Available online at www.sciencedirect.com

Behavior Therapy 41 (2010) 317–328www.elsevier.com/locate/bt

Contribution of Peer Deviancy Training to the Early Developmentof Conduct Problems: Mediators and Moderators

James SnyderAmber McEachernLynn Schrepferman

Christy JustMelissa JenkinsShani Roberts

Ashton LofgreenWichita State University

Three variables were tested as moderators of the relation-ship between peer deviancy training and child antisocialbehavior in a longitudinal study of 267 boys and girls fromages 5.3 to 9.3 years. Deviancy training was directlymeasured by observation of the discourse and play ofchildren with same-gender classmates. Peer deviancy train-ing was significantly related to multi-setting child antisocialbehavior from ages 5.3 to 9.3 years. Child impulsivity, poorparental discipline, and peer rejection were all significantmoderators of that relationship, even in the context of theirdirect association with trajectories of antisocial behaviorand after controlling for deviant peer affiliation. Thesemoderator effects appeared to be associated with children'sincreased sensitivity to peer modeling and reinforcement ofdeviant discourse and play. Not all children are equallyaffected by peer deviancy training, and an array ofintervention strategies are described that may serve toprotect children from deviant peer influence.

RESEARCH HAS REPEATEDLY FOUND that affiliation withdeviant peers substantially increases risk for the

This research was supported by National Institute of MentalHealth Grant MH57342.

Address correspondence to James Snyder, Department ofPsychology, Box 34, Wichita State University, Wichita, KS 67260-0034; e-mail: james.snyder@wichita.edu.0005-7894/10/317–328/$1.00/0© 2010 Association for Behavioral and Cognitive Therapies. Published byElsevier Ltd. All rights reserved.

emergence and persistence of antisocial behavior(Dodge, Dishion, & Lansford, 2006). The negativeeffects of deviant peer affiliation have been ob-served during childhood (Hanish, Martin, Fabes,Leonard, & Herzog, 2005; Snyder, Horsch, &Childs, 1997) adolescence (Patterson, Dishion, &Yoerger, 2000) and early adulthood (e.g., Duncan,Boisjoly, Kremer, Levy, & Eccles, 2005; Gardner,Dishion, & Connell, 2008). Exposure to deviantpeers may result from naturally occurring inciden-tal contacts (e.g., with classmates), selective affili-ation (e.g., in friendships or cliques) among youthin schools and neighborhoods (Dishion, Nelson,Winter, & Bullock, 2004; Kellam, Ling, Merisca,Brown, & Ialongo, 1998), and aggregation of at-risk youth for group interventions (e.g., for skillstraining or residential care) (Lavallee, Bierman,Nix, & CPPRG, 2005; Lipsey, 2006; Poulin,Dishion, & Burraston, 2001). These affiliationsthen result in deviancy training in which youth areexposed to peers' modeling and reinforcement ofdeviant behavior (Dishion, Spracklen, Andrews, &Patterson, 1996).This research has resulted in calls to proactively

manage aggregation of at-risk youth and to imple-ment explicit tactics to mitigate the potentialnegative effects when such aggregation occurs(Dishion & Dodge, 2005). However, to effectivelyrespond to these calls, the social mechanisms thatmediate and the conditions that moderate deviantpeers' influence on trajectories of antisocial behavior

318 snyder et al .

need to be better understood. This report examinesmediators and moderators of deviant peer influencein school settings during early elementary school.More specifically, two interlocking hypotheses aretested. First, peer modeling and reinforcement arehypothesized to be associated with children's displayof deviant discourse and play, and children's deviantdiscourse and play to be associated with increasedrisk for antisocial behavior—a mediator effect.Second, child impulsivity, poor parental discipline,and peer rejection are hypothesized to augment therelationship of peer modeling and reinforcement tochild deviant discourse and play by increasingchildren's sensitivity to peer modeling and reinforce-ment—a moderator effect.

Mediators of the Association Relationship ofDeviant Peer Affiliation and Antisocial BehaviorSelection and shaping are often suggested asmediators of the relationship between deviant peeraffiliation and risk for antisocial behavior (Gifford-Smith, Dodge, Dishion, & McCord, 2005). Theirrelative importance has been studied by applyingauto-regressive cross-lagged or multi-domain latentgrowth curve analyses to multi-method longitudi-nal data to examine the temporal ordering ofdeviant peer affiliation and antisocial behavior.These studies indicate both selection and shapingoperate, reflecting a reciprocal influence betweendeviant peer affiliation and antisocial behavior(Dishion & Owen, 2002; Laird, Criss, Pettit,Dodge, & Bates, 2008; Wills & Cleary, 1999).However, these optimally designed longitudinal

studies do not specify or measure the social-experiential mediating processes by which youths'behavior is shaped as a result of interaction withdeviant peers (Dishion & Dodge, 2005). Although anumber of processes have been postulated, peerreinforcement and modeling of antisocial discourseand behavior (often labeled deviancy training) havereceived the most empirical attention (Dishion &Dodge, 2006). A basic premise of deviancy trainingis that youths' discourse and activities during peerinteraction is shaped by attention and approval.When interacting with deviant peers, youths' dis-course about and rehearsal of norm-violatingbehaviors are likely to result in laughter, approval,and excitement, and to be shaped by directlyexperienced peer contingencies and by vicariousobservation of contingencies received by peers. Assuch, peer deviancy training provides instruction andpractice about how, when, and where to engage innorm-violating behavior, and information aboutwhat consequences may accrue for that behavior.As a result of equivalence functions of language and

verbal rules about the relation of behavior-contin-gency relationships (Drossel, Waltz, & Hayes,2007), discourse and rehearsal of norm-violatingbehaviors and the positive peer consequences theyengender serve as cognitive templates and verbalrules promoting the performance of antisocialbehavior in other social ecologies.Deviancy training mediates the association of

deviant peer affiliation and antisocial behavior viatwo processes. First, the rates at which childrenengage in deviant talk and play are associated withthe rates at which they are reinforced by peers forsuch talk and play, and at which children observepeers getting reinforced for deviant talk and play(Dishion et al., 1996; Snyder et al., 2004). Second,rates at which youth engage in deviant talk and playduring peer interaction is prospectively linked toantisocial behavior at home, in school, and in theneighborhood during childhood (Snyder et al.,2005) and adolescence (Dishion et al., 1996;Patterson et al., 2000). It is our hypothesis thatthree proposed moderator variables (poor inhibi-tory control, poor parental discipline, and peerrejection) operate on the first process by augment-ing children's sensitivity to the excitement, titilla-tion, and attention engendered by deviant discourseand play during peer interaction.

Moderators of the Effects of Deviant Peers onthe Development of Antisocial Behavior

Exposure to and involvement with deviant peersdoes not necessarily or uniformly facilitate antiso-cial behavior; not all youth are equally affected.Three types of variables appear to amplify ormitigate the relationship of deviant peer exposureto the development of antisocial behavior: thechild's experiential history in the peer group (peerrejection), children's temperament (impulsivity),and parenting practices (poor discipline).Research on the social relationships of children

has clearly documented that aggressive children arelikely to be rejected by normative peers and thatsuch rejection increments risk for subsequentantisocial behavior (Kupersmidt & Dodge, 2004).Rejection by normative peers may increment peerdeviancy training in two ways. First, rejectedchildren may spend more time with other aggressivechildren, either as a result of aggregation of childrenwho are excluded from the normative peer groupor as a result of active mutual selection (Dishion,Patterson, & Griesler, 1994). These increased con-tacts with deviant peers may result in a greaterexposure to antisocial discourse and activities, andshape a susceptibility to peer approval of such dis-course and activities (Snyder et al., 2005). Second,

319p e er dev i ancy tra in ing

social rejection and exclusion reduces children'saccess to peer social attention for constructive andskilled behavior (Snyder et al., 2004). As a result,rejected children may actually show increasedsensitivity to peer reinforcement (Dishion, Piehler,& Myers, 2008) and may be more likely to utilizedeviant discourse and play as a means of generatingsuch reinforcement.The facilitating main effect of inadequate inhib-

itory control on antisocial development is welldocumented (Nigg & Huang-Pollock, 2003). Poorinhibitory control also appears to moderate theassociation of deviant peer affiliation with antiso-cial behavior during adolescence. Youth withinadequate effortful control are more susceptibleto the negative influence of deviant friends (Dishion& Patterson, 2006; Gardner et al., 2008). Similarly,the impact of deviant peer affiliation on antisocialbehavior may be moderated by sensitivity to short-term rewards. Goodnight, Bates, Newman, Dodge,and Pettit (2006) reported the relation betweendeviant peer affiliation and later externalizingbehaviors was greater for youth less able to inhibitbehavior in the context of reward than those withbetter inhibitory skills. Deviant peer influenceappears to be engendered by the excitement, exhi-laration, and short-term reward antisocial dis-course and behavior generates among youth, andto be mitigated by an ability to inhibit norm-violating responses supported by short term posi-tive consequences. The degree to which effortful,inhibitory control similarly moderates the relation-ship between deviant peer influence and antisocialbehavior during childhood has not been examined.Socialization in the family also appears to

influence children's sensitivity to deviant peerinfluence. Good parental monitoring has beenfound to mitigate the association of deviant peeraffiliation with antisocial behavior during adoles-cence (Laird et al., 2008) as well as to have powerfulmain effects by limiting youths' affiliation withdeviant peers (Dishion & McMahon, 1998). Themoderator effects of other parenting variables havereceived little empirical attention (Vitaro, Brendgen,& Tremblay, 2000). This is surprising given youths'ability to resist temptation and short-term reward inthe relative absence of adult tracking and contin-gencies is a widely accepted criterion of successfulsocialization. Given nearly all youth are repeatedlyand sometimes extensively exposed to deviant peers,even in relatively normative (Snyder et al., 2005)and advantaged environments (Luthar & Becker,2002), the lack of research on protective factorsresulting from family socialization during childhoodis surprising. Parental discipline has a powerfuldirect effect on antisocial development during

childhood (Snyder, Reid, & Patterson, 2003), andis a prime candidate as a moderator of children'ssusceptibility to deviant peer influence.In summary, not all youth are necessarily and

equally affected by contacts with deviant peers.Some youth may be more vulnerable and othersmore immune to the influence of deviant peers.Identification of variables which attenuate orexacerbate the contribution of deviant peers toantisocial behavior prior to adolescence is clearlywarranted and important. Three promising mod-erator variables are examined in this report: childimpulsivity, parental discipline, and peer rejection.Identification of moderator variables may enablebetter identification of individuals at heightenedrisk for deviant peer influence, and lead to a betterarticulation of intervention tactics that reduce riskfor the early development of antisocial behaviorengendered by exposure to deviant peers.

Methodparticipants

The participants were 133 girls and 134 boyswhose mean age was 5.3 years at the initial datacollection point (entry to kindergarten) and 9.3years at the last data collection point (in fall fourthgrade). This community sample was obtained byrecruiting kindergarten children (and their parents)enrolled in one elementary school over 3 consecu-tive years. The school served a low socio-economicneighborhood in a city with a population of350,000. Seventy-one percent of the children wereEuropean American, 19% African American, 5%Hispanic/Latino, and 3% Native American, and2%were Asian American. At age 5.3 years, 43% ofchildren lived in intact families with two-biologicalparents, 28% in single parent families, 21% inblended family households, and 7% in other familyconfigurations. The median family income was$25,000; 28% of the children lived in families withincomes below the poverty line. Forty-six percent ofthe parents had completed high school, and 20%had less than a high school education. Seventy-fivepercent of two-parent families were comprised ofdual wage earners, and 9% of the families werewithout any employed adult.Data collection was continued after participants

made school or residential moves. Data wereavailable for over 95% of the 267 children meanages 5.8, 6.3 and 6.8 years, and for 77% of thechildren at mean age 9.3 years. Missing data wereestimated in AMOS 4.0 using the full informationmaximum likelihood (FIML) estimation method.FIML does not delete cases missing from one ormore waves of data collection, nor does it delete

320 snyder et al .

cases missing one or more variables within a waveof data collection. This procedure avoids biasedparameter estimates that are likely to occur if pair-wise or list-wise deletion procedures are used tocompensate for missing data (Worthke, 2000).

measures

Deviancy TrainingA construct for peer deviancy training was derivedfrom videotapes of triadic peer interaction of each ofthe participating children, obtained on multipleoccasions from ages 5.3 to 5.8 years. Each child wasinvolved as the target on three occasions, and as anontargeted coparticipating classmate on a mini-mum of six occasions. Combinations of same-gender classmates were sampled across occasionsin a random fashion. On each occasion, the targetchild and his/her same-gender classmates engaged intwo tasks. During the first 15 minutes, the childrenplayed structured age-appropriate interactivegames, and during the second 15 minutes engagedin free play.These multi-occasion samples of peer interaction

(total=90 minutes for each child as target; 180minutes as nontarget) were coded using the Anti-social Content Code (ACC; Oeser & Schrepferman,2002). ACC codes the content of children's dis-course and play and peers' responses to suchbehavior during ongoing interaction. ACC providesreal-time coding of the occurrence of each of sixforms of norm-violating discourse and play: alcoholand tobacco, sex, aggression and property destruc-tion, authority defiance, sneaky actions (stealing,cheating, or lying), and “gross” body functions.ACC also dichotomously codes the responses ofpeers to each display of deviant discourse and playby the target child as positive (agreement, laughter,or other active positive attention) or not positive(ignoring, distress, complaints, dismay, correction,disapproval, or threats to tell an adult). After each15-minute segment of peer interaction, ACC codersmade Likert ratings of how much each nontargetchild engaged in deviant discourse and play andreceived positive responses for their deviant dis-course and play from the other nontarget child.Coders were trained to a criterion of 65%

agreement on real-time coding of specific categoriesof deviant discourse and play and peer responses,and to a criterion intra-class correlation of .70 onthe Likert scale ratings. Weekly recalibration meet-ings were held to minimize observer drift and tomaintain coder reliability. Ongoing coder agree-ment was estimated for 230 (32%) of the 723observation occasions available for this sample.Percent agreement and Kappa indices of coderreliability for the occurrence/nonoccurrence of

specific deviant discourse or play categories were87.5% and .74, respectively. Percent agreement andKappa coder reliability for peer responses todeviant talk or play were 77.1% and .60, respec-tively. The mean intra-class correlation for coders'ratings of deviant discourse and play was .78, andfor peer positive responses to such discourse andtalk was .71.Two composite variables were derived from the

ACC codes and ratings. The first compositeassessed each child's deviant discourse and play.This composite was defined by the sum of thefollowing three variables, after each was averagedacross observation occasions and standardized:rpm observed deviant talk and play when childwas target, ratings of child's deviant discourse andplay when the child was a nontarget, and rating ofpositive responses for peer deviant discourse andplay made by the child as non-target. Using thesame aggregation approach, the second compositefor peer modeling and reinforcement of deviantdiscourse and play was defined by the sum of thefollowing three variables: observed proportion ofpositive responses made by peers for the targetchild's deviant discourse and play, ratings of peers'deviant discourse and play observed by the targetchild, and ratings of the positive responses of onenontarget peer to the deviant talk and play of thesecond non-target peer observed by the target child.

Deviant Peer AffiliationDeviant peer affiliation was a composite constructdefined by the sum of the two standardizedvariables assessed at child age 5.3 years: the meanexternalizing scores on the parent Child BehaviorChecklist (Achenbach, 1991) and the TeacherReport Form (Achenbach, 1991) of the threeclassmates the child nominated as preferred play-mates. The correlation between teacher and parentreports of preferred playmates' externalizing pro-blems was .38 (pb .05)

Child Impulsivity. A composite for a poorexecutive control or impulsivity construct wasdefined by the mean of three measures (after eachwas standardized) obtained from ages 5.3 to 5.8years. The first measure was the mean of four scores(after each was standardized) from two individuallyadministered tests: the Trails Test B time and errors(Kindlon, Mezzacappa, & Earls, 1995), and digitspan forward and backward (reverse scored) fromtheWISC-III (Wechsler, 1991). The secondmeasurewas the mean percent of time off-task during 5-minute periods of academic work, obtained onthree separate occasions (Walker & Severson,1991). Mean observer agreement as indicated byintra-class correlation of off task time was .72. The

321p e er dev i ancy tra in ing

third measure consisted of the mean of assessors'ratings of the degree to which each child wasinattentive across three occasions of individually-administered assessment tasks. Mean observeragreement as indicated by intra-class correlationwas .69. The correlations among the three measurescomprising the impulsivity composite ranged from.30 to .40 (all pb .01).

Peer RejectionClassmates' nominations of three children whomthey preferred and three children whom they didnot prefer as playmates were obtained in the fall ofkindergarten using a picture sociometric technique(Asher, Singleton, Tinsley, & Hymel, 1979). Thenumber of nominations each child received aspreferred and as nonpreferred playmates wassummed and standardized by classroom. Peerrejection scores were defined by subtracting thestandardized score for nominations as preferredplaymates from the standardized score for nomina-tions as nonpreferred playmates.

Parental DisciplineThe interaction of each child with a parent wasvideotaped for 2 hours on each of two occasionsbetween child ages 5.3 to 5.8 years. Each occasionwas structured around a set of verbally cued tasks,including playing a novel interactive game, freeplay, clean up, talking about the child's day atschool, problem solving, and practice arithmeticand word recognition. The interaction was codedusing the Family Peer Process Code (FPP; Stubbs,Crosby, Forgatch, & Capaldi, 1998) and theSpecific Affect Coding System (SPAFF; Gottman,McCoy, Coan, & Collier, 1995). FPP codes theoccurrence of parent and child verbal and non-verbal social behavior into one of 24 mutuallyexclusive and collectively exhaustive categoriesalong a real time line, including both the onsetand offset of the behavior. FPP coders made Likertratings of parent behavior after each 15-minuteepisode of micro-coding. SPAFF codes the occur-rence of parent and child behavior into 19 mutuallyexclusive and collectively exhaustive categoriesalong a real time line, giving coding priority tothe display of seven discrete emotions based on agestalt of facial, vocal, and physical features ofbehavior. FPP and SPAFF coders were trained to acriterion agreement of 75% prior to coding data,and weekly recalibration and coding trainingsessions were used to minimize drift. Fifteen percentof all observations were used to assess coderreliability, and coders were unaware of whichobservation sessions were to be used for reliabilityassessment.

Four indicators from the FPP and SPAFF dataand one from a structured parent interview (Snyder,Cramer, & Patterson, 2005) were combined todefine a composite index of poor parental disci-pline. The first indicator was the rate per minutewith which parents observed (FPP) to directaversive verbal and nonverbal behaviors towardthe child. The intra-class correlation index of coderagreement on the rate per minute parent aversivewas .87. The second indicator was an odds ratioindexing the likelihood that parents gave in to childaversive behavior (FPP) during a discipline episode.The third indicator was the mean of FPP coders'ratings on six items describing parents' use ofineffective disciplinary responses: relies on negativeaffect, uses ridicule and sarcasm, grabs or hits,threatens punishment, is erratic or inconsistent, andoverly strict and oppressive. The intra-class corre-lation index of coder agreement on these Likertratings was .84, and the scale alpha was .79. Thefourth indicator was the cross-occasion mean of therate per minute with which parents were observedto direct anger and contempt (SPAFF) at theirchildren. Average coder agreement on the SPAFFwas 83% (kappa=.73). The fifth indicator wasdefined by parents' rating of likelihood they woulduse “tough” discipline tactics in response to fivecommon problem child behaviors (e.g., swatting achild for throwing a tantrum for candy at the store).The alpha for this scale was .66. The five indicatorsof the poor discipline composite were inter-corre-lated from .27 to .46 (all psb .01) and significantlyloaded on a single construct (all bsN .33, psb .05;model fit p=.32). Standardized scores of the fiveindicators were averaged to create a composite ofpoor discipline.

Multi-Setting Child Antisocial BehaviorChild antisocial behavior was assessed in each ofthree social ecologies (home, classroom, andschool playground) at child ages 5.3, 5.8, 6.3,and 6.8 years. These data were used to estimatea trait construct for children's early antisocialbehavior.

Antisocial behavior on the playground. Anadaptation of the observation system described byWeiss, Dodge, Bates, and Pettit (1992) was used tocode children's behavior on the playground. Thisinterval coding system classifies child behavior intoone of several mutually exclusive and collectivelyexhaustive categories every 10 seconds; the catego-ry used in current research ascertained the targetchild's physical aggression toward peers. Codersalso made a series of 5-point Likert item ratingsafter each 5-minute observation occasion to assesslow base-rate behaviors. One item, “the child

322 snyder et al .

engaged in sneaky behavior,” was used to assesscovert antisocial behavior on the playground.Coders were trained to a minimum Kappa agree-ment of .70 prior to data collection at each of theage periods at which playground observations wereobtained. Observer agreement was assessed on10% of all occasions. Interval-by-interval coderagreement on physical aggression ranged from65% to 69%. Intra-class correlations of the rpmof physical aggression (the metric used in thecurrent analyses) generated by independent coderpairs ranged from .62 to .84 (all psb .001). Coderagreement on ratings of sneaky behavior as indexedby intra-class correlations averaged .77.Behavior of participating children was observed

on the school playground on six separate occasionsat child ages 5.3 and 5.8 years, and on four separateoccasions at ages 6.3 and 6.8 years. On eachoccasion, the behavior of each child was observedand coded for 5 minutes (i.e., 30 consecutive 10-second intervals). The mean cross-occasion rate perminute at which children directed physical aggres-sion toward peers, and the mean ratings of “sneakybehavior” at each of the age periods was calculatedand used as indicators of the trait antisocialbehavior construct.

Antisocial behavior at home. Parents' ratingsof the behavior problems on the aggression anddelinquency scales of the Child Behavior Checklist(Achenbach, 1991) were collected at child ages5.3, 5.8, 6.3, and 6.8 years. Alpha reliabilities ofthese scales in this sample were greater than .70 ateach developmental point. These mean of the twoscales at each age period was calculated and usedas indicators of the construct of trait antisocialbehavior.

Table 1Descriptive Statistics and Correlations

1 2 3

1. Peer rejection2. Impulsivity .523. Poor discipline .13 .244. Child Deviant discourse/play .22 .20 .165. Peer deviancy model-ing & reinforcement .06 .05 -.076. Deviant peer affiliation .07 .06 -.047. Child AS (5.3 yrs) .44 .45 .208. Child AS (5.8 yrs) .42 .34 .189. Child AS (6.3 yrs) .47 .45 .2210. Child AS (6.8 yrs) .42 .37 .2911. Child AS (9.3 yrs) .24 .27 .22Mean .00 .00 .00S.D. 1.00 1.00 1.00Skewness 0.21 1.97 2.04

Note. Correlations shown in bold are reliable at pb .05; means, staaggregation, summation of standard scores and/or square root transfor

Antisocial behavior in the classroom. Tea-chers' ratings of the behavior problems on theaggression and delinquency scales of the TeacherReport Form (Achenbach, 1991) were obtained atchild ages 5.3, 5.8, 6.3, and 6.8 years. Differentteachers typically completed the TRFs at the 5- and6-year-old periods. Internal reliabilities of thesescales in this sample were alpha N .60 at each ageperiod. The mean of the two scales from each of theage periods was calculated and used as indicators ofthe construct of trait antisocial behavior.

Youth Self-Reported Antisocial BehaviorChildren's self-reported behavior problems werecollected at age 9.3 years, using the externalizingscale of the Youth Self Report (YSR; Achenbach,1991). Previous research has thoroughly estab-lished the validity and reliability of the YSR(Achenbach, 1991). The internal reliability of theexternalizing scale in this sample was alpha= .87.

Resultsanalysis plan

Descriptive statistics and correlations for the com-posite variables are shown in Table 1. Aftertransformation, only two variables were skewedslightly more than 2.0, indicating significant devia-tion of distributions from normality was not a majorconcern (Kline, 2005). All subsequent analyses usedsquare root transformed variables. The correlationsindicate two of the proposed moderator variables,impulsivity and peer rejection, shared 27% commonvariance. The correlations of poor parental disciplinewith peer rejection and child impulsivity were low,suggesting a relative lack of redundancy among these

4 5 6 7 8 9 10 11

.62

.22 .28

.35 .23 .21

.29 .18 .21 .55

.45 .27 .28 .71 .70

.30 .13 .13 .59 .68 .74

.12 .30 .12 .33 .38 .26 .22

.04 .00 -.01 1.27 1.39 1.37 1.45 3.17

.91 .79 4.54 0.42 0.42 0.43 0.47 1.082.00 1.54 1.78 2.11 1.55 1.51 1.93 0.99

ndard deviations and skewness for constructs are shown aftermation.

323p e er dev i ancy tra in ing

variables. Each of the proposed moderator variables,child deviant discourse and play, and deviant peeraffiliation were reliably correlated with measures ofchild antisocial behavior at ages 5.3, 5.8, 6.3, 6.8 and9.3 years.

baseline, direct effects models

The fit of a baseline model (shown by solid paths andcomposite variables in Figure 1) without the pro-posed moderator variables (shown as dotted pathsand composite variables) to the data was adequate.As hypothesized, peermodeling and reinforcement ofdeviant discourse andplaywere reliably related to thechild's deviant discourse and play. Children's deviantdiscourse and play, in turn, predicted the constructfor multi-setting antisocial behavior from ages 5.3 to6.8 years. Children's deviant discourse and play alsopredicted self-reported antisocial behavior at age 9.3years, after controlling for earlier antisocial behavior.This supports the mediator hypothesis. As expected,deviant peer affiliation at age 5.3 years wasassociated with antisocial behavior at ages 5.3 to6.8 years, but was not a predictor of antisocialbehavior at 9.3 years of age. The lack of a relationbetween deviant peer affiliation and child deviantdiscourse and play was expected because children's

FIGURE 1 Contribution of Peer Deviancy TrainiModerator Model.

discourse and play were assessed during interactionwith a random selection of classmates rather thanpreferred peers.The results of models including variables hypoth-

esized to moderate the association of peer modelingand reinforcement with children's deviant discourseand play, and to have direct relationships to othervariables in the model as shown by dotted lines inFigure 1 are reported in Table 2, referring to pathsmarked “a,” “b,” “c” and “d” in Figure 1. Eachvariable contributing to the moderator constructswas centered and then multiplied to create aninteraction term. A model for each moderatorvariable was tested separately. Each model (nowincluding all constructs and paths shown in Figure1) fit the data adequately: X2

(33,267) b 45.00,pN .06, CFI N .970, RMSEA b .040. In eachmoderator model, the path from child deviantdiscourse and play to multi-setting antisocialbehavior from ages 5.3 to 6.8 years was reducedslightly from that shown in Figure 1, but alwaysgreater than b=.27 (all psb .001). The path fromantisocial behavior from ages 5.3 to 6.8 years toantisocial behavior at age 9.3 years was alsoreduced slightly in the full moderator models, butalways greater than b=.25 (all psb .01). The path

ng to Child Antisocial Behavior: Baseline and

Table 2Results From the Moderator Models

AddedVariable

Paths Increment in R2

“a” “b” “c” “d” child deviant discourse 5.3-6.8 yrs antisocial 9.3 yr antisocial

impulsivity .14 ⁎ .15 ⁎⁎ .44 ⁎⁎⁎ .12 .04 .17 .00poor discipline .12 ⁎ .20 ⁎⁎⁎ .29 ⁎⁎⁎ .11 ⁎ .07 .10 .02peer rejection .11 ⁎ .15 ⁎⁎ .41 ⁎⁎⁎ .07 .04 .15 .00

Note. Path “a” refers to how the added variable moderates the relationship of peer modeling and reinforcement of deviant discourse and playto child deviant discourse and play; path “b” refers to the direct association of the added variable on child deviant talk and play; path “c” refersto the direct association of the added variable on kindergarten-first grade chronic antisocial behavior: and path “d” refers to the directassociation of the added variable on fourth grade antisocial behavior (see Figure 1 for a graphic display).

⁎ pb .05.⁎⁎ pb .01.

⁎⁎⁎ pb .001.

324 snyder et al .

coefficients and statistical reliability of the remain-ing paths were nearly the same in the baseline andmoderator models.As shown in the second column (path b) in Table

2, child impulsivity (b=.15, pb .01), poor parentaldiscipline (b= .20, pb .001) and peer rejection(b=.15, pb .01) were each directly associated withhigher rates of child deviant discourse and play. Asshown in the third column (path c) in Table 2, childimpulsivity (b=.44, pb .001), poor parental disci-pline (b=.29, pb .001) and peer rejection (b=.41,pb .001) were each directly associated with higherlevels of child antisocial behavior at ages 5.3 to 6.8years. As shown in the fourth column (path d) inTable 2, only parental discipline (b=.11, pb .05)was associated with child self-reported antisocialbehavior at age 9.3 years, after controlling forearlier antisocial behavior at ages 5.3 to 6.8 years.

hypothesized moderator models

The results of the analysis ascertaining the hypoth-esized occurrence of moderator effects on children'ssensitivity to peer modeling and reinforcement of

FIGURE 2 Simple Slopes of the Relationship of Peer Reinforce-ment and Modeling of Deviant Discourse and Play to Rates ofChild Norm-Violating Discourse and Play at High (+1/2 SD) andLow (-1/2 SD) Levels of Impulsivity.

deviant discourse and play are shown in the firstcolumn (labeled “a”) in Table 2. Child impulsivityreliably moderated the relation of peer modelingand reinforcement of deviant discourse and play tothe rates at which children engaged in suchdiscourse and play (b= .14, pb .05). FollowingAiken and West (1991), simple slopes were plottedfor the relationship of peers' reinforcement andmodeling of deviant discourse and play to rates ofchild deviant discourse and play at high (+1/2 SD)and low (-1/2 SD) levels of impulsivity. The resultsare shown in Figure 2. Children with poor relativeto good inhibition were more sensitive to peers'reinforcement and modeling of deviant discourseand play. Parental discipline reliably moderated therelation of peer modeling and reinforcement ofdeviant discourse and play to rates of children'sdeviant discourse and play (b=.12, pb .05). Simpleslopes were plotted for the relationship of peers'reinforcement and modeling of deviant discourseand play to rates of deviant discourse and play athigh (+1/2 SD) and low (-1/2 SD) levels of poordiscipline, and are shown in Figure 3. Children

FIGURE 3 Simple Slopes of the Relationship of Peer Reinforce-ment and Modeling of Deviant Discourse and Play to Rates ofChild Norm-Violating Discourse and Play at High (+1/2 SD) andLow (-1/2 SD) Levels of Ineffective Discipline.

FIGURE 4 Simple Slopes of the Relationship of Peer Reinforce-ment and Modeling of Deviant Discourse and Play to Rates ofChild Norm-Violating Discourse and Play at High (+1/2 SD) andLow (-1/2 SD) Levels of Peer Rejection.

325p e er dev i ancy tra in ing

experiencing poor relative to better discipline weremore sensitive to peers' reinforcement and model-ing of deviant discourse and play. Peer rejectionreliably moderated the relation of peer modelingand reinforcement of deviant discourse and play torates of children's deviant discourse and play(b=.11, pb .05). Simple slopes were plotted forthe relationship of peers' reinforcement and mod-eling of deviant discourse and play to rates of childdeviant discourse and play at high (+1/2 SD) andlow (-1/2 SD) levels of peer rejection, and are shownin Figure 4. Children experiencing high relative tolow levels of peer rejection were more sensitive topeers reinforcement and modeling of deviantdiscourse and play.

DiscussionPeer deviancy training among school classmatesappears to be a powerful risk factor for multi-setting antisocial behavior as early as age 6, and therisk it poses extends to age 9. The potency ofdeviancy training is also apparent in that it wasconcurrently and prospectively associated withantisocial behavior even after accounting for otherwell established risk factors, including deviant peeraffiliation, peer rejection, poor parental discipline,and child impulsivity. Peer deviancy training mayalso serve as an indirect mechanism by which peerrejection, poor discipline, and child impulsivityincrement risk for early onset and persistingantisocial behavior. In summary, the data in thisreport suggest that modeling and reinforcement ofdeviant discourse and play during peer interactionare associated with the development of verbal rulesand cognitive templates which facilitate the perfor-mance of antisocial behavior.

These findings are consistent with those reportedby other investigators (e.g., Patterson et al., 2000),but extend previous findings in two ways. First,peer deviancy training operates as early as thetransition to elementary school as children aresystematically exposed to peers who have diverseprevious experiences with norm-violating modelsand who display a variety of deviant discourse andbehavior. Second, the risk posed by peer deviancytraining is not restricted to friendships or cliquesamong deviant youth. In this study, incidentalinvolvement in and exposure to norm-violatingdiscourse and behavior and its positive socialconsequences during interaction with randomlyselected same-gender elementary school classmateswas similarly associated with increased risk forantisocial behavior that persisted for nearly 3 years.However, not all children appear to be equally

influenced by peer modeling of deviant discourseand play, and by the excitement, attention andtitillation deviant discourse and play generateamong peers. The degree to which children areinfluenced by peer deviancy training appears to bemoderated by three variables from differentdomains: child impulsivity or poor executiveinhibition; ineffective, coercive and non-contingentdiscipline; and peer rejection. Each of these vari-ables appears to amplify the effects of peer deviancytraining by enhancing children's reactivity andsensitivity to the social reinforcement generatedwhen discourse and play turn to topics of sex,drugs, aggression, swearing and stealing, defianceof authority, and bragging about risky exploits andexperiences (Wills & Dishion, 2004). The explicitdescription and assessment of the social processescomprising peer deviancy training is not new, buttheir role as mechanisms by which moderatorsoperate has not been previously tested.Impulsivity appears to increase children's sensi-

tivity to peer attention and reinforcement of deviantdiscourse and play. Inhibition of pre-potentresponses is a key executive function, enablingeffortful control, problem solving, and emotionaland behavioral self-regulation (Barkley, 2001).Executive inhibitory control facilitates the abilityto avoid events and suppress actions associatedwith immediate reward, and to approach eventsand engage in behaviors associated with immediatecues for punishment (Posner & Rothbart, 2000).The data in this report replicate similar findings onthe role of impulsivity in moderating the associationbetween deviant peer affiliation and antisocialbehavior (Gardner et al., 2008; Goodnight et al.,2006) but extends those findings to a youngersample. While the danger engendered by themismatch between inadequate executive control

326 snyder et al .

and increased exposure to environmental risk inearly adolescence has received considerable empir-ical attention (Spear, 2000), a similar mismatchmay appear at other salient transitions such as entryto elementary school.In a similar vein, it is also not surprising that poor

parental discipline would be associated with poorerself-regulation by children in the absence of adulttracking and contingencies (Vitaro et al., 2000). Aswell as replicating its direct and persistent associ-ation with antisocial behavior during early child-hood (Snyder et al., 2003), poor discipline was alsoassociated with increased rates of child deviantdiscourse and play, and with reduced capacity toresist the direct and vicarious pull of peer rewardsassociated with such discourse and play. Althoughnot tested in this study, poor discipline may alsoundermine the self-regulation children need totolerate social disapproval and disparagementmeted out by peers for signs of embarrassment orwithdrawal when exposed to norm-violating dis-course and behavior.The association of peer rejection with increased

sensitivity to peer modeling and reinforcement ofdeviant discourse and play is a novel finding. Therelative rates with which various social responsesare used by children to relate to peers are shapedby the relative rates at which reinforcement isprovided by peers for those response alternatives(Snyder et al., 2004). Rejected children mayexperience low rates of peer reinforcement fornormative discourse and play (Coie & Kupers-midt, 1983). As a consequence, the social excite-ment and attention generated by deviant discourseand rehearsal may be particularly salient andpowerful. Children who are rejected by normativepeers may also spend more time interacting withantisocial peers, providing an experiential historythat further shapes increased responsiveness topeer modeling and reinforcement of deviantdiscourse and activities.It should also be noted that impulsivity, poor

discipline and peer rejection were each directlylinked with multi-setting child antisocial behaviorfrom ages 5.3 to 6.8 years as well as being indirectlylinked through peer deviancy training. The size ofthe moderated and indirect associations of impul-sivity, poor discipline and peer rejection withantisocial behavior was small relative to their directassociations. However, reliable indirect and mod-erated associations of peer deviancy training withantisocial behavior were still apparent in thecontext of their more direct associations and aftercontrolling for deviant peer affiliation, suggestingthat their indirect and moderated associations arehighly reliable and operate early in development.

This report has several strengths. Each of theconstructs was measured by multiple informantsand methods, and with relatively little measurementoverlap among constructs. This reduces measure-ment error and minimizes source variance as analternate explanation for observed empirical rela-tionships. The relation of child deviant discourseand play to peer reinforcement of deviant discourseand play has been reported previously for thissample (Snyder et al., 2004), but was tested morepowerfully in this report by the addition ofindicators of peer modeling of deviant discourseand play, and of target children's reinforcement ofpeers' deviant discourse and play. The relation ofchild deviant discourse and play to concurrent childantisocial behavior has also been reported previ-ously for this sample (Snyder et al., 2005), but hereis extended to child self reported antisocial behaviornearly 3 years later. Thus, the two crucial con-structs of peer deviancy training in this report weremore powerfully measured to reflect multiple directand vicarious social processes, and the persistingimpact of early peer deviancy training was moreclearly substantiated. The longitudinal designprovided the opportunity to incorporate constructsinto models consistent with the temporal sequenceof their operation.These strengths are balanced by a number of

limitations. Strong causal inferences cannot bedrawn from these correlational data. The modelstested in this report may be incomplete, and fail toincorporate a number of other key variables thatwould reduce potential bias in estimates of thecontribution of deviancy training and associatedmoderating variables to antisocial development.The degree to which the findings derived from thissingle-school community sample generalize tochildren, families, schools and communities with abroader range of characteristics is unknown.However, the sample used to obtain data in thisreport represents a relatively full range of childadjustment and family characteristics.The results have a number of potentially impor-

tant implications. Interventions to reduce theoccurrence of peer deviancy training in schoolsettings as early as kindergarten may powerfullyreduce risk for early-onset, persisting antisocialbehavior. These interventions may focus on reduc-ing the occurrence of deviancy training in a verydirect way by increasing adult monitoring of peerinteraction on playgrounds and in hallways andrestrooms to include deviant discourse, activitiesand play as well as more overt forms of aggressionand disruptive behavior. Interventions could alsodirectly alter peer contingencies for deviant dis-course and play, perhaps piggybacking on existing

327p e er dev i ancy tra in ing

effective interventions for aggressive behavior suchas the good behavior game (Stoolmiller, Eddy, &Reid, 2000). However, this report indicates anumber of intervention tactics and targets may beavailable in addition to simply trying to limitaggregation of at-risk children and peer deviancytraining. Effective parenting is a protective factor,reducing children's sensitivity to peer modeling andreinforcement of deviant discourse and play, andtheir co-participation in such discourse and play.More specifically, the promotion of effectiveparental discipline during early childhood may bea potent means to mitigate future negative peerinfluence. It may also be useful to complement skillstraining and peer relationship enhancement inter-ventions for disruptive and aggressive children (e.g.,Lochman, Wells, & Lenhart, 2008) with explicittraining in resisting or managing peers' deviantdiscourse and modeling. These efforts may beparticularly important for children who exhibitdeficient executive control. Finally, recent efforts todirectly enhance children's executive attentionaland inhibitory control (Rothbart & Rueda, 2005),if successful and generalizable, may provide a usefulavenue to address the problems of peer deviancytraining.

ReferencesAchenbach, T. M. (1991). Integrative guide for the 1991 CBCL

4-18, YSR and TRF profiles. Burlington, Vermont: Univer-sity of Vermont Department of Psychiatry.

Aiken, L., & West, S. G. (1991). Multiple regression: Testingand interpreting interactions. Newbury Park, CA: Sage.

Asher, S. R., Singleton, L. C., Tinsley, B. R., & Hymel, S.(1979). A reliable sociometric measure for preschoolchildren. Developmental Psychology, 15, 443–444.

Barkley, R. A. (2001). The executive functions and selfregulation: An evolutionary neuropsychological perspective.Neuropsychology Review, 11, 1–29.

Coie, J. D., & Kupersmidt, J. B. (1983). A behavioral analysis ofemerging social status in boys' groups. Child Development,61, 1400–1416.

Dishion, T. J., & Dodge, K. A. (2005). Peer contagion ininterventions for children and adolescents: Moving towardsan understanding of the ecology and dynamics of change.Journal of Abnormal Child Psychology, 33, 395–400.

Dishion, T. J., & Dodge, K. A. (2006). Deviant peer contagionin interventions and programs: An ecological framework forunderstanding influence mechanisms. In K. A. Dodge, T. J.Dishion, & J. E. Lansford (Eds.), Deviant peer influence inprograms for youth: problems and solutions (pp. 14–43).New York: Guilford Press.

Dishion, T. J., & McMahon, R. J. (1998). Parental monitoringand the prevention of child and adolescent problembehavior: A conceptual and empirical formulation. ClinicalChild and Family Psychology Review, 1, 61–75.

Dishion, T. J., Nelson, S. E., Winter, C., & Bullock, B. M.(2004). Adolescent friendship as a dynamic system: Entropyand deviance in the etiology and course of male antisocialbehavior. Journal of Abnormal Child Psychology, 32,651–663.

Dishion, T. J., & Owen, L. D. (2002). A longitudinal analysis offriendships and substance use: bidirectional influence fromadolescence to adulthood. Developmental Psychology, 38,480–491.

Dishion, T. J., & Patterson, G. R. (2006). The development andecology of antisocial behavior. In D. Cicchetti, & D. J. Cohen(Eds.), Developmental psychopathology. Vol. 3: Risk,disorder and adaptation (pp. 503–541). New York: Wiley.

Dishion, T. J., Patterson, G. R., & Griesler, P. C. (1994). Peeradaptations in the development of antisocial behavior: Aconfluence model. In L. R. Huesmann (Ed.), Currentperspectives on aggressive behavior (pp. 61–95). NewYork: Plenum.

Dishion, T. J., Piehler, T. F., &Myers, M.W. (2008). Dynamicsand ecology of adolescent peer influence. In M. J. Prinstein,& K. A. Dodge (Eds.), Understanding peer influence inchildren and adolescents (pp. 72–93). New York: Guilford.

Dishion, T. J., Spracklen, K. M., Andrews, D. W., & Patterson,G. R. (1996). Deviancy training in male adolescent friend-ships. Behavior Therapy, 27, 373–390.

Dodge, K. A., Dishion, T. J., & Lansford, J. E. (2006). Deviantpeer influences in programs for youth: Problems andsolutions. New York: Guilford Press.

Drossel, C., Waltz, T. J., & Hayes, S. C. (2007). Anintroduction to principles of behavior. In D. W. Woods, &J. W. Kanter (Eds.), Understanding behavior disorders: Acontemporary behavioral perspective (pp. 21–46). Reno,NV: Context Press.

Duncan, G. J., Boisjoly, J., Kremer, M., Levy, D. M., & Eccles,J. (2005). Peer effects in drug use and sex among collegestudent. Journal of Abnormal Child Psychology, 33,375–385.

Gardner, T. W., Dishion, T. J., & Connell, A. M. (2008).Adolescent self-regulation as resilience: Resistance toantisocial behavior within the deviant peer group. Journalof Abnormal Child Psychology, 36, 273–284.

Gifford-Smith, M., Dodge, K. A., Dishion, T. J., & McCord, J.(2005). Peer influence in children and adolescents: Crossingthe bridge from developmental to intervention science.Journal of Abnormal Child Psychology, 33, 255–265.

Goodnight, J. A., Bates, J. E., Newman, J. P., Dodge, K. A., &Pettit, G. S. (2006). The interactive influences of frienddeviance and reward dominance on the development ofexternalizing behavior during middle adolescence. Journalof Abnormal Child Psychology, 34, 573–583.

Gottman, J. M., McCoy, K., Coan, J., & Collier, H. (1995). TheSpecific Affect Coding System (SPAFF) for observingemotion communication in marital and family interaction.Mahwah, NJ: Erlbaum.

Hanish, L. D., Martin, C. L., Fabes, R. A., Leonard, S., &Herzog, M. (2005). Exposure to externalizing peers in earlychildhood: Homophily and peer contagion processes.Journal of Abnormal Child Psychology, 33, 267–281.

Kellam, S., Ling, X., Merisca, R., Brown, H., & Ialongo, N.(1998). The effect of the level of aggression in the first gradeclassroom on the course and malleability of aggressivebehavior into middle school. Development and Psychopa-thology, 10, 165–185.

Kindlon, D., Mezzacappa, E., & Earls, F. (1995). Psychometricproperties of impulsivity measures: Temporal stability,validity, and factor structure. Journal of Child Psychologyand Psychiatry, 34, 645–661.

Kline, R. B. (2005). Principles and practice of structuralequation modeling. New York: Guilford.

Kupersmidt, J., & Dodge, K. A. (2004). Children's peerrelations: From development to intervention. Washington,DC: American Psychological Association.

328 snyder et al .

Laird, R. D., Criss, M. M., Pettit, G. S., Dodge, K. A., & Bates,J. E. (2008). Parents' monitoring knowledge attenuates thelink between antisocial friends and adolescent delinquentbehavior. Journal of Abnormal Child Psychology, 36,299–310.

Lavallee, K. L., Bierman, K. L., & Nix, R. L.Conduct ProblemsPrevention Research Group. (2005). The impact of firstgrade “friendship group” experiences on child socialoutcomes in the Fast Track Program. Journal of AbnormalChild Psychology, 33, 307–324.

Lipsey, M. W. (2006). The effects of community-based grouptreatment for delinquency: A meta-analytic search for cross-study generalizations. In K. A. Dodge, T. J. Dishion, & J. E.Lansford (Eds.), Deviant peer influences in programs foryouth (pp. 162–184). New York: Guilford Press.

Lochman, J. E., Wells, K. C., & Lenhart, L. A. (2008). Copingpower: Child group program. New York: Oxford UniversityPress.

Luthar, S. S., & Becker, B. E. (2002). Privileged but pressured?A study of affluent youth. Child Development, 73,1593–1610.

Nigg, J. T., & Huang-Pollock, C. L. (2003). An early-onsetmodel of the role of executive functions and intelligence inconduct disorder/delinquency. In B. B. Lahey, T. E.Moffitt, & A. Caspi (Eds.), Causes of conduct disorderand juvenile delinquency (pp. 227–253). New York:Guilford Press.

Oeser, J., & Schrepferman, L. (2002). Antisocial Content Code.Unpublished manuscript, Wichita State University, Depart-ment of Psychology.

Patterson, G. R., Dishion, T. J., & Yoerger, K. (2000).Adolescent growth in new forms of problem behavior:Macro- and micro-peer dynamics. Prevention Science, 1,3–13.

Posner,M. I.,&Rothbart,M.K. (2000).Developingmechanismsof self regulation. Development and Psychopathology, 12,247–441.

Poulin, F., Dishion, T. J., & Burraston, B. (2001). Three yeariatrogenic effects associated with aggregating high-riskadolescents in cognitive-behavioral preventive interventions.Applied Developmental Science, 5, 214–224.

Rothbart, M., & Rueda, M. R. (2005). The development ofeffortful control. In U.Mayr, &W. Keele (Eds.),Developingindividuality in the human brain: A tribute to M. Posner(pp. 167–188). Washington, DC: American PsychologicalAssociation Press.

Snyder, J., Cramer, A., & Patterson, G. R. (2005). Thecontributions of ineffective discipline and parental hostileattributions of child misbehavior to the development ofconduct problems at home and school. DevelopmentalPsychology, 4, 1–12.

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

Snyder, J., Reid, J., & Patterson, G. R. (2003). A social learningmodel of child and adolescent antisocial behavior. In B. B.Lahey, T. E. Moffitt, & A. Caspi (Eds.), Causes of conduct

disorder and juvenile delinquency (pp. 27–48). New York:Guilford Press.

Snyder, J., Schrepferman, L., Oeser, J., Patterson, G. R.,Stoolmiller, M., Johnson, K., et al. (2005). Deviancy trainingand association with deviant peers in young children:Occurrence and contribution to early onset conductproblems. Development and Psychopathology, 17,397–413.

Snyder, J., Stoolmiller, M., Patterson, G. R., Schrepferman, L.,Oeser, J., Johnson, K., & Soetaert, D. (2004). Theapplication of response allocation matching to understand-ing risk mechanisms in development: The case of youngchildren's deviant talk and play, and risk for early onsetantisocial behavior. Behavior Analyst Today, 4, 335–344.

Spear, L. (2000). The adolescent brain and age-relatedbehavioral manifestations. Neuroscience and BiobehaviorReview, 24, 417–463.

Stoolmiller, M., Eddy, J. M., & Reid, J. B. (2000). Detectingand describing preventive intervention effects in a universalschool-based randomized trial targeting delinquent andviolent behavior. Journal of Consulting and ClinicalPsychology, 68, 296–306.

Stubbs, J., Crosby, L., Forgatch, M., & Capaldi, D. (1998).Family and peer process code: A synthesis of Oregon SocialLearning Center behavior codes (Training Manual). Avail-able from the Oregon Social Learning Center, 160 East 4th,Eugene, OR 97401.

Vitaro, F., Bredgen, M., & Tremblay, R. E. (2000). Influence ofdeviant friends on delinquency: Searching for moderatorvariables. Journal of Abnormal Child Psychology, 28,313–325.

Walker, H., & Severson, H. (1991). Systematic screening forbehavior disorders (SSBD): User's Guide. Sopris West:Longmont, CO.

Wechsler, D. (1991). Wechsler Intelligence Scale for Children(3rd ed.). San Antonio, TX: The Psychological Corporation.

Weiss, B., Dodge, K. A., Bates, J. E., & Pettit, G. (1992). Someconsequences of early harsh discipline and a maladaptivesocial information processing style. Child Development, 63,1321–1335.

Wills, T. A., & Cleary, S. D. (1999). Peer and adolescentsubstance use among 6th-9th graders: Latent growth curveanalyses of influence versus selection mechanisms. HealthPsychology, 18, 453–463.

Wills, T. A., & Dishion, T. J. (2004). Temperament andadolescent substance use: A transactional analysis ofemerging self control. Journal of Clinical child andAdolescent Psychology, 33, 69–81.

Worthke, W. (2000). Longitudinal and multi-group modelingwith missing data. In T. D. Little, U. Shnabel, & J. Baumert(Eds.), Modeling longitudinal and multi-level data: Practicalissues, applied approaches, and specific examples(pp. 219–240). Erlbaum: Mahwah, NJ.

RECEIVED: January 21, 2009ACCEPTED: May 29, 2009Available online 4 February 2010