18

Click here to load reader

Observed Peer Victimization During Early Elementary School: Continuity, Growth, and Relation to Risk for Child Antisocial and Depressive Behavior

Embed Size (px)

Citation preview

Page 1: Observed Peer Victimization During Early Elementary School: Continuity, Growth, and Relation to Risk for Child Antisocial and Depressive Behavior

Observed Peer Victimization During Early Elementary School: Continuity,

Growth, and Relation to Risk for Child Antisocial and Depressive Behavior

James Snyder, Monica Brooker, M. Renee Patrick, Abigail Snyder, Lynn Schrepferman, and Mike Stoolmiller

The rate at which 266 boys and girls ages 5 to 7 years old were victimized by peers was observed on multipleoccasions in kindergarten and first grade. Individual differences in victimization were observed at kindergartenentry and in growth over the subsequent 2 years. Victimization increased for some children but decreased forothers. Growth in victimization was reciprocally related to growth in teacher-reported antisocial and depressivebehavior for boys. For girls, kindergarten victimization was related to growth in parent-reported antisocialbehavior, teacher-reported depressive behavior to growth in victimization, and growth in victimization toparent-reported depression. At a short-term group level, antisocial behavior had a lagged suppressive effect onvictimization for boys but a facilitating effect for girls.

Aggression and victimization in child and adoles-cent peer groups are significant health problems thatcompromise the safety and development of children.Previous research on peer victimization has primar-ily relied on relatively global child self-reports andpeer reports (Ladd & Kochenderfer-Ladd, 2002).This precludes estimates of the absolute frequencyof and growth in peer victimization. This reportextends previous empirical efforts by applyingobservational methods to the assessment of victimi-zation at the transition to elementary schoolFaperiod when the quality of peer social relations mayafford longer term risk and protection for subse-quent development (Ladd, Birch, & Buhs, 1999).Observational measures are used to examine thecross-occasion consistency, developmental continu-ity, and growth of peer victimization. Causal modelsare tested to ascertain the relationship of peervictimization and child overt antisocial and depres-sive behavior. The phenomenology of peer victimi-zation and its relation to child adjustment arecompared by gender.

Potential Utility of Observation of Peer Victimization

Child self-report and peer nominations are com-monly used to identify children who are chronicallyvictimized by peers and to ascertain the negativeconsequences of victimization in terms of socialwithdrawal, loneliness, peer social problems, con-duct problems, and poor school adjustment (Hanish& Guerra, 2002; Kochenderfer & Ladd, 1996a; Laddand Kochenderfer-Ladd, 2002). In terms of psycho-metric characteristics and predictive validity, thesemeasurement strategies appear to work well bysecond or third grade and can be expanded into amulti-informant approach that includes teacher andparent report. However, peer nominations and selfreports of victimization during kindergarten andfirst grade do not converge, show limited continuity,and predict child adjustment less powerfully (Ladd& Kochenderfer-Ladd, 2002).

Young children may be relatively unable toaccurately distill the consistency with which victi-mization is experienced by individuals in the contextof the frequent, brief disputes that character-ize newly forming peer groups in early element-ary school (Buhs & Ladd, 2001; Parker, Rubin, Price,& DeRosier, 1995). Aggression and victimization inthese groups may be highly situational, and onlyover time does the field of children who areconsistently victimized become narrowed on thebasis of ongoing experience (Perry, Perry, & Boldizar,1990). However, chronic victimization statusemerges during this early dynamic period as asubset of children get caught in a vicious cycle inwhich victimization and maladjustment feed off oneanother (Hodges & Perry, 1999). The impact of peer

r 2003 by the Society for Research in Child Development, Inc.All rights reserved. 0009-3920/2003/7406-0020

James Snyder, Monica Brooker, M. Renee Patrick, AbigailSnyder, Lynn Schrepferman, Department of Psychology, WichitaState University; Mike Stoolmiller, Oregon Social Learning Center.This research was supported by National Institute of Mental

Health Grant R01 57342, ‘‘Child Conduct Problems: CompetingSocialization Models.’’ The contributions of the school involvedare gratefully recognized, especially the support of SusanneSmith, Cindy Schafer, and Ann Clemons. We also thank thefamilies and children who graciously consented to participate inthis research.Correspondence concerning this article should be addressed to

Jim Snyder, Department of Psychology, Box 34, Wichita StateUniversity, Wichita, KS 67260. Electronic mail may be sent [email protected].

Child Development, November/December 2003, Volume 74, Number 6, Pages 1881 – 1898

Page 2: Observed Peer Victimization During Early Elementary School: Continuity, Growth, and Relation to Risk for Child Antisocial and Depressive Behavior

victimization on child adjustment first becomesapparent during this period (Kochenderfer & Ladd,1996b; Ladd & Kochenderfer-Ladd, 2002).

Observational methods complement peer reportand self-report measures by providing relativelysensitive, objective and accurate estimates of theabsolute frequency and situational variation in peerharassment (Cairns & Green, 1979). Observationmay be particularly useful in ascertaining thedynamic shift from high levels of situationalvictimization during initial group formation to amore stable traitlike status in which a smallernumber of children increasingly become the regulartargets of peer aggression. Given that victimizationis sampled on several separate occasions, theamount of variance in victimization at any develop-mental point can be partitioned into situation (state),person (trait), and error (observer disagreement;Jones, Reid, & Patterson, 1975; Stoolmiller, Eddy, &Reid, 2000). Low reliability due to occasion (i.e.,state) can then be treated as useful informationreflecting initial group-assortment processes ratherthan nuisance or error. The persistent experience ofharassment across occasions is directly estimated byvariance due to person and can be construed as atraitlike status as victim. If multioccasion observa-tional data are collected in a panel design, individualgrowth trajectories for trait victimization can beestimated and described. This approach was used inthe current report.

The sensitivity of observation to both situationalvariation and regularity in social behavior aptlyreflects the playground ecology during kindergartenand first grade. School playgrounds are highlyunstructured but nonelective social ecologies inwhich children encounter a large melting pot ofunfamiliar peers, each with unique skills, interests,problems, values, languages, and cultures (Parkeret al., 1995). A wide range of activity and affiliatechoices are available, and these choices shift rapidly(Snyder, Horsch, & Childs, 1997). Initially, childaggression and victimization are likely to be highlysituational. However, over time, hierarchies, alli-ances, biases, and reputations become increasinglyestablished such that some children may increas-ingly experience cross-situational and persistentpeer harassment (Coie & Kupersmidt, 1983).

Stoolmiller et al. (2000) estimated that about 20%of the variance in observed aggressive behavior onthe playground was attributable to stable individualdifferences and 80% to time-varying social environ-mental factors. There was substantial temporalstability (.80) and minimal change in the rates ofplayground aggression during the school year. In the

context of low trait variance, temporal stabilityreflects consistencies both in the individual and inthe individuals’ environment, comparable to thenotions of cumulative and interactional continuity(Caspi, Elder, & Bem, 1987). Comparable, observa-tion-based estimates of trait and situational varia-bility, temporal continuity, and individual trajectoriesfor trait victimization are needed. The means andvariances of these trajectories would describe theevolution of victimization in newly forming peergroups. Variation in intercepts and slopes woulddescribe individual differences in peer victimizationexperiences over time. This variation could be used toascertain sources of risk for persistent and growingvictimization, or the impact of victimization on childadjustment at school and at home. Using thisapproach, three hypotheses concerning the evolutionof peer victimization during early elementary schoolwere tested. The first hypothesis was that therewould be an increase in trait and a decrease insituational victimization with age in the context ofminimal change in group-level rates of victimization.The second hypothesis was that there would be hightemporal stability in trait victimization. The thirdhypothesis was that reliable individual differences ininitial levels and growth in trait victimization wouldbe observed. If these hypotheses are supported, somechildren should show age-related increases andothers should show age-related decreases in cross-occasion victimization over time in the context ofrank-order, within-group temporal stability.

Relation of Peer Victimization to Child-AdjustmentProblems

Previous research on young elementary schoolchildren has established both concurrent and pro-spective associations of persistent victimization withloneliness and peer social relational problems(Alsaker, 1993; Ladd & Kochenderfer-Ladd, 2002;Kochenderfer & Ladd, 1996b), conduct problems(Schwartz, McFadyen-Ketchum, Dodge, Pettit, &Bates, 1998), negative attitudes toward school, andschool avoidance (Kochenderfer & Ladd, 1996a).Assessment of the effects of victimization has reliedprimarily on self-reports and peer nominations ofchild adjustment. Although persistent victimizationappears to have a clear negative impact on childdevelopment, the current empirical picture would been-riched by determining whether persisting victi-mization during early elementary school has gen-eralizable effects on growth in child-adjustmentprob-lems that are also apparent to teachers andparents.

1882 Snyder et al.

Page 3: Observed Peer Victimization During Early Elementary School: Continuity, Growth, and Relation to Risk for Child Antisocial and Depressive Behavior

The causal relationship between victimizationand child adjustment is potentially bidirectional.There are strong concurrent relations between peervictimization and child loneliness, aggression, socialskills deficits, and impulsivity (Ladd & Kochender-fer-Ladd, 2002). Some prospective longitudinalresearch indicates that victimization has a strongereffect on child adjustment than its reciprocal rela-tion (Kochenderfer & Ladd, 1996a), although otherresearch indicates a reciprocal prospective relation-ship in which child aggressiveness and skills deficitslead to peer victimization, which in turn leads toincreasing loneliness, social withdrawal (Boivin,Hymel, & Hodges, 2001), and conduct problems(Schwartz et al., 1998, 1999). This study examinedlongitudinal models in which first peer victimizationand then child adjustment (antisocial and depressivebehavior) are given causal priority over the other.

Gender Differences in Peer Victimization and Its Relationto Child Adjustment

Gender differences in peer victimization and itsrelation to child adjustment during early elementaryschool have received limited empirical attention.Some data indicate few differences. Kochenderferand Ladd (1996b) reported that about 20% of bothboys and girls experience moderate to severe levels ofvictimization in kindergarten. Few gender differencesare apparent when the linkage of victimization toadjustment is assessed using multiple informants(Ladd & Kochenderfer-Ladd, 2002). On the otherhand, boys’ self-reported victimization shows morestability during kindergarten and first grade eventhough the stability of self-report and its concordancewith peer reports increases more quickly for girlsthan for boys during Grades 1 through 3. Earlyvictimization may be more strongly linked to teacher-reported peer problems and peer rejection for boysthan for girls (Ladd & Kochenderfer-Ladd, 2002).

Current findings on gender differences in victim-ization and its relation to adjustment are notcongruent with other research. Boys are consistentlymore aggressive and dominance oriented than girls,and most peer interaction during childhood occursin same-gender groups (Maccoby, 1998). Thus, areasonable hypothesis is that raw rates and levels oftrait victimization would be higher for boys than forgirls. Given their higher levels of impulsivity andirritability (Moffitt, Caspi, Rutter, & Silva, 2001),boys relative to girls might be expected to becomemore frequently involved in coercive exchanges withpeers, creating a stronger linkage of victimizationwith aggressive, overt antisocial behavior. Given the

equivalent rates of depressive behavior in boys andgirls and an absence of gender differences in thelinkage of victimization to loneliness and with-drawal during early elementary school (Ladd &Kochenderfer-Ladd, 2002), gender differences fordepression are not expected. These hypotheses weretested in this study.

Method

Participants

Participants were 132 girls and 134 boys whosemean age at the initial data-collection point (entry tokindergarten) was 5.5 years (range5 5.2 –6.1). Acommunity sample was obtained using a recruitingstrategy targeting all kindergarten children (n5 352,participation rate5 76%) who enrolled in one ele-mentary school in each of 3 consecutive years. Theschool serves a low-socioeconomic, mixed indus-trial-residential neighborhood in a city with apopulation of more than 400,000. The ethnic break-down of the sample was 71% European American,19% African American, 5% Hispanic/Latino, 3%Native American, and 2% Asian American. Forty-three percent of children lived in intact (i.e., two-biological parent) households, 28% in single-parent(predominantly mother) households, 21% in blendedhouseholds, and 7% in other family configurations(grandparental, adoptive). The median per capitafamily income was $8,300; 28% of the children livedin families with incomes below the poverty line, andan additional 23% lived in families with incomesbetween the poverty line and 150% of the povertyline. Forty-six percent of the parents had completedtheir high school education, 20% had less than a highschool education, and 34% had education beyondhigh school. Seventy-five percent of two-parentfamilies were composed of dual wage earners, and9% of the families had no employed adult.

Data collection was continued after participantsmade school or residential moves during kindergar-ten and first grade. At least partial data wereavailable for 97.3%, 96.9%, and 96.6% of the 266children in the spring of kindergarten and fall andspring of first grade, respectively. The most completedata were available for playground observationsat each developmental point. Missing data wereestimated in Amos 4.0 (Arbuckle & Wothke, 1999).Amos uses the full information maximum likelihood(FIML) estimation method. FIML does not deletecases missing from one or more waves of datacollection, nor does it delete cases that are missingone or more variables within a wave of data

Observed Peer Victimization 1883

Page 4: Observed Peer Victimization During Early Elementary School: Continuity, Growth, and Relation to Risk for Child Antisocial and Depressive Behavior

collection. This procedure avoids potential problemssuch as biased parameter estimates that are likely tooccur if pairwise or listwise deletion procedures areused to compensate for missing data (Arbuckle &Wothke, 1999; Wothke, 2000).

Measures

Playground Observation of Peer Victimization

An adaptation of the playground behavior codingsystem developed by Weiss, Dodge, Bates, and Pettit(1992) was used to assess the children’s playgroundbehavior. This interval coding system classifies childbehavior into one of seven mutually exclusive andcollectively exhaustive categories every 10 s accordingto the following rank-order hierarchy: negative inter-action (various forms of verbal and physical aggres-sion), rough play (less intense, incidental negativephysical contact as part of play), positive interaction(other social behavior), parallel play, solitary play,unfocused, and other. When coded, negative interac-tion is further subcategorized into physical (hit, kick,push, pinch, etc.), verbal (face-to-face teasing, namecalling, threats, etc.), and relational forms. Relationalforms of victimization are not considered here as theyare not amenable to observational methods. A morecomplete description of the coding system andmaterial for training can be obtained from the firstauthor. More than one form of aggression may becoded in an interval. The coding system also selec-tively classified the social experiences of the targetchild into one of two (nonexclusive) categories every10 s: being the victim of physical aggression andverbal aggression from peers. These categories areused in the current analyses.

The behavior of participating children was ob-served on the school playground on multipleoccasions at each of four consecutive developmentalpoints, in the fall (September and October) andspring (April and May) of both kindergarten (sixoccasions each; M ages5 5.5 and 6.2 years, respec-tively) and first grade (four occasions each; Mages5 6.5 and 7.2 years, respectively). On eachoccasion, the behavior of the target child was codedfor 5min (30 consecutive 10-s intervals). Fewer than5% of the occasions for a child co-occurred on thesame day and were separated by at least 10min.Victimization by verbal and physical aggression wascalculated as rate per minute (rpm) scores for eachchild on each occasion.

Coders were intensively trained before datecollection. Training entailed instruction and testingon verbal and modeled exemplars of behaviors inthe coding system, and coding of videotapes and

live playground behavior that were sequentiallymore demanding and complex. Training continueduntil a coder reached a minimum kappa agreementof .70 with a master coder. The master coder was anadvanced graduate student in clinical psychologywith coursework in observational assessment andwas trained by the first author who has more than20 years of experience in behavioral observationin naturalistic family and peer settings. Additionaltraining was initiated before each of the develop-mental points at which playground observationsoccurred. Recalibration sessions were held biweeklyto minimize observer drift. A total of 11 coderscollected data, with considerable variation in theinvolvement of specific coders across developmentalassessment points. Only two coders, including themaster coder, engaged in observation at all fourpoints. Children were assigned to coders andobservation dates in a quasirandom fashion, basedon coder and child availability.

Two observers coded children’s behavior on 10%of all occasions to assess observer agreement.Assignment of occasions for reliability assessmentwas quasirandom according to both coders andchildren. Coder agreement was obtained propor-tionally to the number of observation occasions ateach of the developmental points (total occa-sions5 520; total hours5 43.3). Interval-by-intervalindexes of coder agreement for the behavior cate-gories reached acceptable levels at each of thedevelopmental assessment points. Kappa agreementfor verbal victimization and physical victimizationranged from .65 to .75, and kappa agreement for theaggregate category combining verbal and physicalranged from .70 to .80. Intraclass correlations of therpm of physical victimization and verbal victimiza-tion on each observation occasion (the level used inthe subsequent analyses) generated by the variousindependent coder pairs ranged from .79 to .84 (allpso.001). Comparable correlations for the aggregatecategory combining verbal and physical victimiza-tion ranged from .83 to .87. About 15% to 20% of thevariance in rpm victimization was due to coder(measurement) error. There were no statisticallyreliable variations in coder agreement by childgender, coder, or developmental point.

Procedurally, coders identified the target childduring free play periods on the school playground,primarily occurring at noon recess. Behavior wascoded using a paper-and-pencil system in which the5-min observational period was divided into thirty10-s intervals. The end of each interval was signaledby a visual-audio timing device, at which time thecoder recorded the appropriate categories. Physical

1884 Snyder et al.

Page 5: Observed Peer Victimization During Early Elementary School: Continuity, Growth, and Relation to Risk for Child Antisocial and Depressive Behavior

recording of categories was rapid so that intervalsrepresent roughly contiguous periods. Observerslocated themselves at a distance (typically within 6ft) and perspective (frontal view) to see and hear thesocial exchanges between children. When a childwas not visible or relatively proximal to the observerand relevant social behavior was unavailable, codingwas interrupted until proximal contact could bereestablished.

Children’s reactivity was minimized in threeways. First, 5 to 10 days of practice coding sessionswere used at the initiation of data collection at eachdevelopmental point so that children could (re)ac-climate to the presence of the observers. Second,observers were trained to be literally nonresponsiveto child overtures and queries, and children’sattempts to engage the observers were quicklyextinguished. It was our experience as well as others(Pellegrini, 2001) that young children are remarkablynonreactive to observation given a few hours ofexposure to coders who literally do not respond tochild behavior. There were very occasional excep-tions to this nonresponsiveness stance, involvingcircumstances in which children were in danger ofserious physical harm. When such occasions oc-curred, most of these interventions did not involvethe child or peers who were immediate targets ofobservation. Third, when a child and the peers withwhom he or she was interacting evidenced sustained(more than 1min) behavioral reactivity (whispering,running away, etc.), observation was terminated andreestablished on a new occasion.

Observations were conducted on a playgroundadjacent to the school that was roughly the size oftwo football fields and enclosed by a fence. Itcontained two sets of typical playground equipment,a small cement area, sandy areas in which thechildren dug and played, and a large grassy area.Approximately one third of the children in theschool (85 children, half each from kindergarten andfirst grade) were on the playground at one time andwere typically supervised by two untrained adultparaprofessionals. Adults made few attempts toorganize activities. Interaction was relatively free ofadult constraints and was largely determined by thenatural interests and affiliations of the children andthe contingencies that they provided one another.

Child Overt Antisocial and Depressive Behavior

Teacher ratings. Teachers’ ratings of child overtantisocial and depressive behavior were obtained inthe fall and spring of kindergarten and in the falland spring of the first grade, using an adaptation of

the Teacher Report Form (TRF; Achenbach, 1991).Teachers were asked to rate a child’s behavior overthe previous 2-month period. Different teacherstypically completed these scales in kindergartenand first grade. A score for overt antisocial (aggres-sive) behavior was derived at each developmentalpoint, composed of 26 items (e.g., argues, cruel/bullies, disobeys, teases, tantrums). Internal relia-bility of the scale was alpha 4.93 at each assessmentpoint. A score for child depressive behavior wasderived at each developmental point, composed of12 items (e.g., lonely, sad, withdrawn). Internalreliability of the scale was alpha 4.69 at eachassessment point.

Parent ratings. Parents’ perceptions of child overtantisocial and depressive behavior were obtained inthe fall and spring of kindergarten and in the fall andspring of the first grade using an adaptation of theChild Behavior Checklist (CBCL; Achenbach, 1991).Parents were asked to rate their child’s behavior overthe previous 2-month interval. A scale for overtantisocial behavior was derived at each develop-mental point, composed of 23 items (e.g., argues,cruel/bullies, disobeys, fights, teases, and tantrums).Internal reliability was alpha4.90 at each assessmentpoint. A score for child depressive behavior wasderived at each developmental point, composed of 11items (e.g., sad, guilty, withdrawn). Internal reliabilitywas alpha 4.74 at each assessment point.

Child self-reported depression. The Computer-As-sisted Child Interview (CACI; Bank, 2000) was usedto obtain child self-reports of depressive affect andbehavior at the end of the first grade. The CACIentails the visual presentation of a series of picturesand accompanying verbal descriptors of variousbehaviors, affects, and social events relevant toyoung children’s experience. Children rate thedegree to which they have experienced or displayedeach, using a visual device (number of jellybeans in ajar), from not at all (empty) to some (half-full) to a lot(full). The depression scale of the CACI is composedof nine items (e.g., sad, lonely, worried). This scale isinternally consistent (alpha5 .73) and stable over a1-week interval (test – retest r5 .80).

Results

Observed Rates of Peer Victimization on the SchoolPlayground

Substantial rates of victimization were observed.On average, children were targets of peer physical orverbal harassment about once every 3 to 6min (Table1). Physical and verbal harassment occurred at

Observed Peer Victimization 1885

Page 6: Observed Peer Victimization During Early Elementary School: Continuity, Growth, and Relation to Risk for Child Antisocial and Depressive Behavior

comparable rates, and these rates changed little fromfall to spring of each school year, or from kinder-garten to first grade. A repeated measures multi-variate analysis of variance (MANOVA) indicated asignificant main effect for gender. F(15, 251)5 6.06,po.01, but not for time or the Gender� Timeinteraction. Follow-up analyses of variance (ANO-VAs) indicated that boys were more often subjectedto verbal and physical harassment but that theseeffects were marginal (ps5 .09 and .07, res-pectively).

Estimates of Trait and State Victimization

The distributions for rpm victimization at eachobservation occasion were significantly skewed; 30%to 35% of the children were not victimized by peersduring any given 5-min observation period consti-tuting an occasion. However, only 5% to 10% werenever victimized during the multiple observationoccasions at each developmental point. Victimiza-tion rates for each occasion were log transformed.First-order latent constructs were estimated sepa-rately for physical and verbal victimization at eachdevelopmental point, using as indicators the trans-formed rpm scores at each of the six (at both fall andspring kindergarten) or four (at both fall and springfirst grade) occasions during which observationswere conducted. These first-order constructs wereestimated separately for boys and girls using a strictparallel reliability model in which all factor loadingswere set to 1 and all variances across occasions at aspecific developmental point were set to be equal.This model treats each observation session as anequally good estimate of a child’s victimizationexperiences, and these constructs reflect the cross-occasion consistency of victimization at each devel-opmental point (Stoolmiller et al., 2000). Each ofthese models fit the data (p4.05, root mean square

error of approximation (RMSEA) o.05, comparativefit index (CFI) 4.95).

Correlations between the latent factors for cross-occasion verbal and physical victimization at con-current developmental assessment points were veryhigh (4.75). Models in which the correlationsbetween verbal and physical victimization factorswere constrained to 1 were found to fit the data aswell as their comparable correlated two-factormodels. Subsequent analyses were based on singleconstructs that aggregate verbal and physical formsof cross-occasion victimization at each developmen-tal point. The first-order latent constructs reflectindividual differences in children’s victimizationthat are stable across observation occasions at eachdevelopmental point and are hereafter referred to aschildren’s trait victimization status.

After corrections for observer unreliability, thepercentage of true variance (80% or average observerreliability) due to cross-occasion or trait victimiza-tion was: fall kindergarten5 18% for boys and 13%for girls; spring kindergarten5 17% for boys and11% for girls; fall first grade5 26% for boys and 20%for girls; and spring first grade5 32% for boys and28% for girls. The proportion of trait varianceroughly doubled over the 18-month developmentalperiod. The remaining true variance was due to day-to-day situational factors on the playground. Con-tinuity in trait victimization was estimated by Lag 1autoregressive paths separately for boys and girls.All standardized path coefficients were greater than.60 (po.001), indicating substantial rank-order con-tinuity in trait victimization for boys and girls.

Multigroup (by child gender) linear growthmodels (with intercept set at fall kindergarten) werefit to the first-order latent constructs for traitvictimization across the four developmental points.Model parameters and significance tests are shownin the top row of Table 2. The model fit the data

Table 1

Mean Observed Rates per Minute of Peer Victimization

Kindergarten First grade

Fall Spring Fall Spring

Boys Girls Boys Girls Boys Girls Boys Girls

Physical .36 .19 .35 .20 .30 .17 .40 .20

(.28) (.19) (.27) (.15) (.24) (.22) (.38) (.24)

Verbal .25 .16 .23 .19 .24 .15 .28 .19

(.23) (.14) (.17) (.17) (.22) (.18) (.24) (.17)

Note. Standard deviations are in parentheses.

1886 Snyder et al.

Page 7: Observed Peer Victimization During Early Elementary School: Continuity, Growth, and Relation to Risk for Child Antisocial and Depressive Behavior

adequately. Between-child variations in initial statusat fall kindergarten and in trajectory slope from fallkindergarten to spring of first grade for traitvictimization were highly significant, with theexception of slope for girls (marginal at po.10).The means of the slopes for boys and girls werepositive but not significant, indicating no group-level average growth in trait victimization. In thecontext of no average growth, significant variabilityin slopes indicates that trait victimization increasesfor some children over time but decreases for others.Boys exhibited higher rates and more variation intrait victimization at school entry than did girls. Nosignificant gender differences were found in slopemeans or variability. When Lag 1 autoregressivepaths were added to the linear growth model, pathcoefficients between adjacent constructs for traitvictimization were not significant. Rank-order tem-poral continuity in victimization was incorporatedinto the growth model.

Child Overt Antisocial and Depressive Behavior

Children in the sample displayed substantial ratesof overt antisocial behavior (comparable to the

aggression scales on the CBCL and TRF). Meanparent-reported antisocial behavior at fall kinder-garten was 12.1 (SD5 7.6) for boys and 10.5(SD5 6.7) for girls; both means were more than .5SD above normative means. Sixteen percent of theboys and 9% of the girls had T scores greater than 70(98th percentile). The mean of teacher-reportedantisocial behavior at fall kindergarten was 6.6(SD5 8.4) for boys and 4.7 (SD5 6.5) for girls; bothmeans were more than .25 SD above normativemeans. Eleven percent of the boys and 5% of thegirls had T scores for aggression greater than 70 (98thpercentile).

Parent- and teacher-reported depression scalesderived for this research have no direct counterpartson the CBCL and TRF. Instead, items were selectedto reflect overt depressive symptoms as specified inthe Diagnostic and Statistical Manual of MentalDisorders–4th Revision (DSM–IV) and the ChildDepression Inventory (CDI; Kovacs, 1980). Boys’ andgirls’ mean scores on the 11-item parent scale at fallkindergarten were 3.2 (SD5 3.1) and 2.8 (SD5 2.7),respectively. Boys’ and girls’ mean scores on the 11-item teacher scale at fall kindergarten were 1.9(SD5 2.4) and 1.5 (SD5 2.0), respectively. The

Table 2

Fit Indexes and Parameters for Linear Growth Models of Trait Victimization, and Child Overt Antisocial and Depressive Behavior

Boys Girls

Intercept Slope Intercept Slope

Growth model M SD M SD M SD M SD

Trait victimization (rpm); model fit: w2/df5 1.21, CFI5 .975, RMSEA5 .028

� 0.560a 0.232a � 0.004 0.063 � 0.872a 0.187a 0.003 0.056

(� 18.19) (9.04) (� 0.26) (2.22) (� 30.91) (7.60) (0.20) (1.70)

Overt antisocial behavior

Teacher report; model fit: w2/df5 1.63, CFI5 .997, RMSEA5 .041

0.283a 0.361 0.027 0.118 0.153a 0.238 0.045 0.122

(9.17) (16.52) (2.12) (10.60) (7.20) (15.84) (3.76) (13.44)

Parent report; model fit: w2/df5 1.73, CFI5 .997, RMSEA5 .048

0.673a 0.197 � 0.006a 0.039 0.631a 0.205 � 0.023a 0.055

(33.60) (13.44) (� 0.79) (2.44) (31.90) (7.60) (� 3.14) (5.64)

Depressive behavior

Teacher report; model fit: w2/df5 1.71, CFI5 .987, RMSEA5 .043

0.314a 0.161 0.024 0.055 0.243a 0.158 0.049 0.063

(16.75) (10.32) (2.83) (5.80) (12.21) (9.42) (5.00) (6.54)

Parent report; model fit: w2/df5 0.95, CFI5 .998, RMSEA5 .001

0.546 0.190 � 0.004 0.054 0.589 0.219 � 0.015 0.044

(24.89) (11.02) (� 0.41) (4.90) (25.02) (12.52) (� 1.80) (3.70)

Note. All models are represented as linear growth. Measures for overt antisocial and depressive behavior were all square root transformedto reduce skewness. Victimization was measured as a construct at each time point using observations across multiple observationoccasions as indicators in a strict parallel reliability model. Critical ratios are shown in parentheses. rpm5 rate per minute;CFI5 comparative fit index; RMSEA5 root mean square error of approximation.aGender differences where boys exceed girls, po.05.

Observed Peer Victimization 1887

Page 8: Observed Peer Victimization During Early Elementary School: Continuity, Growth, and Relation to Risk for Child Antisocial and Depressive Behavior

possible range for both scales was 0 to 22. In thefall kindergarten, 5% of the boys and 2% of thegirls received an average item rating greaterthan 1 (somewhat or sometimes true) on the parentscale, and 3% of the boys and 2% of the girls re-ceived an average rating greater than 1 on theteacher scale.

All teacher and parent scales were log trans-formed. Linear growth models for teacher- andparent-reported child antisocial and depressivebehavior fit the data adequately (see Table 2). Ineach case, significant variation was observed ininitial levels (fall kindergarten) and slopes for boysand girls. This variation indicates substantial in-dividual differences in child antisocial and depres-sive behavior, requisite to examining their relation-ships to growth in trait victimization. When genderdifferences were present, boys exhibited highermean levels, more variance, and larger slopes inantisocial and depressive behavior.

Relationship of Trait Victimization and ChildAdjustment: Analytic Plan

Three alternate causal models were fit to the paneldata for victimization and for child antisocial anddepressive behavior: an autoregressive bivariateMarkov simplex model, a bivariate growth model,and a combination of the two (Curran & Bollen,2001; Curran, Stice, & Chassin, 1997). The Markovmodel assessed the time-specific relations betweentwo constructs. It contained the bivariate correlationbetween trait victimization and antisocial (or de-pressive) behavior in fall kindergarten, autoregres-sive paths from Time t to Time t11 for eachconstruct, and Time t to Time t11 cross-laggedbidirectional paths (see Figure 1a). Autoregressiveand cross-lagged paths representing the sameinfluence at different intervals were constrained tobe equal, assuming that temporally focal groupprocesses operate in the same way during each ofthe 6-month intervals constituting the 18-monthdevelopmental span under consideration. Thecross-lagged paths represent longitudinal predictionof one construct from the other beyond the auto-regressive prediction of that construct from itself.Autoregressive and cross-lagged effects representgroup-level, average processes operating over ashort, 6-month time frame. This model maps closelyonto past research in which child adjustment waspredicted by earlier victimization after controllingfor earlier adjustment. The cross-lagged causalrelation may be reciprocal.

In the bivariate growth model, continuous in-dividual trajectories for victimization and for anti-social and depressive behavior are represented byinitial levels and change over time. The relation ofbetween-individual variation in one trajectory withvariation in the other trajectory is estimated bydirectional paths from the intercept and slope of oneconstruct (e.g., trait victimization) to the slope of thesecond construct (e.g., child antisocial behavior)while controlling for all other (correlational) rela-tionships (see Figure 1b). The direction of the pathscan be reversed to test the alternate, reciprocalmodel, but causal priority is given to one trajectoryover the other in each case.

Finally, parameters for bivariate cross-lagged andbivariate growth models were estimated simulta-neously in a combined model (Figure 1c). Thismodel assesses sources of variation due to bothshort-term group processes and long-term individ-ual trajectories, each in the context of the other.Parameters from this model can be compared witheach of the two simpler models. If sources ofinfluence in simpler models are diminished in thecombined model, redundancy might be inferred. Ifsources of influence not apparent in the simplermodels become reliable in the combined model, asynergistic effect might be inferred. If parameters inthe combined model remain unchanged relative thesimpler models, additive effects might be inferred.

Bivariate Markov Simplex Models

Bivariate simplex models including mean struc-ture and separate equality constraints for eachautoregressive and cross-lagged path (by childgender) fit the data (w2/dfo1.30, RMSEAo.032,CFI4.965) for models that linked trait victimizationwith parent- and teacher-reported child antisocialand depressive behavior. Standardized coefficientsare shown in Table 3. There were significantcorrelations between trait victimization and teacherand parent reports of child antisocial behavior in fallkindergarten for boys and girls (first column). Theautocorrelational paths for each adjustment indexand for trait victimization (second and thirdcolumns, respectively) were all large and statisticallysignificant for boys and girls. Trait victimization waspredictive of subsequent parent-reported antisocialbehavior for boys and of parent-reported depressivebehavior for girls (fourth column). There weresignificant positive cross-lagged relationships fromteacher- and parent-reported antisocial behavior,and from teacher-reported depressive behavior totrait victimization for girls (last column).

1888 Snyder et al.

Page 9: Observed Peer Victimization During Early Elementary School: Continuity, Growth, and Relation to Risk for Child Antisocial and Depressive Behavior

Bivariate Linear Growth Models

Measurement models (Anderson & Gerbing,1988) were tested for each of the multigroup (by

gender) models relating trait victimization and eachchild adjustment index. All measurement models fitthe data (w2/dfo1.32, RMSEAo.035, CFI4.954). Aset of causal models (see Figure 1b) was tested that

Fig. 1a. Schematics of bivariate Markov simplex model. FK5 fall kindergarten; SK5 spring kindergarten; FF5 fall first grade; SF5 springfirst grade.

Fig. 1b. Schematics of bidirectional bivariate growth models predicting child adjustment from trait victimization. Shown only with causalpriority given to victimization over child adjustment; a complementary model giving causal priority to child adjustment overvictimization was also tested. FK5 fall kindergarten; SK5 spring kindergarten; FF5 fall first grade; SF5 spring first grade.

Fig. 1c. Schematics of directional bivariate growth model with embedded cross-lagged simplex paths between adjustment andvictimization. For clarity, shown without paths or correlations between intercepts and slopes, which would include those from Figure 1b.FK5 fall kindergarten; SK5 spring kindergarten; FF5 fall first grade; SF5 spring first grade.

Observed Peer Victimization 1889

Page 10: Observed Peer Victimization During Early Elementary School: Continuity, Growth, and Relation to Risk for Child Antisocial and Depressive Behavior

gave victimization causal priority over child adjust-ment in the presence of all the other relationshipsamong intercepts and slopes (i.e., with directionalpaths from intercept and slope of victimization toslope adjustment, and from intercept to slopeadjustment). A set of reciprocal causal models wastested that gave causal priority to child adjustmentover trait victimization in the presence of all theother relationships among intercepts and slopes (i.e.,with directional paths from intercept and slopeadjustment to slope victimization, and from inter-cept to slope victimization). These directional mod-els are equivalent to the measurement model andhave the same fit parameters, but they provide the

basis for stronger causal inferences about howindividual trajectories for victimization and for childadjustment exert influence on each other and aboutthe direction in which that influence is exerted.

The results of the directional bivariate growthmodels are shown in Table 4, with path parametersfor the effects of victimization on adjustment on theleft, and path parameters for the effects of adjust-ment on victimization on the right. For boys (top ofTable 4), growth in victimization was predictive ofgrowth in teacher-reported antisocial behavior (.35).There was a reciprocal predictive relation betweengrowth in victimization and growth in teacher-reported depressive behavior (.46 and .41). For girls

Table 4

Standardized Path Coefficients for Bivariate Growth Models: The Effects of Victimization on Adjustment and of Adjustment on Victimization

Effects of victimization on adjustment Effects of adjustment on victimization

Adjustment index

Intercept

victim to slope

adjustment

Slope victim

to slope

adjustment

Intercept

to slope

adjustment

Intercept

adjustment to

slope victim

Slope adjustment

to slope victim

Intercept

to slope victim

Boys

AntisocialFteacher .15 .35w � .47� � .08 .28 � .56��a

AntisocialFparent � .09a .16 .07 � .06 .10 � .58��a

DepressionFteacher .03 .46wa � .16 � .10a .41wa � .50��a

DepressionFparent � .30 .27 .05 � .18 .15 � .51��a

Girls

AntisocialFteacher .25 � .03 � .49� .07 � .04 .06a

AntisocialFparent .55wa � .30 .04 � .23 � .32 .08a

DepressionFteacher � .01 � .29a .00 .50wa � .28a .06a

DepressionFparent .10 .38w .03 � .20 .26 .03a

aGender difference at po.05. wpo.10. �po.05. ��po.01.

Table 3

Standardized Coefficients for the Bivariate Simplex Models of the Relation of Trait Victimization to Child Adjustment

Adjustment index

Correlation with

victimization at fall

kindergarten

Time t to t11

autocorrelation:

child adjustment

Time t to t11

autocorrelation:

victimization

Lagged path:

victimization to

adjustment

Lagged path:

adjustment to

victimization

Boys

AntisocialFteacher .28� .70�� .63�� � .06 .06

AntisocialFparent .29� .68�� .66�� .10� � .01

DepressionFteacher .09 .49�� .65�� .05 � .01

DepressionFparent .24� .57�� .63�� � .01 .01

Girls

AntisocialFteacher .24� .69�� .77�� .07 .16��

AntisocialFparent .27� .75�� .81�� � .05 .11�

DepressionFteacher .01 .50�� .78�� .07 .19��

DepressionFparent .08 .63�� .75�� .10� � .01

�po.05. ��po.01.

1890 Snyder et al.

Page 11: Observed Peer Victimization During Early Elementary School: Continuity, Growth, and Relation to Risk for Child Antisocial and Depressive Behavior

(bottom of Table 4), victimization in fall kindergartenwas predictive of growth in parent-reported anti-social behavior (.55), growth in victimization waspredictive of growth in parent-reported depressivebehavior (.38), and teacher-reported depression infall kindergarten was predictive of victimization(.50). Gender differences in path parameters accom-panied most of the significant effects for boys andgirls.

Combined Bivariate Growth and Cross-Lagged MarkovModels

Teacher-reported overt antisocial behavior. The slopeof victimization reliably predicted the slope ofteacher-reported antisocial behavior for boys

(b5 .53, top row in Table 5) but not for girls (b5– .23, fifth row in Table 5), a significant genderdifference. In the reciprocal directional model, thepath from slope of antisocial behavior to traitvictimization was positive and significant for boys(b5 .51, top row in Table 6) but not for girls ( – .29,fifth row in Table 6), a significant gender difference.The cross-lagged simplex paths from child antisocialbehavior to trait victimization in both directionalmodels (see last column in Tables 5 and 6) werenegative and significant for boys and positive andsignificant for girls, a significant gender difference.A schematic diagram of the significant effects isshown in Figure 2.

For boys only, the inclusion of cross-lagged paths(relative to models without their inclusion, see Table

Table 5

Standardized Path Coefficients for Bivariate Directional Models With Markov Simplex Cross-Lag Relationships: The Effect of Victimization on

Adjustment

Adjustment index

Intercept victim

to slope adjustment

Slope victim to

slope adjustment

Intercept adjustment

to slope adjustment

Cross-lagged

victim to adjustment

Cross-lagged

adjustment to victim

Boys

AntisocialFteacher .23 .53��a � .50�� � .12 � .29��a

AntisocialFparent � .22a .12 .00 .11� � .22��a

DepressionFteacher .03 .46�a � .16 .01 .04

DepressionFparent � .30 .07a � .07 .06�a � .05

Girls

AntisocialFteacher .26 � .23a � .40�� � .06 .14�a

AntisocialFparent .35�a .07 � .11 .00 .11�a

DepressionFteacher � .01 � .29a .00 � .01 .08

DepressionFparent .07 .42�a .01 � .04 .01

aGender differences reliable at po.05. �po.05. ��po.01.

Table 6

Standardized Path Coefficients for Bivariate Directional Models With Markov Simplex Cross-Lag Relationships: The Effect of Adjustment on

Victimization

Adjustment index

Intercept adjustment

to slope victim

Slope adjustment

to slope victim

Intercept victim

to slope victim

Cross-lagged

victim to adjustment

Cross-lagged

adjustment to victim

Boys

AntisocialFteacher .21 .51��a � .53�a � .02a � .27��a

AntisocialFparent .12 .07 � .64��a .13��a � .16��a

DepressionFteacher � .05 .43�a � .55��a .01 � .06

DepressionFparent � .07 .04 � .59��a .05�a � .04

Girls

AntisocialFteacher � .18 � .29a .03a .01 .16�a

AntisocialFparent � .29 � .09 � .05a .06a .14�a

DepressionFteacher .53�a � .33a .00a .07 � .01

DepressionFparent � .13 .32 � .04a � .04 .01

aGender differences reliable at po.05. �po.05. ��po.01.

Observed Peer Victimization 1891

Page 12: Observed Peer Victimization During Early Elementary School: Continuity, Growth, and Relation to Risk for Child Antisocial and Depressive Behavior

4) altered a number of parameters in growth modelsfor teacher-reported antisocial behavior, suggestinga synergistic effect. The slope-to-slope path coeffi-cients increased in size from marginal or nonsigni-ficance to significance (from .35 to .53 for the pathfrom slope victimization to slope antisocial behavior,and from .28 to .51 for the path from slope antisocialbehavior to slope trait victimization), and reducedresidual variance in the slope for antisocial behaviorfrom .72 to .49, and in the slope for trait victimizationfrom .50 to .45. Cross-lagged paths from antisocialbehavior to victimization became significant andnegative. The suppressive impact of antisocialbehavior on victimization at 6-month-lag intervalsaligned the individual growth trajectories for victi-mization and overt antisocial behavior, and theinclusion of relations between the growth trajectoriesuncovered group processes that were otherwise notdiscerned. The significant positive cross-laggedpaths from teacher-reported antisocial behavior tovictimization for girls identified in the bivariateMarkov model (see Table 3) were maintained, and

the slope-to-slope coefficients changed from nearly 0to negative (though nonsignificant) in the context ofthe bivariate growth model.

Parent-reported overt antisocial behavior. Trait victi-mization in fall kindergarten was a significant,positive predictor of the slope in parent-reportedantisocial behavior for girls (b5 .35, sixth row inTable 5) but negative and nonsignificant for boys(b5 – .22, second row in Table 5), a significantgender difference. Other directional paths betweenintercepts and slopes were not significant. Bothcross-lagged autoregressive paths were significantfor boys (second row in Tables 5 and 6), with anegative relation from victimization to antisocialbehavior and a positive relation from antisocialbehavior to victimization. For girls, the cross-laggedpath from antisocial behavior to victimization waspositive (sixth row in Tables 5 and 6) and opposite ofthat for boys, a significant gender difference. Adiagram of the significant effects is shown in Figure 2.

Teacher-reported depressive behavior. The slope oftrait victimization reliably predicted the slope of

Fig. 2. Schematics of significant paths in the bivariate Markov simplex plus bivariate growth models. Solid lines designate significant pathsfor boys, and dotted lines designate significant paths for girls. Cross-lagged paths for relations between only two time-specific measuresare shown for exposition clarity. The symbol 1 indicates a statistically significant positive relationship, – indicates a statisticallysignificant negative relationship, and � indicates a statistically significant gender difference for congruent paths.

1892 Snyder et al.

Page 13: Observed Peer Victimization During Early Elementary School: Continuity, Growth, and Relation to Risk for Child Antisocial and Depressive Behavior

teacher-reported depressive behavior for boys(b5 .46, third row in Table 5) but not for girls(b5 – .29, seventh row in Table 5), a significantgender difference. The slope of depressive behaviorreliably predicted the slope for victimization forboys (b5 .43, third row in Table 6) but not for girls(b5 – .33, seventh row in Table 6), a significantgender difference. The intercept of depressionreliably predicted slope in victimization for girls(b5 .53, seventh row in Table 6) but not for boys(b5 – .05, second row in Table 6), a significantgender difference. No cross-lagged simplex pathswere significant. A schematic diagram of thesignificant effects is shown in Figure 2.

Parent-reported depressive behavior. The slope ofvictimization reliably predicted slope in depressivebehavior for girls (b5 .42, eighth row in Table 5) butnot for boys (b5 .07, fourth row in Table 5), asignificant gender difference. The slope of depres-sive behavior did not predict the slope for victimiza-tion for girls or boys. A reliable positive cross-laggedpath from victimization to depressive behavior wasobserved for boys (b5 .06, fourth row in Table 5) butnot for girls (b5 – .04, eighth row in Table 5), asignificant gender difference. The cross-lagged pathfrom depression to victimization for girls found inthe Markov model (Table 3) was nonsignificantin the context of the bivariate growth model.A diagram of the significant effects is shown inFigure 2.

Prospective Relation of Trait Victimization to Child Self-Reported Depressive Experiences

Given parents’ and teachers’ potential difficultyin accessing children’s depressive experiences, afinal model was tested to determine whether growthtrajectories for observed victimization during kin-dergarten and first grade predicted child self-reported depressive experiences at the end of firstgrade. Boys reported more depressive experiencesthan girls, t(136)5 3.75, po.01. Given that a compar-able child self-report measure of depression was notavailable at earlier assessment points, observerratings of child sadness made every 15min during4 hr of parent – child interaction obtained in thekindergarten year were inserted into the model asa proxy measure to control for earlier child depres-sive experiences. The model examined directionalpaths from intercept and slope of victimization atschool, and from observed sadness during parent–child interaction to child self-reported depressiveexperiences in the spring of first grade. All otherrelations among were represented as correlations.

The multigroup (by gender) model fit the data (w2/df5 1.24, RMSEA5 .030, CFI5 .975). The slope ofvictimization was reliably related to depressiveexperiences for girls (b5 .55, po.05) but not forboys (b5 .13), a reliable gender difference. Theintercept for victimization was not reliably relatedto depressive experiences in the spring of first gradeeither for girls (b5 – .17) or for boys (b5 .09). Thecorrelation between sadness during parent – childinteraction in kindergarten and the slope forvictimization at school was significant for girls(r5 .41, po.05) but not for boys (r5 – .20).

Discussion

Peer Victimization From an Observational Perspective

The high rates of harassment on the schoolplayground during kindergarten and first gradeare consistent with previous data derived fromobservations (Stoolmiller et al., 2000) and from peerreports and child self-reports (Kochenderfer & Ladd,1996b). The high rates of peer harassment duringearly elementary school may reflect young children’sevolving abilities to regulate their own emotions andbehavior and to respond in constructive and skilledways to peer conflict (Parker et al., 1995). They alsoreflect the relative absence of effective adult con-tingencies to modulate social interaction on theschool playground (Walker, Colvin, & Ramsey,1995).

Both more (e.g., physical hitting, coercive threats)and less intense (e.g., poking with a finger, namecalling) verbal and physical forms of harassmentwere used to define rates of victimization for threereasons. First, the rates of verbal and physicalvictimization by peers are highly correlated. Second,more frequent but less intense aversive experiencesmay have deleterious effects similar to those that areless frequent and more intense (Patterson, 1982).Third, physical aggression and coercive threats aretypically embedded in conflicts that escalate fromlower intensity aggressive acts (Snyder, Edwards,McGraw, Kilgore, & Holton, 1994). More research isneeded to ascertain whether low- and high-intensityverbal and physical forms of harassment havesimilar determinants and detrimental effects.

The application of a strict parallel reliabilitymodel to cross-occasion observation suggests thatmuch of peer victimization during kindergarten andfirst grade is situationalFit depends on what a childhappens to be doing, where the child happens to beplaying, and whom the child happens to be with ona particular day or at a particular time. Part of peer

Observed Peer Victimization 1893

Page 14: Observed Peer Victimization During Early Elementary School: Continuity, Growth, and Relation to Risk for Child Antisocial and Depressive Behavior

experience entails learning how effectively to resolveconflict and how selectively to avoid individuals andsettings that set the occasion for aggression andharassment (Shantz & Hobart, 1994). However,victimization is not totally situational; reliableaverage group levels and individual differences incross-occasion victimization are discernable early inkindergarten. Some children rapidly acquired atraitlike status as a victim early in kindergartenwhereas others did not. Though mean levels of traitvictimization did not change during kindergartenand first grade, substantial continuity and reliableindividual differences in trajectories for trait victi-mization were observed. There is a continuingassortment and differentiation in children’s statusas victims during kindergarten and first grade (Ladd& Kochenderfer-Ladd, 2002). Some children experi-enced harassment with increasing regularity. Otherchildren appeared to respond effectively to aggres-sion by peers such that harassment experiencesbecame increasingly intermittent and situational.Social interactional analyses are needed to under-stand how some children learn to effectively copewith or avoid repeated victimization whereas othersdo not.

Boys experienced higher rates and exhibitedlarger individual differences in trait victimizationthan did girls at kindergarten entry. These differ-ences were maintained over the subsequent 18months and may reflect gender segregation in peerplay in which interaction in boys’ relative to girls’groups is characterized by more dominance, compe-tition, and challenge (Maccoby, 1998). However,significant and substantial mean levels and individ-ual differences in trait victimization were also obser-ved for girls and may have been underestimated inthe absence of measures of indirect or relationaggression (Crick & Grotpeter, 1996). Peer victimiza-tion is a substantial reality early in elementaryschool. Interventions to reduce situational peeraggression and victimization at the transition toschool may effectively alter trajectories for aggres-sion and victimization before they evolve into morepersistent and chronic forms.

Relationship of Victimization and Multisetting ChildAdjustment: Causal Models

The longitudinal panel design permitted theexamination of a series of complementary modelsin which causal priority was first assigned tovictimization and then to child antisocial or depres-sive behavior to assess their relative prospectiveinfluence. The models examined gender differences

and the relationship of observed victimization andadjustment as measured from multiple perspectives(parent report, teacher report, and self-report),reflecting both proximal (school) and distal (home)settings. There was no shared-method variancebetween measures of victimization and adjustment.The data are discussed separately for antisocial anddepressive behavior.

Trait Victimization and Antisocial Behavior

Boys. For boys, trait victimization and antisocialbehavior at school were powerfully related at thelevel of both long-term continuous individualtrajectories and short-term group processes. In fact,short-term group and long-term individual effectsappeared to operate in a synergistic fashion topromote and sustain one another. In the shortterm, frequent displays of antisocial behavior atschool suppressed subsequent victimization bypeers. The functional value of aggressive behaviorin reducing peer victimization implies a negativereinforcement effect (Patterson, Littman, & Bricker,1967). However, this short-term benefit came at along-term cost in that individual trajectories forrecurring victimization and teacher-reported anti-social behavior were increasingly aligned andreciprocally causal. This suggests that a bilateralreputational-defensive process (Coie & Kupersmidt,1983) operates in the school setting at the level ofindividual trajectories as well as at a more immedi-ate social interactional level (Snyder, 2002) in whichfrequent aggressive behavior and persisting victimi-zation become linked in a repeating coercion-counter-coercion feedback loop.

Boys who were highly antisocial at home wereprospectively less likely to experience recurringvictimization by peers, replicating the short-termnegative reinforcement function of antisocial behav-ior at school. Persistent victimization at school alsoled to short-term increases in antisocial behavior athome. Boys who are aggressive at both home andschool have more severe adjustment problems thanthose displaying problems in only one setting(Loeber & Dishion, 1984) and consequently mayhave even greater difficulty in establishing effectivepeer relations in early elementary school. Thereciprocal short-term effects between peer victimiza-tion and antisocial behavior at home suggests ageneralization process; the lessons boys learn inresponding to peer victimization affects their behav-ior at home, and their level of training in antisocialbehavior at home affects how they are received bypeers at school.

1894 Snyder et al.

Page 15: Observed Peer Victimization During Early Elementary School: Continuity, Growth, and Relation to Risk for Child Antisocial and Depressive Behavior

Girls. For girls, trait victimization early in kin-dergarten had a generalized long-term effect ongrowth in antisocial behavior at home, indicating theimportance of establishing effective peer relation-ships early after the transition to school. At a groupprocesses level, girls who displayed antisocialbehavior at both home and school were prospec-tively at increasing risk for persistent peer victimiza-tion in the short run, an effect opposite of thatobserved for boys. Tough and aggressive behaviormay be more salient and gender discordant amonggirls, who value cooperation, harmony, and closedyadic relationships (Maccoby, 1998). Highly ag-gressive girls may have difficulty in successfullyaccommodating the prevailing peer social ethosfrom very early in formation of the peer groupwithout experiencing blatant rebuffs in terms ofreciprocated verbal and physical aggression. This isinconsistent with past research in which girls’aggression appeared to be largely ignored by peers(Fagot, Hagan, Leinbach, & Kronsberg, 1985).

Trait Victimization and Depressive Behavior

Boys. From the teachers’ perspective, growth intrait victimization and depressive behavior werereciprocally related. Boys who exhibited sadness,withdrawal, and reticence were increasingly targetsof verbal and physical aggression by peers, perhapsbecause of their passivity and failure to fit into thecompetitive male ecology. Ongoing victimizationexacerbated growth in sadness, withdrawal, andworry, which in turn may have been used by peers tomark children for continued victimization. Thisdevelopmental feedback loop is similar to thatdescribed by Boivin et al. (2001). The impact of traitvictimization on distress was sufficient to generalizeto the home setting, at least in the short run.

Girls. The relationship of persistent victimizationand depressive behavior was also apparent for girls.Sadness, withdrawal, and worry at kindergartenentry powerfully facilitated growth in trait victimi-zation over the next 18 months, similar to previousresearch (Kochenderfer & Ladd, 1996a). As withantisocial behavior, girls who are unable quickly toestablish supportive peer relationships appear to beat increased risk for recurring and persisting ostra-cism and rejection. The experience of increasingvictimization was sufficiently powerful to be appar-ent to the girls themselves as well as to their parents.

A Broader Perspective

Consistent with previous research (Alsaker, 1993;Kochenderfer & Ladd, 1996a, 1996b; Ladd &

Kochenderfer-Ladd, 2002; Schwartz et al., 1998),day-in, day-out victimization by peers is a clearlydiscernable feature of the peer environment in earlyelementary school. The data in this report extendprevious research in several ways. First, the effects oftrait victimization clearly extend beyond children’sawareness of their own misery; victimization incre-ments child distress sufficiently to be apparent toteachers and parents. Second, like many insidiousprocesses that mediate long-term change, the dataindicate that the causal relation is reciprocal. Sad-ness, social disengagement, and worry are apparentto peers. Children who frequently display suchbehavior are likely marks for peer rejection andaggression (Boivin et al., 2001), perhaps because theylack sufficient skills to cope with harassment (Perryet al., 1990) or a supportive peer network thatmitigates victimization (Hodges, Boivin, Vitaro, &Bukowski, 1999).

Third, gender differences appear to be importantbut complicated and conditional on the aspect ofadjustment to which victimization is related. Thetiming and processes by which victimization anddistress are linked vary by gender even when the neteffect of the linkage is similar for boys and girls. Themultivariate, reciprocal impact of skills deficits,aggressive behavior, and disturbed peer relationsmay evolve more quickly after the transition toschool for girls than for boys. Boys’ persistingvictimization by peers appears to be an importantsource of learning that more gradually promotesgrowth in antisocial behavior that is readily appar-ent to teachers and, to a lesser degree, to parents.Reciprocally, long-term growth in victimization maybe one consequence of boys’ generalized, frequent,and increasing use of aggressive, annoying behaviorto relate to other people, including peers.

The contrasting short-term suppressive effect ofchild antisocial behavior on victimization by peersand the long-term bilateral, facilitative effects ofgrowth in antisocial behavior and growth in traitvictimization for boys are different from previouslyreported data (Schwartz et al., 1998). The proximalfunction and distal developmental consequences ofantisocial behavior are not necessarily the same, adisparity that serves as a trap to support the long-term persistence and growth of aggressive, antisocialbehavior as a result of its short-term function.Aggressive and disruptive behavior may not servethe same short-term peer group function for girls,and in fact may be ‘‘punished’’ by increasing cross-occasion victimization. This seeming punishmenteffect may actually dampen antisocial developmentin girls, serving as a process that partially accounts

Observed Peer Victimization 1895

Page 16: Observed Peer Victimization During Early Elementary School: Continuity, Growth, and Relation to Risk for Child Antisocial and Depressive Behavior

for gender differences in antisocial behavior thatcontinue to emerge during this developmentalperiod (Moffitt et al., 2001). The differential short-term peer contingencies experienced by boys andgirls for aggressive and assertive behavior, and themanner in which they respond to peers’ rejectionand ostracism, may serve as early processes thatinitiate long-term gender divergence in antisocialand depressive behavior, similar to effects found forsibling relationships (Compton, Snyder, Schrepfer-man, Bank, & Wu-Shortt, 2003).

These findings were derived from a series ofmodels representing short-term group processes andlong-term individual processes, and their combina-tion. To our knowledge, this approach has been usedInfrequently. As such, the current findings needreplication. However, the use of the combined modelis congruent with the notion that developmentalprocesses occur at multiple levels and emerge overvarious time frames. Modeling each of the multipleprocesses one at a time may lead to a less completeunderstanding than efforts to specify their indepen-dent, additive, redundant, and synergistic effects(Curran & Bollen, 2001).

Limitations and Implications for Future Research

The models tested in this report treat childrenwho experience cross-occasion peer victimization asa homogeneous group. Previous research suggeststhat harassment among children is more aptlycaptured by several distinct subgroups: passivevictims, aggressive victims, aggressors only, andchildren who are neither aggressors nor victims.Children in these subgroups bring distinct behavior-al repertoires to the peer group that differentiallyaffect rates of victimization and its growth over time.Passive and aggressive victims may show differentdevelopmental responses to persistent victimization(Schwartz, Proctor, & Chien, 2001). Treating hetero-geneous groups as a homogeneous class mayobfuscate the interrelation of victimization and childadjustment. We will address this issue in subsequentresearch by using latent class growth analysis toascertain the conjoint occurrence of individualtrajectories for trait victimization and trait aggres-sion as observed on the school playground.

The application of a strict parallel reliabilitymodel to observational data provided a useful toolto estimate trait victimization. As such, observationsmay serve as another anchor in a multi-informantand multimethod assessment approach to victimiza-tion, especially for young children (Ladd & Kochen-derfer-Ladd, 2002). However, observational data can

also be used in another way. Inferences about thesocial processes and contingencies by which peervictimization grows for some children and dimini-shes for others, and its relation to child adjustmentduring the transition to school, should be comple-mented by sequential analyses of interaction toascertain what it is that children actually do (or failto do) that evokes peer aggression and how theylearn to behave and to make activity and affiliatechoices that diminish and deflect victimization.

Causal inferences that can be drawn from thesedata remain tentative. The linkage of victimizationand child adjustment may be conditional on theirshared relationship with a third variable such aschild temperament. The models presented here aresurely incomplete and fail to represent a range ofmediating and moderating variables (Hodges et al.,1999; Schwartz et al., 1999), including larger groupcontagion processes occurring at the level of class-room or schools (Kellam, Ling, Merisca, Brown, &Ialongo, 1998). Finally, the degree to which thefindings derived from this single-school communitysample generalize to children, playgrounds, schoolsand communities with a broader range of character-istics is unknown. The sample used here represents arelatively full range of child adjustment in terms ofantisocial behavior, but a relatively restricted rangein terms of child depressive behavior.

References

Achenbach, T. M. (1991). Integrative guide for the 1991 CBCL/4-18, YSR, and TRF profiles. Burlington, VT: University ofVermont, Department of Psychiatry.

Alsaker, F. D. (1993, April). Bully-victim problems in day-carecenters: Measurement issues and associations with children’spsychosocial health. Paper presented at the biennialmeeting of the Society for Research in Child Develop-ment, New Orleans, LA.

Anderson, J. C., & Gerbing, D. W. (1988). Structuralequation modeling in practice: A review and recom-mended two-step approach. Psychological Bulletin, 103,411 – 423.

Arbuckle, J. L., & Wothke, W. (1999). Amos 4.0 user’s guide.Chicago: SmallWaters.

Bank, L. (2000, November). Computer Assisted Child Inter-view for young children (CACI): Psychometric characteristicsand validity. Paper presented at the annual meeting ofthe Association for the Advancement of BehaviorTherapy, New Orleans, LA.

Boivin, M., Hymel, S., & Hodges, E. V. E. (2001). Toward aprocess view of peer rejection and harassment. In J.Juvonen & S. Graham (Eds.), Peer harassment in school:The plight of the vulnerable and victimized (pp. 265 – 289).New York: Guilford.

1896 Snyder et al.

Page 17: Observed Peer Victimization During Early Elementary School: Continuity, Growth, and Relation to Risk for Child Antisocial and Depressive Behavior

Buhs, E. S., & Ladd, G. W. (2001). Peer rejection as amediator of young children’s school adjustment: Anexamination of mediating processes. DevelopmentalPsychology, 37, 550 – 560.

Cairns, R. B., & Green, J. A. (1979). How to assesspersonality and social patterns: Observations or ratings.In R. B. Cairns (Ed.), The analysis of social interactions:Methods, issues and illustrations (pp. 209 – 226). Hillsdale,NJ: Erlbaum.

Caspi, A., Elder, G. H., & Bem, D. J. (1987). Moving againstthe world: Life course patterns of explosive children.Developmental Psychology, 23, 303 – 308.

Coie, J. D., & Kupersmidt, J. B. (1983). A behavioralanalysis of emerging social status in boys’ peer groups.Child Development, 54, 1400 – 1416.

Compton, K., Snyder, J., Schrepferman, L., Bank, L., & Wu-Shortt, J. (2003). The contribution of parents and siblingsto antisocial and depressive behavior in adolescence: Adouble jeopardy coercion model. Development andPsychopathology, 15, 162 – 182.

Crick, N. R., & Grotpeter, J. K. (1996). Children’s treatmentby peers: Victims of relational and verbal aggression.Development and Psychopathology, 8, 367 – 380.

Curran, P. J., & Bollen, K. A. (2001). The best ofboth worlds: Combining autoregressive and latentgrowth models. In L. M. Collins & A. G. Sayer(Eds.), New methods for the analysis of change (pp. 105 –136). Washington, DC: American PsychologicalAssociation.

Curran, P. J., Stice, E., & Chassin, L. (1997). The relationbetween adolescent alcohol use and peer alcohol use: Alongitudinal random coefficients model. Journal ofConsulting and Clinical Psychology, 65, 130 – 140.

Fagot, B., Hagen, R., Leinbach, M. D., & Kronesberg, S.(1985). Differential reactions to assertive and commu-nicative acts of toddler boys and girls. Child Develop-ment, 56, 1499 – 1505.

Hanish, L. D., & Guerra, N. G. (2002). A longitudinalanalysis of patterns of adjustment following peervictimization. Development and Psychopathology, 14,69 – 91.

Hodges, E. V. E., Boivin, M., Vitaro, F., & Bukowski, W. M.(1999). The power of friendship: Protection against anescalating cycle of peer victimization. DevelopmentalPsychology, 75, 94 – 101.

Hodges, E. V. E., & Perry, D. G. (1999). Personal andinterpersonal antecedents and consequences of victimi-zation by peers. Journal of Personality and SocialPsychology, 76, 677 – 685.

Jones, R. R., Reid, J. B., & Patterson, G. R. (1975).Naturalistic observation in clinical assessment. In P.McReynolds (Ed.), Advances in psychological assessment(Vol. 3, pp. 42 – 95). San Francisco: Jossey-Bass.

Kellam, S. G., Ling, X., Merisca, R., Brown, C. H., &Ialongo, N. (1998). The effect of the level of aggression inthe first grade classroom on the course and malleabilityof aggressive behavior into middle school. Developmentand Psychopathology, 10, 165 – 185.

Kochenderfer, B. J., & Ladd, G. W. (1996a). Peer victimiza-tion: Cause or consequences of school maladjustment?Child Development, 67, 1305 – 1317.

Kochenderfer, B. J., & Ladd, G. W. (1996b). Peer victimiza-tion: Manifestations and relations to school adjustmentin kindergarten. Journal of School Psychology, 34, 267 –283.

Kovacs, M. (1980). Rating scales to assess depression inschool-aged children. Acta Paedopsychiatria, 46, 305 – 315.

Ladd, G. W., Birch, S. H., & Buhs, E. (1999). Children’ssocial and scholastic lives in kindergarten: Relatedspheres of influence? Child Development, 70, 1373 – 1400.

Ladd, G. W., & Kochenderfer-Ladd, B. (2002). Identifyingvictims of peer aggression from early to middle child-hood: Analysis of cross-informant data for concordance,estimation of relational adjustment, prevalence ofvictimization, and characteristics of identified victims.Psychological Assessment, 14, 74 – 96.

Loeber, R., & Dishion, T. J. (1984). Boys who fight at homeand school: Family conditions influencing cross-situa-tional consistency. Journal of Consulting and ClinicalPsychology, 52, 759 – 768.

Maccoby, E. E. (1998). The two sexes: Growing up apart,coming together. Cambridge, MA: Belnap Press.

Moffitt, T. E., Caspi, A., Rutter, M., & Silva, P. A. (2001). Sexdifferences in antisocial behaviour: Conduct disorder, delin-quency, and violence in the Dunedin longitudinal study.New York: Cambridge University Press.

Parker, J. G., Rubin, K. H., Price, J. M., & DeRosier, M. E.(1995). Peer relationships, child development andadjustment: A developmental psychopathology per-spective. In D. Cicchetti & D. J. Cohen (Eds.), Develop-mental psychopathology. Vol. 2: Risk, disorder and adaptation(pp. 65 – 92). New York: Wiley.

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

Patterson, G. R., Littman, R. A., & Bricker, W. (1967).Assertive behavior in children: A step towards a theoryof aggression. Monographs of the Society for Research inChild Development, 32(5), 1 – 43 (Serial No. 5 – 10).

Pellegrini, A. D. (2001). Sampling instances of victimiza-tion in the middle school: A methodological compar-ison. In J. Juvonen & S. Graham (Eds.), Peer harassment inschool: The plight of the vulnerable and victimized (pp. 125 –144). New York: Guilford.

Perry, D. G., Perry, L. C., & Boldizar, J. P. (1990). Learningof aggression. In M. Lewis & S. Miller (Eds.), Handbook ofdevelopmental psychopathology (pp. 135 – 146). New York:Plenum.

Schwartz, D., McFadyen-Ketchum, S., Dodge, K. A., Pettit,G. S., & Bates, J. E. (1998). Peer group victimization as apredictor of children’s behavior problems at home andin school. Development and Psychopathology, 10, 87 – 99.

Schwartz, D., McFadyen-Ketchum, S. A., Dodge, K. A.,Pettit, G. S., & Bates, J. E. (1999). Early behavior prob-lems as a predictor of later peer group victimization:Moderators and mediators in the pathways of socialrisk. Journal of Abnormal Child Psychology, 27, 191 – 201.

Observed Peer Victimization 1897

Page 18: Observed Peer Victimization During Early Elementary School: Continuity, Growth, and Relation to Risk for Child Antisocial and Depressive Behavior

Schwartz, D., Proctor, L. J., & Chien, D. H. (2001). Theaggressive victim of bullying: Emotional and behavioraldysregulation as a pathway to victimization by peers. InJ. Juvonen & S. Graham (Eds.), Peer harassment in school:The plight of the vulnerable and victimized (pp. 147 – 174).New York: Guilford.

Shantz, C. U., & Hobart, C. J. (1994). Social conflict anddevelopment. In C. U. Shantz (Ed.), Conflict in child andadolescent development (pp. 71 – 94). New York: Cam-bridge University Press.

Snyder, J. (2002). Reinforcement and coercion mechanismsin the development of antisocial behavior: Peer relation-ships. In J. B. Reid, G. R. Patterson, & J. Snyder (Eds.),Antisocial behavior in children and adolescents: A develop-mental analysis and model for intervention (pp. 101 – 122).Washington, DC: American Psychological Association.

Snyder, J., Edwards, P., McGraw, K., Kilgore, K., & Holton,A. (1994). Escalation and reinforcement in mother-childconflict: Social processes associated with the develop-ment of physical aggression. Development and Psycho-pathology, 6, 305 – 321.

Snyder, J., Horsch, E., & Childs, J. (1997). Peer relationshipsof young children: Affiliative choices and the shaping ofaggressive behavior. Journal of Clinical Child Psychology,26, 145 – 156.

Stoolmiller, M., Eddy, J. M., & Reid, J. B. (2000). Detectingand describing preventive intervention effects in auniversal school-based randomized trial targeting de-linquent and violent behavior. Journal of Consulting andClinical Psychology, 68, 296 – 306.

Walker, H. M., Colvin, G., & Ramsey, E. (1995). Antisocialbehavior in school: Strategies and best practices. PacificGrove, CA: Brooks/Cole.

Weiss, B., Dodge, K. A., Bates, J. E., & Pettit, G. (1992).Some consequences of early harsh discipline and amaladaptive social information processing style. ChildDevelopment, 63, 1321 – 1335.

Wothke, W. (2000). Longitudinal and multi-group model-ing with missing data. In T. D. Little, K. U. Shnabel, & J.Baumert (Eds.), Modeling longitudinal and multi-level data:Practical issues, applied approaches, and specific examples(pp. 219 – 240). Mahwah, NJ: Erlbaum.

1898 Snyder et al.