Temperament alters susceptibility to negative peer influence in early adolescence

Abstract

The role of deviant peers in adolescent antisocial behavior has been well documented, but less is known about individual differences in susceptibility to negative peer influence. This study examined whether specific temperament dimensions moderate the prospective relationship between peer deviance and delinquent behavior in early adolescence. Participants included 704 adolescents recruited from the community. At baseline, parents provided information on adolescents’ temperament and youth reported on their own and their friends’ delinquent behavior. Self-reports of adolescents’ delinquent behavior were collected again 16 months later. Peer deviance was related to delinquent behavior over time more strongly for adolescents with low levels of task orientation, flexibility, and positive mood, compared to youth with high levels of task orientation, flexibility, and positive mood. Analyses of gender differences indicated that low flexibility increased susceptibility to negative peer influence only for males, but not females. General activity level and sleep rhythmicity did not moderate the effect of peer behavior on delinquency.

The role of peers in the development of externalizing behavior problems in youth has been well documented. Affiliation with deviant peers increases the risk of subsequent aggressive and disruptive behavior in childhood (Berndt & Keefe, 1995; Keenan, Loeber, Zhang, Stouthamer-Loeber, & Van Kammen, 1995; Mrug, Hoza, & Bukowski, 2004), as well as delinquency and substance use in adolescence (Monahan, Steinberg, & Cauffman, 2009; Selfhout, Branje, & Meeus, 2008; Wills & Cleary, 1999). In contrast, less is known about factors that moderate individuals’ susceptibility to peer influence. Nevertheless, a few studies have identified multiple individual and social environment characteristics that alter the prospective relationship between peer deviance and externalizing behaviors. Specifically, a history of antisocial behavior and poor parenting amplify the negative effect of deviant peers on adolescent antisocial behavior (Galambos, Barker, & Almeida, 2003; Mrug & Windle, 2009; Vitaro, Brendgen, & Tremblay, 2000). Another key child characteristic that has been proposed to alter susceptibility to peer influences is temperament (e.g., Wills & Dishion, 2004). However, few empirical studies have examined this issue. Also, existing research has focused primarily on substance use, so little is known about the role of temperament in the development of adolescent delinquency. Thus, the main goal of this study was to identify specific temperamental dimensions that moderate the relationship between deviant peer affiliations and subsequent delinquent behavior in early adolescence, a developmental period of high susceptibility to negative peer influence (Steinberg & Monahan, 2007).

Temperament refers to individual differences in behavior, reactivity, and self-regulation that are biologically-based, emerge early in life, and are stable over time (Rothbart & Bates, 2006). Although specific dimensions of temperament and their definitions differ across studies, several conceptually and empirically distinct aspects of temperament have been frequently addressed in research. These dimensions include activity level, negative emotionality, effortful control of attention, regulation of emotions, rhythmicity, and social approach vs. inhibition (Sanson, Hemphill, & Smart, 2004). Many of these dimensions have been related to adolescent externalizing behaviors. Low flexibility, poor attention regulation, and high activity level were associated with delinquency (Windle, 1992), and negative affect and activity level were related to substance use (Molina, Chassin, & Curran, 1994; Wills & Cleary, 1999). Similarly, multiple aspects of temperament distinguished adolescent drug abusers from their peers, including high activity level and negative affect, low attention control and flexibility, and arrythmicity (Tarter, Laird, Kabene, Bukstein, & Kaminer, 1990). In addition, adolescents with a greater number of difficult temperament dimensions reported higher levels of delinquency and more frequent substance use (Windle, 1991). Longitudinal studies have extended these cross-sectional relationships over time, linking poor attention regulation, high activity level, negative affect, and social inhibition in childhood with aggressive and antisocial behaviors and substance use in adolescence (Caspi, Henry, McGee, Moffitt, & Silva, 1995; Caspi, Moffitt, Newman, & Silva, 1996; Lerner & Vicary, 1984; van der Laan, Veenstra, Bogaerts, Verhulst, & Ormel, 2010; Vitaro, Brendgen, & Tremblay, 2002).

Theoretical conceptualizations agree that temperament translates into behavior problems not directly, but indirectly through transactions with the social environment (e.g., Bates, Bayles, Bennett, Ridge, & Brown, 1991; Keenan & Shaw, 1997). For instance, it has been hypothesized that difficult temperament contributes to poor self-control, which in turn leads to lower quality of parent-child relationships, more negative life events, lower academic, cognitive, and social competencies, and association with deviant peers (Wills & Dishion, 2004). These factors interact with one another and provide a more proximal foundation for the development of antisocial behavior and substance use in adolescence. In addition to these mediating processes, it has been proposed that difficult temperament and deficient self-control contribute to poor outcomes through increasing youths’ vulnerability to negative social influences and risk factors, including deviant peer influences (Wills & Dishion, 2004; Wills, Sandy, & Yaeger, 2000). Focusing on the joint roles of temperament and peers in the development of externalizing behaviors, theoretical models thus specify two types of relationships. First, affiliation with deviant peers serves as one pathway linking specific aspects of temperament with behavioral outcomes (mediating relationship). Second, difficult temperament and deficits in self-control increase adolescents’ vulnerability to social influences from deviant friends (moderating relationship).

A number of investigations have provided support for deviant peer affiliations mediating the links between temperament and adolescent substance use (e.g., Creemers et al., 2010; Giancola & Parker, 2001; Wills & Cleary, 1999; Wills, Windle, & Cleary, 1998). By contrast, only two studies examined temperament as a moderator of that relationship. One study focused on substance use, finding that protective temperament (a combination of attention control and positive emotionality) attenuated the effect of peer substance use on growth in early adolescents’ substance use (Wills, Sandy, Yaeger, & Shinar, 2001). Another study found that self-regulation (a complex construct involving attention control, flexibility, and behavioral inhibition) reduced the link between peer deviance and subsequent delinquency among older adolescents (Gardner, Dishion, & Connell, 2008). These studies suggest that adolescents with more difficult temperament are not only more likely to affiliate with deviant peers, but are also more susceptible to these peers’ negative social influences. However, because these studies examined combinations of a limited set of temperament dimensions, it is not clear which specific aspects of temperament are relevant for adolescents’ susceptibility to negative peer influence.

Another important issue that has been entirely omitted from the existing literature on temperament and peer influences is the possibility of gender differences. Because peers and adults may respond differently to the same temperament trait when expressed by boys vs. girls, specific temperament dimensions may present greater liability for one gender compared with the other (Prior, 1992). Indeed, empirical evidence confirms that temperament relates to social outcomes differently for boys and girls (e.g., Williams et al., 2010). In particular, task orientation and flexibility are more strongly predictive of social skills for boys compared with girls (Eisenberg et al., 1993; Sanson et al., 2004), so it is possible that these characteristics attenuate susceptibility to negative peer influences in boys more so than they do in girls.

Thus, the primary aim of this study was to determine if (and which) specific temperament dimensions moderate the link between peer deviance and adolescents’ delinquent behavior. A secondary goal was to examine gender differences in these processes. Based on previous research, we hypothesized that poor task orientation and low flexibility would be associated with stronger negative peer influence. Consistent with previous findings of greater importance of task orientation for boys’ than girls’ problem behavior (Sanson et al., 2004), we hypothesized that task orientation and possibly other temperament dimensions would increase negative peer influence effects to a greater degree in males than females. To provide a stronger basis for causal inferences, we investigate the prospective relationship between peer deviance and subsequent delinquency, while controlling for previous levels of delinquent behavior. To reduce problems with shared-method variance, information about adolescents’ temperament was collected from parents, while data on peer deviance and delinquent behavior were based on adolescent self-report.

Method

Participants and Procedures

This study included 704 early adolescents (52% male, 76% African-American, 22% Caucasian, 2% other) and their primary caregivers who participated in the Birmingham Youth Violence Study, conducted in Birmingham, Alabama, between 2003 and 2005. Students were initially recruited from 5^th grade classrooms in 17 Birmingham area schools selected through a two-stage probability sampling process. In the first stage, schools were randomly selected based on probabilities designed to achieve a sample that would be representative of all students attending public schools in the Birmingham metropolitan area in terms of racial/ethnic, gender, and socioeconomic composition. In the second stage, all individual students at selected schools were invited to participate. As a result of this sampling procedure, the demographic make-up of the sample closely resembled the demographic composition of the Birmingham metropolitan area (74% African American, 24% Caucasian). The study was approved by the Institutional Review Board of the university, and all families provided parental consent and child assent. Caregivers and adolescents were interviewed separately by trained interviewers in private spaces, either at the university or in the family home. Sensitive questions (e.g., about delinquency) were answered privately through computer-assisted technology. All participants were compensated financially for their time. A total of 704 families completed individual interviews at Wave 1 (42% participation rate); 603 families returned for Wave 2 interviews, on average 16 months later (86% retention rate). The retained sample did not differ from those lost through attrition in age and gender, but included a slightly higher proportion of African Americans, families with lower income, and adolescents with higher levels of delinquency, higher general activity level, and lower flexibility (small effect sizes). There were no differences in other temperament characteristics or peer deviance. The average age of youth participants was 11.8 (SD=.8) at Wave 1 and 13.2 years (SD=.9) at Wave 2. The sample was socioeconomically heterogeneous, with family income ranging from below $5,000 to over $90,000 (median $25,000–$30,000) at both time points.

Measures

Temperament

At Wave 1, parents responded to five subscales of the Revised Dimensions of Temperament Survey (DOTS-R; Windle & Lerner, 1986). Item were rated on a 4-point scale (1 - ‘usually false’, 4 - ‘usually true’) and summed. Five temperament dimensions were assessed: 1) general activity level (7 items; e.g., “My child moves around a lot” α =.88); higher scores indicate high levels of energy and activity; 2) flexibility-rigidity (5 items; e.g. “My child resists changes in routine”; α=.78); higher scores indicate greater flexibility in responding to external changes; 3) mood quality (7 items; e.g. “My child smiles often”; α=.78); high scores indicate more positive affect; 4) rhythmicity - sleep (6 items; e.g. “My child wakes up at different times”; α=.74); high scores indicate highly regular sleep-wake cycle; and 5) task orientation (8 items; e.g., “My child persists at a task until it is finished”; α=.81); high scores indicate greater persistence and low distractibility.

Peer deviance

At Wave 1, adolescents responded to six items asking how many of their friends engaged in delinquent behavior and substance use. Specifically, these questions asked about friends’ damaging property, hitting or threatening to hit someone, getting into physical fights, carrying a knife or a gun, using alcohol, and using marijuana or other illegal drugs. These items were rated on a 5-point scale (1 - ‘none of them’, 5 - ‘all of them’) and summed (α =.73).

Delinquency

At Wave 1, adolescents were asked how often they engaged in 8 delinquent acts in the last 12 months. The delinquency items (Elliott, Huizinga, & Ageton, 1985) included questions on fighting, running away from home, truancy, stealing, and breaking into a vacant home or building. Responses were rated on a 3-point scale but were recoded to a dichotomous format (0 – no engagement, 1 – one or more times) to improve the variable’s distribution, and summed (α =.56). At Wave 2, the delinquency scale was expanded to appropriately measure a broader range of delinquency expected in the older sample. One fighting question was removed and 20 questions were added, for a total of 27 items. The additional items came from the same source as the Wave 1 items (Elliott et al., 1985) and inquired about theft, destruction of property, fighting and assaults, selling drugs, public disorder, and robbery. For each question, adolescents were asked to estimate the number of times they engaged in that behavior in the last 12 months. As with the Wave 1 data, the answers were dichotomized and summed (α =.80).

Demographics

Adolescent age in years, gender, ethnicity, and family income at Wave 1 were included as demographic covariates. Gender was coded 0 for boys and 1 for girls; race/ethnicity was coded 0 for Caucasian and 1 for African American or other. Family income was assessed using 13 categories (1 – <$5,000, 13 – >$90,000 ). Wave 2 income data were substituted when Wave 1 income was missing.

Results

Descriptive Statistics and Bivariate Correlations

Descriptive statistics and zero-order correlations among all variables appear in Table 1. All but one temperament dimension (positive mood) were associated with delinquency at Wave 2. Higher delinquency levels were also reported by males, racial/ethnic minorities, lower income adolescents, and those with more deviant friends. Peer deviancy was related to lower flexibility and higher activity level. Males reported higher levels of peer deviance and were described by parents as less flexible, less task oriented, more active, and having less positive affect than females, but the magnitude of these differences was small.

Table 1.

Descriptive Statistics and Bivariate Correlations for All Variables

	Mean (SD)	1.	2.	3.	4.	5.	6.	7.	8.	9.	10.	11.
1. Delinquency, W2	2.34 (2.79)	1.00
2. Delinquency, W1	1.15 (1.15)	.45*	1.00
3. Age, W1	11.76 (.76)	.23*	.19*	1.00
4. Female gender, N (%)	335 (48)	−.11*	−.13*	−.07	1.00
5. Ethnic minority, N (%)	548 (78)	.16*	.17*	.21*	.10*	1.00
6. Family income, W1	6.50 (3.81)	−.14*	−.22*	−.27*	−.10*	−.37*	1.00
7. Peer deviance, W1	9.42 (3.44)	.43*	.56*	.23*	−.11*	.20*	−.20*	1.00
8. Flexibility, W1	15.77 (3.67)	−.11*	−.14*	−.17*	.10*	−.05	.19*	−.11*	1.00
9. Positive mood, W1	25.78 (3.07)	−.06	−.08*	−.11*	.10*	−.02	.12*	−.06	.26*	1.00
10. Task orientation, W1	19.69 (5.73)	−.10*	−.11*	−.05	.07*	.09*	−.01	−.04	.16*	.05	1.00
11. Activity level, W1	18.34 (6.24)	.10*	.18*	.07	−.13*	.06	−.13*	.10*	−.43*	−.03	−.34*	1.00
12. Rhythm sleep, W1	16.71 (4.49)	−.10*	−.07	−.07	−.03	−.08*	.12*	−.03	.07	.13*	.15*	−.12*

Note: W1 – Wave 1; W2 – Wave 2;

^*p<.05 or lower

Temperament and Peer Deviance Predicting Delinquency

The primary analyses included hierarchical multiple regressions predicting Wave 2 delinquency. All control variables (delinquency at Wave 1, age, gender, ethnicity, and family income), peer deviance, and the five temperament dimensions were entered in step 1, and interactions between peer deviance and each temperament dimension in step 2. Peer deviance and the temperament variables were centered prior to computing the interaction terms, and each interaction was examined in a separate analysis to prevent multicollinearity problems (Aiken & West, 1991). To interpret significant interactions, simple slopes for peer deviance were computed at high and low levels of temperament (1 SD above and below the mean). All analyses were conducted in Mplus utilizing the full-information-maximum-likelihood (FIML) method for handling missing data. This method estimates model parameters and standard errors using all available information, and is superior to listwise deletion and commonly used missing data imputation methods (Enders, 2001). However, the amount of missing data was small. Altogether, 594 cases (84% of 704) had complete data on all variables, but only 1.5% of all data points (i.e., each variable for each participant) were missing. In addition to the predictive paths, all models estimated covariance parameters among all predictors. As a result, all models were saturated (just-identified, with df=0 and χ²=0) and thus had perfect fit.

Regression coefficients and incremental R² values for each step are listed in Table 2. At the first step, delinquency was uniquely predicted by older age, previous levels of delinquency, and peer deviance. After controlling for these and other variables, none of the temperament dimensions was uniquely associated with delinquency. At the second step, flexibility, positive mood, and task orientation moderated the effect of peer deviance on subsequent delinquency. Decomposition of these interactions indicated that affiliation with deviant peers predicted higher levels of delinquency more strongly at low levels of flexibility (β =.30, p<.001), task orientation (β=.31, p<.001), and positive affect (β =.30, p<.001), compared with high levels of flexibility (β=.15, p<.01), task orientation (β =.11, p<.05), and positive affect (β =.13, p<.01) (see Figure 1).

Table 2.

Hierarchical Linear Regressions Predicting Delinquency at Wave 2

	Predictor	β	ΔR2
Step 1	Age at W1	.11**	.28***
	Ethnic minority	.06
	Female gender	−.05
	Family income at W1	.01
	Delinquency at W1	.27***
	Peer deviance at W1	.24**
	Flexibility at W1	−.02
	Positive mood at W1	.00
	Task orientation at W1	−.05
	Activity level at W1	−.02
	Rhythm sleep at W1	−.05
Step 2a	Friends deviance x Flexibility	−.09*	.01*
Step 2b	Friends deviance x Positive mood	−.07*	.01*
Step 2c	Friends deviance x Task orientation	−.09**	.01*
Step 2d	Friends deviance x Activity level	.05	.00
Step 2e	Friends deviance x Rhythm sleep	−.01	.00

Note: W1 – Wave 1; W2 – Wave 2;

+ p<.10;

^*p<.05;

^**p<.01;

^***p<.001

Figure 1.

Flexibility, mood, and task orientation moderate the effect of peer deviance on delinquency.

Gender Differences in the Moderating Effects

Gender differences in the moderating effects were tested with 3-way interactions between gender, peer deviance, and each temperament dimension added to the regression models. All main effects and 2-way interactions of these variables were included in each model. Separate analyses were conducted for each temperament dimension and all predictors were centered prior to computing the interaction terms. Only the 3-way interactions with flexibility reached statistical significance (β =−.12, p<.01). This interaction is plotted in Figure 2. All other 3-way interactions were non-significant (p>.10). Of the 2-way interactions included in these models, only the interaction between peer deviance and task orientation was significant (β=−.08, p<.05).

Figure 2.

Interaction between peer deviance, flexibility, and gender.

To interpret the significant 3-way interaction, we computed simple slopes for peer deviance at high vs. low levels of flexibility (1 SD above and below the mean) for males vs. females using procedures developed by Preacher, Curran, and Bauer (2006). Peer deviance was most strongly related to delinquency for males with low levels of flexibility (b=.31, p<.001), followed by girls with high and low levels of flexibility (b=.19, p<.001, and b=.09, p<.05, respectively). However, peer deviance was not related to delinquency for males with high levels of flexibility (b=.07, ns). Additional analyses of simple slope differences (Dawson & Richter, 2006) indicated that the low vs. high flexibility slopes differed for males but not females (males: t=3.57, p<.001; females: t=−1.21, ns), and the slopes for males vs. females differed at low flexibility (t=2.90, p<.01), but not high flexibility (t=−1.45, ns). Altogether, these analyses indicate that low flexibility is associated with increased susceptibility to negative peer influence in males, but not females.

Discussion

This prospective, multi-informant study examined temperament as a moderator of the link between deviant peer affiliation and subsequent delinquent behavior in early adolescence. Of the five temperament dimensions examined, low task orientation and low positive affect were associated with higher vulnerability to negative peer influence for both males and females, whereas low flexibility amplified the negative effects of deviant peers only for males. Conversely, high task orientation and positive affect were characteristic of adolescents who were less strongly affected by deviant peer influences, with high flexibility also characteristic of resilient males. The temperament dimensions of general activity level and sleep rhythmicity did not moderate peer deviance effects.

These results corroborate previous reports that good attention control, positive emotionality, and flexibility protect adolescents with antisocial and substance using friends from increased engagement in delinquency and substance use (Gardner et al., 2008; Wills et al., 2001). However, unlike previous studies that examined combinations of multiple aspects of temperament, we have provided a more fine-grained analysis of specific temperament dimensions interacting with the effects of peer deviance over time. Furthermore, this was the first study that evaluated the possibility of gender differences in these interactive effects.

Consistent with previous research on temperament (Sanson et al., 2004) and supporting our focus on individual temperament dimensions, the five aspects of temperament included here were empirically distinct, as indicated by their weak intercorrelations. The correlations between temperament dimensions that yielded similar patterns of moderation were particularly low (r’s ranged from .05 to .26). By contrast, activity level which was most highly correlated with other dimensions (highest |r|=.42) did not parallel the moderating patterns of the other temperament variables. These results suggest that different facets of temperament make distinct contributions to resilience to negative peer influence.

High levels of task orientation, or attention control, likely protect adolescents through better ability to self-regulate behavior. These adolescents stay on task and are not easily distracted by extraneous influences, exhibiting high levels of inhibitory control. Thus, in peer group situations conducive to antisocial behavior, these youth are more likely to suppress behaviors that may lead to short-term rewards (e.g., antisocial behavior leading to immediate peer reinforcement) in favor of behaviors consistent with long-term goals (typically, conventional behaviors). Conversely, low attention control is associated with less inhibition and more impulsive behavior, which increases children’s sensitivity to reinforcement of deviant peer interactions (Snyder et al., 2010).

Attention control, as well as flexibility, may also protect adolescents through their contributions to the development of social skills and prosocial behavior (Eisenberg et al., 1995; Guerin, Gottfried, Oliver, & Thomas, 1994). Prosocial behavior and low levels of antisocial behavior are associated with resilience to negative peer influence (Mrug & Windle, 2009; Vitaro et al., 2000), and may decrease adolescents willingness to engage in antisocial activities that may endanger their prosocial goals and relationships. In addition, good social skills may help these youth successfully negotiate peer group interactions despite low conformity to deviant group norms.

Additionally, attention control and flexibility are associated with academic success (Bramlett, Scott, & Rowell, 2000; Guerin et al., 1994) and more positive relationships with parents, teachers, and peers (Eisenberg et al., 1995; Mobley & Pullis, 1991; Rudasill & Rimm-Kaufman, 2009). These experiences of success and quality relationships may serve as important sources of positive reinforcement. Positive feedback from antisocial peers for deviant behavior, an important mechanism of negative peer influence (Dishion, Spracklen, Andrews, & Patterson, 1996), may thus be less influential. Close relationships with parents and others also contribute to the development of empathy (Zahn-Waxler & Radke-Yarrow, 1990) which protects adolescents from engaging in antisocial behavior (Lovett & Sheffield, 2007), and may also diminish the influence of deviant peers. Although not directly addressed in this study, these speculations suggest that attention control and flexibility may protect adolescents from deviant peer influences through multiple pathways.

Further analyses of gender differences indicated that the moderating role of flexibility held only for boys, but not girls. This finding is consistent with existing research showing that low flexibility is more strongly related to problem behavior and difficulties in peer relationships for boys than girls (Prior, Smart, Sanson, & Oberkleid, 2001; Sanson et al., 2004). These differences may reflect greater difficulties with flexibility observed for boys (also replicated in the present study) and/or greater socialization expectations of boys to be behaviorally flexible compared to girls. However, the gender differences observed in this and other studies have been small, and the present results of low flexibility increasing susceptibility to negative peer influence for boys but not girls need to be replicated in other investigations before they can be accepted as reliable.

Although negative affect was not directly measured in this study, low levels of positive affect (which was measured) may be indicative of depressive symptoms. Depressive symptoms often co-occur with delinquent behavior in adolescence (Loeber & Keenan, 1994; Sigfusdottir, Farkas, & Silver, 2004) and contribute to escalating trajectories of antisocial behavior over time (Beyers & Loeber, 2003; Loeber, Russo, Stouthamer-Loeber, & Lahey, 1994). This escalation may be due to high levels of involvement with deviant peers found in adolescents with both externalizing and internalizing problems (Dishion, 2000; Patterson & Capaldi, 1990). Adolescents with comorbid externalizing and internalizing problems also demonstrated the highest levels of deviancy training and later arrests (Dishion, 2000). Our results add to this literature, suggesting that low positive affect also amplifies youths’ vulnerability to positive reinforcement from deviant peers. Low positive emotionality may also lead to less closeness with parents and others and thus diminish positive reinforcement from prosocial sources. Although friendships with deviant peers may be similarly affected by a lack of positive affect, ensuing lower relationship quality seems to amplify negative peer influence and lead to greater escalation in antisocial behavior (Poulin, Dision, & Haas, 1999). Finally, low positive affect may also make the youth less empathic and more reactive to deviant peer influences. Again, these speculations suggest multiple ways through which low positive affect may increase susceptibility to negative peer influence that can be tested in future studies.

This was the first study that examined sleep rhythmicity as a possible moderator of negative peer influences. The results suggest that this temperamental trait does not alter early adolescents’ susceptibility to peer influences. However, other aspects of sleep (e.g., sleep amount or quality) have been linked with adolescents’ behavior problems and substance use (Wong, Brower, & Zucker, 2009) and may be explored as moderators of peer influences in future research. Our finding that general activity level did not moderate the effect of deviant peers on delinquency is discrepant from Wills et al. (2001) who found that a combination of high activity level and negative emotionality amplified the effect of peer substance use on early adolescents’ use. It is possible activity level plays a greater role in promoting peer influences on substance use than delinquency, or that the moderation reported by Wills et al. was primarily due to negative mood than activity level. Future studies should examine separate contributions of these temperament dimensions to peer influence susceptibility for both delinquency and substance use outcomes.

Several limitations need to be noted. The magnitude of the obtained effects was small, although it was comparable to results of other studies examining temperament and other moderators of peer influence (Gardner et al., 2008; Mrug & Windle, 2009; Vitaro et al., 2000). The study focused on selected aspects of temperament that may only partly overlap with other conceptualizations of this multifaceted construct. For instance, the dimension of effortful control that has received much research attention recently (e.g., Eisenberg et al., 2004) involves attention control and flexibility which we studied, as well as inhibitory and activational control not studied here. Similarly, we did not consider other temperament dimensions identified by previous research, such as social approach-withdrawal or impulsivity. Another limitation is the use of adolescents’ reports of their friends’ deviant behavior, which are typically biased in the direction of increased similarity (Berndt & Keefe, 1995). A replication using other informants for these reports (e.g., friends’ self-reports or teacher reports) would be helpful. The use of an expanded measure of delinquency at Wave 2 could also be viewed as a limitation. Although measuring delinquency more extensively in early adolescence than in late childhood is developmentally appropriate, it weakens the inference that peer deviance was related to changes in delinquent behavior over time. Using the same measure of delinquent behavior for both waves would lead to stronger conclusions about the temporal sequencing of peer deviance and changes in antisocial behavior. Also, although the sample was diverse in its sociodemographic characteristics, the majority of participants were low income, urban African American adolescents from a single geographic area in the U.S. Thus, generalizability of the results to other populations (e.g., mostly Caucasian suburban or rural adolescents) may be limited. The results may also be less generalizable to youth who were more likely to drop out from the study, that is, Caucasians, those with higher income, lower delinquency and activity level, and higher flexibility. It is also important to note that the results of the present study may not generalize to other developmental periods. Early adolescence is characterized by generally high susceptibility to peer influence (Steinberg & Monahan, 2007), which may enhance the effect of moderating characteristics. It would be informative to address the moderating role of temperament for peer influence in late childhood and later adolescence.

Despite these limitations, the findings provide novel insights into factors explaining individual differences in susceptibility to peer influences. Task orientation, positive affect, and for males also flexibility characterize adolescents who are resilient to negative peer influences, while distractibility, low positive affect, and for males rigidity increase youths’ vulnerability. Interventions that increase children’s self-regulation and emotional well-being may contribute to less antisocial behavior in adolescence through increasing resilience to negative peer influences. Additionally, temperamentally vulnerable individuals may be at greater risk of deviant peer training in group settings with high concentrations of deviant peers (Gifford-Smith, Dodge, Dishion, & McCord, 2005). Reducing opportunities for negative peer influence combined with providing youth with skills that promote resilience to negative peer influences may be the most effective strategy for reducing the impact of deviant peers on antisocial behavior. Interventions and developmental theory would be further informed by research examining the specific mechanisms through which particular aspects of temperament alter resilience to negative peer influences in adolescent boys and girls.