Results and Discussion
Our goals were to replicate the findings from the Social Development Project regarding the direct role of early social rejection in relation to later aggression and to extend these findings by examining sex differences, differences between reactive and proactive aggression, and the possible moderator roles of early social rejection in later behavior problems. The results that follow report Grade 3 TRF Aggression outcomes. Analyses using Grade 3 reactive aggression and proactive aggression scores as separate outcomes were highly consistent with analyses for TRF Aggression. The few exceptions are integrated when the corresponding results for TRF Aggression are described.
First, we found robust negative correlations between social preference scores in kindergarten (age 5) and in Grades 1, 2, and 3 (ages 6, 7, and 8) and TRF Aggression scale scores in Grade 3 (age 8), r(484) = −.36, r(430) = −.39, r(450) = −.33, and r(430) = −.42 for kindergarten, first, second, and third grades (ages 5, 6, 7, and 8), respectively, all ps<.001. These correlations did not vary significantly across sites, cohorts, sexes, and racial groups.
ANOVA with rejection status (no or yes), gender, site, and cohort as factors revealed that children who were socially rejected in kindergarten (age 5) had mean TRF Aggression scores in Grade 3 (age 8) that were 2.8 times higher than those of the nonrejected group, F(1,482) = 53.24, p<.001. Nearly identical findings held for rejected groups identified in first grade (age 6), F(1, 437) = 50.69; second grade (age 7), F(1, 469) = 30.89; and third grade (age 8), F(1, 437) = 94.94, all ps<.001. A significant main effect of gender on TRF Aggression revealed that teachers reported higher levels of aggression for boys than for girls, but gender did not interact significantly with social rejection in the prediction of aggression.
Next, we computed partial correlations, in which third grade (age 8) TRF Aggression scores were predicted from earlier social preference scores after partialling out the effects of all previous TRF Aggression scores. These partial correlations were robust, r(477) = −.21, r(403) = − .17, and r(420) = − .15 for kindergarten, first, and second grades (ages 5, 6, and 7) social preference scores, respectively, all ps<.01. The same findings held when we used ANCOVA with the rejected and nonrejected groups, controlling for prior aggression scores, at kindergarten (age 5), F(1, 477) = 13.01; at first grade (age 6), F(1, 431) = 20.81; and at second grade (age 7), F(1, 439) = 7.38, all ps<.01 or better. Again, no significant interactions between gender and rejection were found. The evidence to this point replicates our previous findings supporting the direct role of social rejection in later TRF Aggression.
When we partialled out TRF school performance, r(460) = − .33, p<.001, and TRF internalizing, r(477) = − .36, p<.001, scores collected from kindergarten teachers, we still found a direct effect of early rejection on Grade 3 (age 8) TRF Aggression. The effect also held when we partialled out aggressive behavior nomination scores collected from kindergarten (age 5) peers, r(481) = − .14, p<.01. When reactive and proactive aggression were examined separately, however, after partialling out aggressive behavior nominations provided by kindergarten peers, the correlation between kindergarten (age 5) rejection and Grade 3 (age 8) proactive aggression was no longer significant, r(483) = − .06, although the correlation between kindergarten (age 5) rejection and Grade 3 (age 8) reactive aggression remained significant, r(483) = − .16, p<.001. Thus, early peer rejection appears to exacerbate growth in reactive rather than proactive aggressive behavior. Furthermore, we could not statistically explain away the incremental effect of kindergarten (age 5) social rejection beyond other early factors in predicting Grade 3 (age 8) aggression, and the findings held for both boys and girls.
Next, we replicated the path analyses evaluating the incremental role of social preference scores at each grade level in this prediction (see ), including Gender × Kindergarten Aggression and Gender × Social Preference interactions for each year. Of the variance in Grade 3 (age 8) aggression, 34% was accounted for by kindergarten (age 5) aggression and the child’s social preference history. None of the interactions involving gender was significant. Social preference scores in each successive year made a significant incremental contribution in predicting later TRF Aggression, beyond the prediction afforded by knowledge of early aggression and previous preference scores. When these analyses were conducted using Grade 3 (age 8) reactive and proactive aggression outcomes, the Grade 2 (age 7) social preference score did not significantly increment the prediction of Grade 3 (age 8) proactive aggression after taking into account kindergarten and Grade 1 social preference scores (β = .06, ns); however, Grade 2 (age 7) social preference did significantly increment this prediction for Grade 3 (age 8) reactive aggression (β = −.20, p<.01). In combination with the partial correlation results, these findings suggest that the link between social rejection and growth in aggression is stronger for reactive than for proactive aggression.
Figure 2 Child Development Project prediction of Grade 3 TRF Aggression from kindergarten aggression and previous social preference scores. First numbers shown are standardized path coefficients controlling for all earlier factors. Numbers after the slash control (more ...)
ANOVA was conducted on Grade 3 (age 8) TRF Aggression scores as a function of a grouping variable that classified children according to the number of years they experienced rejection, with gender, race, cohort, and site as other factors. Mean Grade 3 (age 8) TRF Aggression scores for groups of children who met criteria for rejection in 0, 1, 2, or 3 years of elementary school before Grade 3 (age 8) were 4.00, 9.19, 14.66, and 19.43, respectively. These differences were highly significant, F(3, 447) = 26.97, p<.001. Children who had been rejected continuously for 3 years received Grade 3 aggression scores that were 5 times greater than those of children who had never been rejected. Cell contrasts (p<.01) revealed that 1 year of rejection led to worse outcomes than 0 years, and 2 years was worse than 1 year. Two years and 3 years did not differ. In these analyses, neither the main effect of gender nor the interaction between gender and the number of years of rejection was significant in the prediction of Grade 3 (age 8) TRF Aggression.
At this point, the replication supports the hypothesis that the experience of rejection plays a contributing role, beyond that of a mere marker variable, in the development of later behavior problems. We next examined whether this effect is moderated by another factor. If social rejection acts as a general stressor to exacerbate a child’s maladjustment, it is plausible that the effect of rejection on later aggressive behavior would be greatest for children who were initially inclined toward aggression. We asked the question: Is the effect of social rejection equally strong for children initially inclined toward aggression and children initially inclined away from aggression? We used a median split of the sample based on the kindergarten TRF Aggression scores to generate four groups of children: nonrejected-nonaggressive, rejected-nonaggressive, nonrejected-aggressive, and rejected-aggressive. This classification resembles groupings that are increasingly standard in the field (see Bierman et al., 1993
; Bierman & Wargo, 1995
; Cillessen, Van Ijzendoorn, Van Lieshout, & Hartup, 1992
We then conducted an ANOVA on Grade 3 (age 8) TRF Aggression scores, with kindergarten (age 5) rejected status and kindergarten aggression status as factors, along with gender, race, cohort, and site. We found three significant and unique effects that held across gender, race, cohort, and site. We found strong main effects of early social rejection, F(1, 473) = 10.52, p<.001, and early aggression status, F(1, 473) = 40.94, p<.001. As hypothesized, there was also a significant interaction of rejection and early aggression, F(1, 473) = 6.03, p<.05. The effect of rejection was much stronger among those children who were initially inclined toward aggressive behavior (i.e., above the kindergarten median for aggression) than among those who were inclined initially away from aggressive behavior.
When Grade 3 (age 8) reactive aggression was analyzed as the outcome, the main effect of kindergarten (age 5) rejected status, F(1, 467) = 19.92, p<.001, and kindergarten (age 5) reactive aggression, F(1, 467) = 14.58, p<.001, was each significant, but the interaction between rejection and kindergarten reactive aggression was not significant, F(1, 467) = .82, ns. With Grade 3 (age 8) proactive aggression as the outcome, the main effect of kindergarten (age 5) rejected status was significant, F(1, 469) = 6.80, p<.01, but the main effect of kindergarten (age 5) proactive aggression was not significant, F(1, 469) = 1.66, ns. The interaction between kindergarten (age 5) rejection and kindergarten (age 5) proactive aggression was significant, F(1, 469) = 4.04, p<.05, indicating that the effect of early peer rejection held for proactively aggressive kindergarteners but not for children who were not disposed toward aggression in kindergarten.
The next analysis examined the cumulative effects of chronic rejection as a function of early aggressive disposition. An ANOVA with Grade 3 (age 8) TRF Aggression as the dependent variable and cumulative rejection (0, 1, 2, or 3 years) and kindergarten (age 5) aggression status (above or below the median) as independent variables (along with site, cohort, gender, and race) revealed a main effect of rejection, F(3, 464) = 7.55, p<.001, a main effect of kindergarten aggression status, F(1, 464) = 27.13, p<.001, a significant Gender × Kindergarten (age 5) aggression interaction effect, F(1, 464) = 7.23, p<.01, and a Rejection × Kindergarten aggression status interaction effect, F(3, 464) = 3.79, p<.01. The Gender × Aggression effect indicated that Grade 3 (age 8) aggression scores differed across high and low kindergarten (age 5) aggression groups to a greater degree for boys than girls.
shows that even cumulative rejection does not increase aggressive behavior in children who were not initially inclined toward aggression. Graphed here are the Grade 3 (age 8) TRF Aggression scores by the number of years of rejection that have been experienced by a child, with separate graphs for children initially below the median in kindergarten (age 5) aggression and for children above the median. As shown, the effects of rejection are dramatic and cumulative for children initially above the median in aggression, but these effects are minimal for children initially below the median in aggression. There were no children who were initially below the median on aggression who were rejected for all 3 years. This figure shows that among children who had been initially inclined toward aggressive behavior, if the peer group is accepting of the child continuously across the early elementary school years, that child’s later aggression will not escalate and will remain in an average range.
Child Development Project mean Grade 3 TRF Aggression scores by number of years of social rejection and level of early aggression.
We next examined reactive and proactive aggression outcomes. For Grade 3 (age 8) reactive aggression, significant main effects were found for both cumulative rejection, F(3, 460) = 7.51, p<.001, and kindergarten (age 5) reactive aggression, F(1, 460) = 9.92, p<.01. The cumulative Rejection × Kindergarten (age 5) reactive aggression interaction was not significant, F(3, 460) = 1.06, ns. In the prediction of Grade 3 (age 8) proactive aggression, the main effect of cumulative rejection, F(3, 462) = 8.15, p<.001, was significant, but there was not a significant main effect of kindergarten (age 5) proactive aggression, F(1, 462) = .78, ns, or a significant cumulative Rejection × Kindergarten (age 5) proactive aggression interaction, F(3, 462) = .98, ns.
Although the analyses of TRF Aggression suggest that the group of children who were initially nonaggressive do not become more aggressive following peer rejection, it is still possible that they might suffer other adverse reactions (e.g., growth in social withdrawal). The Grade 3 (age 8) TRF Withdrawal scale was subjected to the same ANOVA, with early rejection status, early aggression status, gender, race, cohort, and site as independent variables. All tests of main effects and interaction effects involving rejection yielded nonsignificant Fs less than 1.
These analyses raise the question of whether peer rejection moderates the effects of other early behaviors on later outcomes. It is plausible that peer rejection will exacerbate any problem for which a child is initially predisposed or at risk. The work of Rubin and colleagues (Rubin et al., 1991
; Rubin & Mills, 1988
) suggests that socially withdrawn children might react to peer rejection with increased social withdrawal or isolation (rather than aggression). This was found to be the effect of rejection on children’s behavior in contrived play groups (Dodge, 1983
To test this possibility, we used a median split on kindergarten (age 5) TRF Social Withdrawal scores to create a group of children initially predisposed toward withdrawal and a group not so predisposed. ANOVA on Grade 3 (age 8) TRF Social Withdrawal scores, with kindergarten (age 5) rejected status and kindergarten (age 5) withdrawal status as factors, along with gender, race, cohort, and site revealed a significant main effect of kindergarten (age 5) social withdrawal status on Grade 3 (age 8) TRF Social Withdrawal scores, F
(1, 473) = 17.26, p
<.001, and a significant main effect of gender, F
(1, 473) = 5.19, p
<.05. However, there was no significant main effect of kindergarten rejection, F
(1, 473) = 3.03, ns
, or rejection by early withdrawal interaction, F
(1, 473) < 1, ns
, on Grade 3 (age 8) TRF Social Withdrawal scores. Panak and Garber (1992)
have conducted additional work regarding effects of social rejection on internalizing outcomes. It remains possible that social rejection leads to other, unmeasured negative outcomes for nonaggressive children. The rest of our focus is on the effects of social rejection on later aggressive behavior.
These findings provide support for the hypothesis that the experience of being rejected by peers in early elementary school plays an incremental role in the later development of aggressive behavior problems, beyond the contributions made by factors leading to peer rejection in the first place. The causal role of rejection cannot be proven by correlational data, of course, because still other variables (e.g., physical attractiveness, verbal ability) might lead both to rejection and to growth in aggressive behavior.
In addition, these findings support the hypothesis that this effect is moderated by a child’s disposition toward aggression. Thus, it does not appear that peer rejection serves merely as an early marker of later behavior problems. Presuming that these direct and moderating effects are real, it makes sense to turn to our next question: What is the process through which the experience of peer rejection has an impact on later antisocial development?