam which behaviors are moreappropriate and effective for their own gender,and which tactics are most useful when inter-acting with other-gender peers. These issuesshould be explicitly addressed in future re-search during early childhood. In addition,future research should investigate the role thatgender schemas and gender role identity playin the development of subtypes of aggressivebehavior. Presumably, a gender difference inrelationally aggressive behaviors would onlybe apparent after children have an understand-ing of their own gender membership and thesocial characteristics of their gender group.These developmental precursors have been as-sumed to exert an important infiuence, butthey are unanswered questions.

Past research has documented stabilityfor girls and boys in physically aggressivebehavior (see Cairns, Cairns, Neckerman, Fer-guson, & Gariepy 1989; Coie & Dodge, 1998)and for relational aggression over time forschool-aged children (Crick, 1996; Crick, Os-trov, & Werner, 2006), but the developmentalliterature has not previously documented sta-bility in these behaviors during early child-hood. In addition, no known studies of stabil-ity have been conducted that have assessed,via observations, proactive and reactive func-tions of relational and physical aggression.This study revealed that low to moderate lev-els of stability emerged for proactive and re-active functions of physical aggression. Simi-lar findings were revealed for reactive func-tions of relational aggression. However,stability was not seen for all behaviors. Thatis, proactive relational aggression was not sig-nificantly stable and future research is neededto explore why there was no continuity in thedisplay of this particular type of aggressionacross the academic year. In general, the find-ings suggest that these behaviors are consis-tent across time and further indicate that with-out intervention these behaviors may continue.Moreover, these behaviors will most likelycontinue to be displayed and transferred to

38


different peer groups across time. To date, nolong-term prospective studies have been con-ducted that track relational aggression fromearly childhood into middle childhood. Thefield of school psychology needs these typesof studies to understand the continuity or law-ful discontinuity that may be present for thesebehaviors.

In general, the findings, to some extent,revealed that proactive and reactive functionsof relational aggression are uniquely associ-ated with future adjustment problems. Specif-ically, analyses revealed that relational aggres-sion categories predicted future adjustmentproblems (e.g., peer rejection and student-teacher confiict) above and beyond the influ-ence of physical aggression categories, whichis consistent with findings with school-agedsamples (see Crick, 1996; Crick, Ostrov, &Wemer, 2006). In addition, the analyses re-vealed that individual function types of rela-tional aggression (i.e., proactive relational ag-gression) had unique predictive ability evenwhen controlling for the other function type.These initial findings during early childhoodawait replication, but they indicate that func-tions of proactive and reactive aggression mayhave utility in the differential prediction offuture adjustment problems (see Poulin &Boivin, 2000; Vitaro et al., 1998; cf. Bushman& Anderson, 2001; Waschbusch et al., 1998)for girls and boys. It is important to reveal thatnot all findings demonstrated unique effects—for example, both proactive and reactive rela-tional aggression were associated with futurelevels of student-teacher conflict above andbeyond the role of the respective physical ag-gression type. These findings suggest that thestudent-teacher relationship may also beharmed when girls display relational aggres-sion at school. Given that children who haveconflictual relationships with their teachers areat increased risk for a host of academic andbehavioral problems as well as special educa-tion referral (Pianta, 1997), it is recommendedthat future research replicate and continue toexplore this important finding. Despite someevidence for unique effects, caution should beexercised when interpreting these study find-ings, as there were several null effects that

deserve attention. For example, reactive rela-tional aggression was not predictive of futurepeer rejection, when controlling for the infiu-ence of proactive relational aggression andreactive physical aggression (in separate mod-els). Reactive relational aggression also didnot predict student-teacher conflict when con-trolling for proactive relational aggression. Toreduce the number of analyses, proactivephysical aggression was not entered as a co-variate because it did not overlap in form orfunction with reactive relational aggression.Thus, future school-based research is clearlyneeded to continue to test the unique predic-tive role of the functions types of physical andrelational aggression during early childhood.

The present study had several limita-tions that should be addressed in future re-search. First, in order to expand the study ofrelational aggression to younger ages, childrenas young as 30 months were included. Argu-ably, there are qualitative differences in thetypes of relationally aggressive behaviors thata 30-month-old and a 50-month-oId woulddisplay. Future research is needed to investi-gate this possibility. Second, the sample (i.e.,116) resulted in somewhat diminished powerfor testing some of the effects when analyzingby gender (i.e., 62 girls and 54 boys). In ad-dition, to maximize the sample size this studywas not able to explore effects beyond 1 aca-demic year; future research is needed with alonger time frame to explore the prospectiveassociations between variables as well as thestability of aggressive behaviors across timeand in particular as children make the transi-tion, from early childhood centers to kinder-garten. Third, because family income datawere collected from deidentified school re-ports, it is not possible to determine if thechildren of families living below the federalpoverty level might have also been those thatwere learning English as a second language,which is a potential threat to the external va-lidity of the sample. Fourth, the effect sizes inthe present study, although in keeping withpast studies with similar samples and studieswith older children, were for the most partsmall in magnitude and must be replicated infuture work. A fifth limitation concems the

39


low level of aggressive behaviors observedacross the study. In reviewing the descriptivestatistics in Table 2, it is apparent that evenwhen accounting for the converted nature ofthe means (i.e., multiply the means by the totalnumber of sessions, or eight sessions in mostsituations), the means are often below 1.0.These findings of low base rates of aggressivesubtypes are consistent with past observationalwork (see Leff & Lakin, 2005), similar to paststudies using the current observational scheme(e.g., Ostrov & Keating, 2004; Ostrov, 2006)and in keeping with other school-based obser-vational schemes for recording relational andphysical aggression in early childhood (Mc-Evoy et al., 2003; McNeilly-Choque et al.,1996). To avoid these methodological prob-lems, future school-based observational stud-ies should rely on multiple methods of assess-ment, including semistmctured confiict stimu-lating observational tasks (e.g., Ostrov et al.,2004). In addition, there is the possibility ofsome outlier effects and caution should beexercised when interpreting the present find-ings. Additional research with clinical sampleswill be needed to determine the clinical sig-nificance and utility of the observationalscheme. Sixth, the present study did not in-clude an assessment of verbal aggression (i.e.,mean names and verbal insults; cf. Ostrov &Keating, 2004) in order to reduce the numberof analyses that were run and in tum controlfor Type I error. Verbal aggression has beenfound to be an important variable for predict-ing adjustment problems for both girls andboys in past research (e.g., Ostrov et al.,2004). In addition, the association betweenfunctions of relational and physical aggressionand functions of verbal aggression are impor-tant issues for future research. Finally, a lim-ited range of adjustment problems was in-cluded to reduce the number of analyses, andfuture studies should include more indices ofpsychopathology that are specifically relevantfor functioning both within school settings(i.e., attention deficit hyperactivity disorder;Zalecki & Hinshaw, 2004) and in othercontexts.

Implications for School PsychologyPractice

School practitioners are well equippedto use the present observational scheme intheir current assessment and evaluation of ag-gressive youth. The basic observationalscheme used in the present study was designedfor ease of use and has been positively evalu-ated by school psychology scholars for use inearly childhood classroom and playgroundcontexts (see Leff & Latkin, 2005). The ob-servational methods described in the presentstudy provide a level of analysis that shortscreening instruments and teacher reportmethods typically do not capture. For exam-ple, the" present observational procedures ac-count for the gender of the victim and thesequence of events (e.g., if relational aggres-sion follows relational victimization), whichare important criteria to target when generat-ing intervention plans. Although basic trainingand time are involved in conducting observa-tional assessments, they may be conductedwith limited fiscal costs and low levels ofreactivity; further, as part of a comprehensiveevaluation, they may serve as valid indicatorsfor the need for school-based peer interventionor as a reliable measure of intervention effec-tiveness in ongoing school-based preventionand intervention efforts. Finally, the presentstudy underscores the need for school-basedintervention programs during early childhood,which are specifically designed to addressboth proactive and reactive functions of phys-ical and relational aggression.

The current findings advance the field inseveral ways. First, the present results suggestsome utility for the distinction betweenschool-based proactive and reactive aggres-sion during early childhood. During earlychildhood these behaviors appear to be reli-ably detected, stable, only slightly correlated,and in some cases differentially predictive ofadjustment problems. The current findingstaken together provide additional confirmationof the importance of including assessments ofboth form and function (Little, Jones et al.,2003; Little, Brauner et al., 2003) across de-velopmental periods. In general, this multim-

40


ethod, multiinformant school-based study sug-gests that proactive and reactive functions ofrelational and physical aggression subtypesmay be harmful and potentially predictive ofincreased risk for future psychopathologicalsymptoms for girls and boys during the earlyschool years, which further indicates the needfor empirically based preventative interven-tions for aggression across developmental pe-riods (Leff, Power, Manz, Costigan, & Na-bors, 2001).

References

Atlas, R. S., & Pepler, D. J. (1998). Observations ofbullying in the classroom. Journal of Educational Re-search. 92, 86-99.

Bonica, C , Yeshova, K., Amold, D. H., Fisher, P. H., &2^1jo. A. (2003). Relational aggression and languagedevelopment in preschoolers. Social Development, 12,551-562.

Bushman, B. J., & Anderson, C. A. (2001). Is it time topull the plug on the hostile versus instrumental aggres-sion dichotomy? Psychology Review, 108, 273-279.

Caims, R. B., Caims, B. D., Neckerman, H. J., Ferguson,L. L., & Gariepy, J. L. (1989). Growth and aggression:Childhood to early adolescence. Developmental Psy-chology, 25. 320-330.

Cillessen, A. H. N., & Mayeux, L. (2004). From censureto reinforcement: Developmental changes in the asso-ciation between aggression and social status. ChildDevelopment, 75, 147-163.

Coie J. D..& Dodge, K. A. (1998). Aggression and anti-social behavior. In N. Eisenberg (Ed.), Handbook ofchild psychology: Vol. 3. Social, emotional, and per-sonality development (pp. 779-862). New York: JohnWiley.

Copeland-Mitchell, J., Denham, S. A., & DeMulder, E. K.(1997). Q-sort assessment of child-teacher attachmentrelationships and social competence in the preschool.Early Education and Development, 8, 27-39.

Crick, N. R. (1996). The role of overt aggression, rela-tional aggression, and prosocial behavior in the predic-tion of children's future social adjustment. Child De-velopment, 67, 2317-2327.

Crick, N. R., Casas, J. F., & Mosher, M. (1997). Rela-tional and overt aggression in preschool. Developmen-tal Psychology, 33, 579-587.

Crick, N. R., & Dodge, K. A. (1996). Social information-processing mechanisms in reactive and proactive ag-gression. Child Development, 67, 993-1002.

Crick, N. R., & Grotpeter, J. K. (1995). Relational aggres-sion, gender, and social-psychological adjustment.Child Development, 66, 710-722.

Crick, N. R., Ostrov, J. M., Appleyard, K., Jansen, E., &Casas, J. F. (2004). Relational aggression in earlychildhood: You can't come to my birthday party un-less In M. Putallaz & K. Bierman (Eds.), DukeSeries in Child Development and Public Policy: Vol. I.Aggressive, antisocial behavior, and violence amonggirls: A developmental perspective, (pp. 71-89). NewYork: Guilford Publications.

Crick, N. R., Ostrov, J. M., Burr, J. E., Jansen, E. A,Cullerton-Sen, C , & Ralston, P. (2006). A longitudinalstudy of relational and physical aggression in pre-school. Journal of Applied Developmental Psychol-ogy. 27, 254-268.

Crick, N. R., Ostrov, J. M., & Wemer, N. E. (2006). Alongitudinal study of relational aggression, physicalaggression and children's social-psychological adjust-ment. Journal of Abnormal Child Psychology, 34.131-142.

Crick, N. R., & Zahn-Waxler, C. (2003). The develop-ment of psychopathology in females and males: Cur-rent progress and future challenges. Devetopment andPsychopathology, 15. 719-742.

Cullerton-Sen, C , & Crick, N. R. (2005). Understandingthe effects of physical and relational victimization: Theutility of multiple perspective in predicting social-emotional adjustment. School Psychology Review, 34.147-160.

Cummings, E. M., Iannotti, R. J., & Zahn-Waxler, C.(1989). Aggression between peers in early childhood:Individual continuity and developmental change. ChitdDevelopment, 60, 887-895.

Dodge, K. A. (1991). The structure and function of reac-tive and proactive aggression. In D. J. Pepler & K. H.Rubin (Eds.), The devetopment and treatment of child-hood aggression (pp. 201-218). Hillsdale, NJ: Law-rence Eribaum Associates.

Dodge, K. A., & Coie, J. D. (1987). Social-informationprocessing factors in reactive and proactive aggressionin children's peer groups. Journat of Personatity andSocial Psychology. 53. 1146-1158.

Dodge, K. A., Lochman, J. E., Hamish, J. D., Bates, J. E.,& Pettit, G. S. (1997). Reactive and proactive aggres-sion in school children and psychiatrically impairedchronically assaultive youth. Journat ofAbnormat Psy-chology. 106. 37-51.

Hamre, B. K., & Pianta, R. C. (2001). Early teacher-childrelationships and the trajectory of children's schooloutcomes through eight grade. Child Development, 72.625-638.

Hart, C. H., Nelson, D. A., Robinson, C. C , Olsen, S. F.,& McNeilly-Choque, M. K. (1998). Overt and rela-tional aggression in Russian nursery-school-age chil-dren: Parenting style and marital linkages. Devetop-mental Psychology. 34. 687-697.

Hawley, P. H. (2003). Strategies of control, aggression,and morality in preschoolers: An evolutionary perspec-tive. Journal of Experimental Child Psychotogy. 85.213-235.

Hubbard, J. A., Dodge, K. A., Cillessen, A. H. N., Coie,J. D., & Schwartz, D. (2001). The dyadic nature ofsocial information processing in boys' reactive andproactive aggression. Journat of Personatity and So-ciat Psychotogy. 80. 268-80.

Hubbard, J. A., Smithmyer, C. M., Ramsden, S. R.,Parker, E. H., Flanagan, K. D., Dearing, K. F., Relyea,N., & Simons, R. F. (2002). Observational, physiolog-ical, and self-report measures of children's anger: Re-lations to reactive versus proactive aggression. ChitdDevetopment. 73. 1101-1118.

Hughes, J. N., Cavell, T. A., & Jackson, T. (1999). Influ-ence of the teacher-student relationship on childhoodconduct problems: A prospective study. Journal ofCtinicat Chitd Psychotogy. 28, 173-184.

41


Hughes, J. N., Cavell, T. A., & Willson, V. (2001).Further support for the developmental significance ofthe quality of the teacher-student relationship. Journalof School Psychology, 39, 289-301.

Keenan, K., & Shaw, D. (1997). Developmental and so-cial influences on young girls' early problem behavior.Psychologicat Bulletin. 121, 95-113.

Kupersmidt, J. B., Bryant, D., & Willoughby, M. T.(2000). Prevalence of aggressive behaviors among pre-schoolers in Head Start and community child careprograms. Behavioral Disorders, 26. 42-52.

Leff, S. S., & Lakin, R. (2005). Playground-based obser-vational systems: A review and implications for prac-titioners and researchers. School Psychology Re-view, 34, 475-489.

Leff, S. S., Power, T. J., Manz, P. H., Costigan, T. E., &Nabors, L. A. (2001). School-based programs foryoung children: Current status and implications forviolence prevention. School Psychology Review. 30,343-360.

Little, T. D., Brauner, J., Jones, S. M., Nock, M. K., &Hawley, P. H. (2003b). Rethinking aggression: A ty-pological examination of the functions of aggression.Merrill Palmer Quarterly. 49. 343-372.

Little, T. D., Jones, S. M., Henrich, C. C , & Hawley,P. H. (2003a). Disentangling the "whys" from the"whats" of aggressive behavior. International Journalof Behavioral Development. 27. 122-183.

McEvoy, M. A., Estrem, T. L., Rodriguez, M. C , &Olson, M. L. (2003). Assessing relational and physicalaggression among preschool children: Inter-methodagreement. Topics in Early Childhood Speciat Educa-tion. 23, 53-63.

McGraw, K. C , & Wong, S. P. (1996). Forming infer-ences about some intra-class correlation coefficients.Psychotogicat Methods. 1. 30-46.

McNeilly-Choque, M. K., Hart, C. H., Robinson, C. C ,Nelson, L., & Olsen, S. F. (1996). Overt and relationalaggression on the playground: Correspondence amongdifferent informants. Journat of Research in ChildhoodEducation, 11, 47-67.

Meehan, B. T., Hughes, J. N., & Cavell, T. A. (2003).Teacher-student relationships as compensatory re-sources for aggressive children. Child Develop-ment. 74. 1145-1157.

Moffitt, T. E. (1993). Adolescent-limited and life-coursepersistent antisocial behavior: A developmental taxon-omy. Psychologicat Review, 100. 674-701.

National Institute of Child Health and Human Develop-ment Early Child Care Research Network. (2004). Tra-jectories of physical aggression from toddlerhood tomiddle childhood. Monographs of the Society for Re-search in Chitd Development, 69, (Serial No. 278).

Ostrov, J. M. (2006). Deception and subtypes of aggres-sion in early childhood. Journal of Experimental ChildPsychology. 93. 322-336.

Ostrov, J. M., & Keating, C. F. (2004). Gender differencesin preschool aggression during free play and structuredinteractions: An observational study. Social Devetop-ment. 13. 255-277.

Ostrov, J. M., Woods, K. E., Jansen, E. A., Casas, J. F., &Crick, N. R. (2004). An observational study of deliv-

ered and received aggression, gender, and social-psy-chological adjustment in preschool: "This white crayondoesn't work." Early Chitdhood Research Quar-terty. 19, 355-371.

Pellegrini, A. D. (1989). Categorizing children's rough-and-tumble play. Play & Cutture. 2. 48-51.

Pellegrini, A. D. (2004). Observing children in their nat-urat worlds: A methodological primer (2nd ed.). Mah-wah, NJ: Eribaum.

Pellegrini, A. D., & Long, J. D. (2003). A sexual selectiontheory longitudinal analysis of sexual segregation andintegration in early adolescence. Journat of Experi-mentat Chitd Psychotogy. 85, 257-278.

Pianta, R. C. (1997). Adult-child relationship processesand early schooling. Early Education and Develop-ment. 8. 11-26.

Pianta, R. C , Steinberg, M. S., & Rollins, K. B. (1995).The first two years of school: Teacher-child relation-ships and deflections in children's classroom adjust-ment. Devetopment and Psychopathotogy. 7, 295—312.

Pianta, R. C , & Stuhlman, M. W. (2004). Teacher-childrelationships and children's success in the first years ofschool. Schoot Psychology Review. 33. 444-458.

Poulin, F., & Boivin, M. (2000). Reactive and proactiveaggression: Evidence of a two-factor model. Psycho-logical Assessment, 12, 115-122.

Price, J. M., & Dodge, K. A. (1989). Reactive and pro-active aggression in childhood: Relations to peer statusand social context dimensions. Journat of AbnormatChitd Psychotogy. 17. 455-471.

Prinstein, M. J., & Cillessen, A. H. N. (2003). Forms andfunctions of adolescent peer aggression associated withhigh levels of peer status. Merritt-Patmer Quar-terty, 49, 310-

Vitaro, F., Gendreau, P. L., Tremblay, R. E., & Oligny, P.(1998). Reactive and proactive aggression differen-tially predict later conduct problems. Journal of ChildPsychology and Psychiatry, 39, 377-385.

Waschbusch, D. A., Willoughby, M. T., & Pelham,W. E. (1998). Criterion validity and the utility ofreactive and proactive aggression: Comparisons toattention deficit hyperactivity disorder, oppositionaldefiant disorder, conduct disorder, and other mea-sures of functioning. Journal of Ctinicat Chitd Psy-chotogy. 27, 396-405.

Young, E. L., Boye, A. E., & Nelson, D. A. (2006).Relational aggression: Understanding, identifying, andresponding in schools. Psychotogy in the Schoots, 43,297-312.

Zalecki, C. A., & Hinshaw, S. P. (2004). Overt andrelational aggression in girls with attention deficit hy-peractivity disorder. Journat of Ctinicat Chitd & Ado-tescent Psychotogy, 33, 125-137

Zahn-Waxler, C. (1993). Warriors and worriers: Genderand psychopathology. Development and Psychopathol-ogy. 5, 79-89.

Date Received: June 24, 2006Date Accepted: November 20, 2006

Action Editor: Shane Jimerson •

42


Jamie M. Ostrov, PhD, is assistant professor of psychology in the Clinical PsychologyProgram at the University at Buffalo, The State University of New York. He is also thedirector of the University at Buffalo Social Development Laboratory. He is on theeditorial boards or is a consulting editor of Child Development. Social Development,Journal of Clinical Child and Adolescent Psychology, Journal of Youth and Adolescence,Early Education and Development, and Early Childhood Research Quarterly. His re-search focuses on the development of relational and physical aggression during early andmiddle childhood.

Nicki R. Crick, PhD, is director and professor at the Institute of Child Development,University of Minnesota. She is a fellow of the American Psychological Association andhas received several other professional honors including the Distinguished ScientificAward for Early Career Contribution to Psychology from American Psychological As-sociation (2002), the Boyd McCandless Award from American Psychological Association(1995), a Faculty Scholars Award from the William T. Grant Foundation (1995-2000),and a First Independent Research Support and Transition (FIRST) award from theNational Institute of Mental Health (1995-2000). She is also a Distinguished McKnightUniversity Professor and a Birkmaier Professor of Educational Leadership. Her researchfocuses on the antecedents, correlates, and consequences of relational and physicalaggression and victimization across the life span.

43

Ostro & Crick, 2007

Documents

school classroom

aggression categories

school psychology review

extensive research of

utility ofincluding

young sample

university ofminnesota

personthis study