Abstract This study tests the hypothesis that self-regulation serves as a resiliency factor in buffering youth from negative influences of peer deviance in middle to late adolescence. The interactive effects between peer deviance and self- regulation were investigated on change in antisocial behav- ior from age 17 to 19 years in an ethnically diverse sample of adolescents. A multi-agent construct was created using adolescent, parent, and teacher reports of self-regulation and peer deviance. Results indicated that self-regulation shows convergent validity and covaries as expected with developmental patterns of adolescent antisocial behavior. Self-regulation moderated the association of peer deviance with later self-reported adolescent antisocial behavior after controlling for prior levels of antisocial behavior. The implications of these findings for models for the develop- ment of antisocial behaviors and for intervention science are discussed.

Adolescent Self-Regulation as Resilience: Resistance
to Antisocial Behavior within the Deviant Peer Context

Theodore W. Gardner & Thomas J. Dishion &
Arin M. Connell

Published online: 25 September 2007
# Springer Science+ Media, LLC 2007

Abstract This study tests the hypothesis that self-regulation
serves as a resiliency factor in buffering youth from negative
influences of peer deviance in middle to late adolescence.
The interactive effects between peer deviance and self-
regulation were investigated on change in antisocial behav-
ior from age 17 to 19 years in an ethnically diverse sample of
adolescents. A multi-agent construct was created using
adolescent, parent, and teacher reports of self-regulation
and peer deviance. Results indicated that self-regulation
shows convergent validity and covaries as expected with
developmental patterns of adolescent antisocial behavior.
Self-regulation moderated the association of peer deviance
with later self-reported adolescent antisocial behavior after
controlling for prior levels of antisocial behavior. The
implications of these findings for models for the develop-
ment of antisocial behaviors and for intervention science are
discussed.

Keywords Self-regulation . Deviant peers .

Antisocial behavior. Adolescence

A plethora of longitudinal and intervention research points
to the central role of deviant peers in the early emergence
and course of antisocial behavior from childhood through
adolescence (for reviews see Dishion and Patterson 2006;
Dodge et al. 2006). Affiliation with a deviant peer group is
associated with both early- and late-onset trajectories of
antisocial behavior, amplifying and helping to maintain
antisocial behavior manifested in childhood as well as
promoting emergence of antisocial behavior during adoles-
cence (e.g., Patterson et al. 1998; Simons et al. 1994;
Snyder et al. 2005; Vitaro et al. 1997). The findings relating
deviant peer influence range across a number of measure-
ment methods as well as analytic techniques. For example,
in an analysis of developmental trajectories, high and
chronic levels of antisocial behavior among males were
partly characterized by deviant peer affiliation (Wiesner and
Capaldi 2003). In two mixed-gender samples van Lier et al.
(2005) reported that a subclass of youth characterized by
chronic-high antisocial behavior was nearly exclusively
male, tended to affiliate more with antisocial peers, and
experienced more peer rejection relative to moderate- and
low-antisocial behavior groups.

Although it is clear peers can influence the development
and course of antisocial behavior, especially among youth
who are moderately involved in such behavior (Vitaro et al.
1998), it is also interesting that many youth who have
regular, repeated contact and friendships with ‘deviant’
peers do not seem to be negatively influenced. Indeed,
one major theoretical orientation on the development of
crime is that those who engage in problem behaviors lack
‘self control’, and that this personal deficit is the ‘cause’,
and peer deviance is simply the venue (e.g., Gottfredson
and Hirschi 1990; Hirschi 2004). In reviewing the literature
on the development of antisocial behavior and substance
use in children and adolescents, it has been proposed that

J Abnorm Child Psychol (2008) 36:273–284
DOI 10.1007/s10802-007-9176-6

T. W. Gardner (*) : T. J. Dishion : A. M. Connell
Child and Family Center, University of Oregon,
Eugene, Oregon, USA
e-mail: [email protected]

T. W. Gardner : T. J. Dishion : A. M. Connell
Department of , University of Oregon,
Eugene, Oregon, USA

T. W. Gardner
Pepperdine University,
24255 Pacific Coast Highway,
Malibu, CA 90263, USA

A. M. Connell
Department of , Case Western Reserve University,
Cleveland, OH, USA

children’s self-regulation potentially serves as both a main
effect on the development of antisocial behavior (as in the
Hirschi model) and also as a moderator (Wills and Dishion
2004; Dishion and Patterson 2006). That is, children and
adolescents with low self-regulation were hypothesized to
be more vulnerable to pathogenic relationship dynamics in
general, and to peer influence in particular. Conversely,
youth high in self-regulation are more able to resist
temptations of peers, and keep track of long-term goals
despite opportunities for short-term high intensity social
rewards often provided by the adolescent peer network (see
Dishion et al. 2004). Figure 1 provides a conceptual
overview of the hypothesis of the role of adolescent self-
regulation in the development and course of antisocial
behavior.

The concept of self-regulation is considered to be an
individual-difference dimension that includes goal setting,
planning, task persistence, and environmental management
as well as modulation of behavioral, emotional, and atten-
tional reactivity (Rothbart and Posner 2005). Self-regulation
involves “initiating, avoiding, inhibiting, maintaining, or
modulating the occurrence, form, intensity, or duration of
internal feeling states, emotion-related physiological, atten-
tional processes, motivational states, and/or the behavioral
concomitants of emotion in the service of accomplishing
affect-related biological or social adaptation or achieving
individual goals” (Eisenberg and Spinrad 2004, p. 338).
Self-regulation develops over time through transactional
processes among constitutionally based differences in
reactivity and regulation, maturation of the executive
attention system, and socialization through educational and
social experiences in school, family, and peer contexts (see
Dishion and Patterson 2006; Rothbart and Ahadi 1994;
Rothbart and Posner 2005; Wills and Dishion 2004).

The analysis of various facets of self-regulation in children
and adolescents has an extensive recent history in develop-
mental psychology. Eisenberg and Spinrad’s (2004) definition
of self-regulation draws heavily upon the temperamental

construct of effortful control, which involves the “ability to
inhibit a dominant response to perform a subdominant
response” (Rothbart and Bates 1998, p. 137). This ability
to inhibit an initial prepotent response is a key executive
function that creates a delay in responding to the immediate
context and creates the temporal space requisite for effortful,
or volitional, goal-directed actions (Barkley 2001). In this
fashion systems of effortful control “allow the approach of
situations in the face of immediate cues for punishment, and
avoidance of situations in the face of immediate reward”
(Posner and Rothbart 2000, p. 434).

Executive control of attention is considered to be an
important underpinning of effortful control and of effective
self-regulation. The executive attention network is often
studied using tasks involving conflict, most often variants
of the Stroop task. Resolving conflict activates anterior
cingulate cortex and lateral prefrontal cortex, which are
involved in the executive attention network (Botvinick et
al. 2001; Bush et al. 2000; Fan et al. 2003). Neuroimaging
studies have provided evidence that the executive attention
network is involved in self-regulation of both negative and
positive emotion (e.g., Beauregard et al. 2001). The
executive attention system appears to begin to come online
around 2 years of age and changes rapidly during the third
year of life and is a central component of the broader
temperamental dimension of effortful control (Posner and
Rothbart 1998). Executive control of attention is valuable
in terms of self-regulation because it allows the individual
to distract him or herself from noxious stimuli in the
environment (i.e., volitional disengagement). The conflict
studies also demonstrate that executive attention is impor-
tant for filtering out distracting stimuli, and thus maintain-
ing focus in the presence of multiple stimuli competing for
attentional resources.

Relevant to the empirical question posed in this paper
regarding the development of antisocial behavior, higher
levels of effortful control have been linked to the
development of conscience (Kochanska 1991, 1995) and
empathy (Rothbart et al. 1994). Higher effortful control is
associated with fewer conduct problems and better social
adjustment in childhood (Eisenberg et al. 2001a, 2000a,
2001b). Similarly, low effortful control is associated with
externalizing problems in preadolescence (Oldehinkel et al.
2004). Because many of the neural systems thought to
underlie emotion regulation appear to mature throughout
adolescence (Spear 2000), it is expectable that individual
differences in self-regulation during adolescence play an
important role in adolescent psychosocial adjustment as
well.

Empirical evidence clearly indicates that self-regulation,
variously defined and measured, is a reliable main effect in
the development and manifestation of antisocial behavior
across childhood and adolescence and into adulthood. For

Fig. 1 Self-regulation in the amplification of problem behavior in
adolescence

274 J Abnorm Child Psychol (2008) 36:273–284

example, in a study of boys with ADHD, highly aggressive
boys scored low on several indices of emotion regulation,
whereas boys with ADHD who were not aggressive scored
within the range of non-diagnosed boys on every assessed
category of emotion regulation (Melnick and Hinshaw
2000). Similarly, poor behavioral self-control (i.e., delay of
gratification) has been linked specifically to both aggressive
and delinquent externalizing disorders in early adolescence
(Krueger et al. 1996). Henry et al. (1996) reported that a
temperamental dimension assessed at ages 3 and 5, which
they labeled lack of control, was related to externalizing
behavior problems at ages 9 and 15 and predicted
conviction for a violent offense by age 18. Among adults,
lower levels of antisocial behavior have been associated
with the broad personality dimension of constraint (Krueger
et al. 2001), which has a self-regulatory component to it
labeled control. One limitation of the extant literature
addressed in the current study is that much of the research
linking self-regulation to problem behavior often include
only one reporting agent and there is less measurement work
defining individual differences in youth self-regulation from
a multi-agent and multi-method framework.

Resilience: Person by Environment Interaction

Although main effects of both deviant peer affiliation and
self-regulation on antisocial behavior have been well
established in previous research, less is known empirically
regarding the transactional relation of these two factors in
the development and expression of antisocial behavior.
From a development and psychopathology perspective,
there is an explicit effort to integrate individual difference
constructs with environmental factors when explaining the
emergence and course of both adaptation and maladaptation
(Cicchetti 1990; Sroufe and Rutter 1984). Accordingly, the
research question of interest here is not one of main effects
but rather a question of how individual differences in self-
regulation modulate the risk for growth in antisocial
behavior conveyed by a deviant peer environment.

From a learning perspective, evidence indicates that the
toxic influence of the deviant peer group appears to operate,
at least in part, via a process of deviancy training, in which
delinquent dyads of youth primarily react positively to
deviant talk, and nondelinquent dyads ignore deviant talk in
favor of normative discussions (Dishion et al. 1996).
Observational data indicate that deviancy training in
delinquent dyads is associated with increased probability
of substance use initiation and self-reported delinquency
during a two-year period (Dishion et al. 1995, 1996), as
well as increases in violent behavior (Dishion et al. 1997).
Deviancy training also predicts growth in new forms of
antisocial behavior during adolescence (Patterson et al.

2000). Thus, reinforcement contingencies present in the
peer ecology appear to play a central role in the emergence,
maintenance, and growth of antisocial behavior during
adolescence.

Conceptually, because effortful control, and thus the
broader construct of self-regulation, facilitates avoidance of
situations in the face of immediate reward it is reasonable to
propose that individual differences in self-regulation mod-
erate the relationship between deviant peer affiliation and
antisocial behavior. Within the deviant peer context, as
discussed previously, there are often contingencies that
serve to reinforce youth antisocial behavior. Logically,
within the deviant peer ecology, the behaviors of youth with
greater self-regulation would likely be less influenced by
the immediate reinforcement contingencies present in the
deviant peer group. In this manner self-regulation may be
considered a source of resilience—relatively positive
adaptation in the face of heightened risk for maladaptation
(Luthar 2006). Consistent with the characterization of self-
regulation as a source of resilience, the central hypothesis
of this study is that individual differences in self-regulation
moderate the relation between peer deviance and growth in
antisocial behavior during middle- to late-adolescence.

Some evidence of a transactional relation between self-
regulation and risk for antisocial behavior conveyed by
environmental factors was reported by Henry et al. (1996)
who reported significant statistical interactions between
temperamental lack of control and two environmental
variables: single-parent home at age 13 and number of
changes in the family configuration since the last assess-
ment. The presence of both low control and the experience
of a single-parent home at age 13 were associated with
increased likelihood of conviction for a nonviolent offense.
The interaction between temperamental lack of control and
the number of changes in family configuration was
associated with increased likelihood of conviction for a
violent offense. Also relevant to the conceptualization of
self-regulation as a protective factor that moderates envi-
ronmental risk is evidence that effortful control, a temper-
amental construct considered central to the regulation of
emotion and behavior (Eisenberg et al 2000a; Rothbart and
Bates 1998), is associated with resiliency in the presence of
multiple environmental risk factors (Eisenberg et al. 2004).

In a very recent study Goodnight et al. (2006) examined
the moderating role of self-regulation with respect to
deviant peer influences on adolescent delinquency. In this
study the authors focused on the construct of reward
dominance, which involves a greater sensitivity to reward
than to punishment. Individuals higher in reward domi-
nance are less likely to interrupt goal-directed behavior to
evaluate its potential negative consequences and, conse-
quently, are less likely to regulate their impulsive behavior
(Newman and Wallace 1993). The results indicated that

J Abnorm Child Psychol (2008) 36:273–284 275275

peer deviance is associated with later externalizing behavior
among youth with medium and high levels of reward
dominance, but not among youth who are low in reward
dominance (Goodnight et al. 2006). The authors suggest
that the observed relations among these variables may be
attributable, at least in part, to heightened susceptibility of
reward dominant youth to the rewards provided by deviant
friends in the context of the peer groups’ deviancy training
process.

The primary aim of the current study was to test the
prediction that for understanding adolescent antisocial
behavior, a transactional, ecological perspective is most
appropriate. From this perspective, three hypotheses are
proposed: First, that self-regulation varies as a function of
the developmental history of a youth’s antisocial behavior.
This hypothesis is designed to provide evidence of the
ecological validity of the multi-informant self-regulation
construct, and we specifically predict that low levels of self-
regulation are associated with an early-onset, persistent
pattern of antisocial behavior while high levels of self-
regulation are associated with low levels of antisocial
behavior throughout adolescence. Second, in predicting
change in antisocial behavior from middle to late adoles-
cence, both self-regulation and peer deviance will serve as
main effects. Third, that there will be a statistically reliable
interaction between self-regulation and peer deviance in
accounting for change in antisocial behavior from middle to
late adolescence. In essence, we predict that youth with
high levels of self-regulation during adolescence are less
vulnerable to the influence of deviant peers with respect to
growth in antisocial behavior.

Method

Participants

The data presented here are from an ongoing longitudinal
study. The sample and methodology have been described in
detail in other publications (Dishion and Kavanagh 2003;
Dishion et al. 2002, 2003). In brief, participants were
recruited from the entire population of 6th-grade students in
three middle schools in an ethnically diverse metropolitan
community. A total of 999 children and their families
completed the initial assessment at Wave 1 (T1). Families
were randomly assigned to intervention and control
conditions. Each year student surveys were conducted
primarily in the school context. If students moved out of
their original schools, they were followed to their new
location. At T6 and T7, students were assessed through the
mail and through the school system.

The current analyses focus on data from T6 (11th grade,
approximately 17 years old) and T7 (approximately 19

years old). Out of the original sample, 803 provided data
at T6, and 802 provided data at T7. Attrition from the
study was not related to demographic factors, including
gender, ethnic background, or socioeconomic indicators.
The sample at T6 and T7 included 44.4% European
American adolescents, 30.9% African American adoles-
cents, 5.7% Hispanic adolescents, 3.3% Asian American
adolescents, 2.3% Native American adolescents, 1.6%
Pacific Islander adolescents, and 11.7% adolescents with
multiple ethnic or racial backgrounds. Forty-nine percent
of the adolescents were female, and 34.7% were from
single-parent families. Gross annual household income
ranged from less than $4999 (9.0%) to over $90,000
(12.8%), and the median annual household income was
$30,000$39,999. The majority of primary caregivers
completed at least high school (88.4%) and many
(22.2%) had graduated college.

Measures

Self-regulation Three indicators of self-regulation were
derived from the T6 assessment based on youth, parent
and teacher reports. Adolescents and their parents complet-
ed the short form of the Early Adolescent Temperament
Questionnaire-Revised (EATQ-R; Ellis and Rothbart 2002)
to assess youths’ effortful control. Effortful control is
composed of 16 items that assess three aspects of self-
regulation: activation control, attention, and inhibitory
control. Questions that assess activation control provide
information regarding whether the youth typically performs
an action even when there is a strong proclivity to avoid it.
Questions regarding attention assess youths’ abilities to
concentrate, to filter out distractions, to attend to multiple
tasks, and to shift from one task to another. Questions that
evaluate youths’ inhibitory control provide information
about whether the youth sticks with plans and goals, is
able to keep secrets, and is able to control impulses (Ellis
2002). Together these capacities form the core of effective
self-regulation (Eisenberg et al. 2000b). Items were rated on
a 5-point scale (1=almost always untrue of you, 2=usually
untrue of you, 3=sometimes true, sometimes untrue of you,
4=usually true of you, 5=almost always true of you). Internal
consistency of the effortful control scale was acceptable for
both youth report (α=0.62) and parent report (α=0.76).

The Teacher Scale of Self-Regulation used for this study
included four items adapted from Humphrey (1982) that
provide teacher perceptions of youth behaviors and activ-
ities within the school environment. The scale provides
information regarding whether the youth: (1) thinks ahead
about the consequences of his or her actions, (2) plans
ahead before acting, (3) pays attention to what s/he is
doing, and (4) sticks to what s/he is doing until it is

276 J Abnorm Child Psychol (2008) 36:273–284

finished, even on unpleasant tasks. Items were rated on a
5-point scale (1=never, 2=almost never, 3=sometimes, 4=
often, 5=always). High internal consistency was found for
this scale (α=0.95).

Mother, youth, and teacher scales were significantly
intercorrelated. Due to the differences in perspectives (i.e.,
self and other) and contexts (i.e., non-home and home) the
concordance among the three reporters is expectably within
the low to moderate range, with the convergence between
self- and other reports predictably being the lowest
(Achenbach et al. 1987). The low to moderate correlation
between informants is likely reflective of valid measure-
ment of the unobservable self-regulation construct without
inflation of intercorrelation due to shared perspective or
context (Kraemer et al. 2003). The youth report scale
correlated significantly with both the parent scale (r=0.31,
p<0.05) and the teacher scale (r=0.21, p<0.05), while the parent report correlated significantly with the teacher report (r=0.41, p<0.05). To create the composite score, z-scores for mother, youth, and teacher scales were calculated, and the self-regulation composite scale was calculated as the mean of these z-scores. A principal components factor analysis constrained to a single solution revealed factor loadings ranging from 0.63 to 0.80, which indicates that these three indices are an acceptable index of the latent self-regulation factor. The self-regulation composite scores approximated a normal distribution. Deviant peer affiliation Deviant peer affiliation at T6 was measured by adolescent, parent, and teacher reports on four items that assess the number of youths’ peers who engage in a variety of deviant behaviors including misbehaving, breaking rules, and substance use. Items were rated on a 5-point scale, with higher scores reflecting more frequent affiliation with deviant peers. Internal consistency of this scale was acceptable for all reporters (youth: α=0.69; parent: α=0.66; teacher: α=0.84). The youth report scale correlated significantly with both the parent scale (r=0.34, p<0.05) and the teacher scale (r=0.31, p<0.05), while the parent report correlated significantly with the teacher report (r=0.39, p<0.05). To create the composite score, z-scores for mother, youth, and teacher scales were calculated, and the peer deviance composite scale was calculated as the mean of these z-scores. A principal components factor analysis constrained to a single solution revealed factor loadings ranging from 0.72 to 0.77, which indicates that these three indices are an acceptable index of the latent peer deviance factor. The peer deviance composite scores were slightly positively skewed, but reasonably approximated a normal distribution. Antisocial behavior Adolescent antisocial behavior was measured at T6 via self-report of nine items that assess frequency of antisocial behaviors during the past month on a 6-point scale (1= “never,” 6= “more than 20 times”). The frequencies of the following behaviors were assessed: (a) lying to parents, (b) skipping school, (c) staying out all night without permission, (d) stealing, (e) panhandling, (f) carrying a weapon (g) hitting or threatening to hit a person, (h) damaging property, and (i) spending time with gang members as friends. Internal consistency of this scale was acceptable (α=0.73). For a subset of analyses designed to evaluate the relation of the composite measure of self-regulation to the devel- opmental status of the adolescent, individuals were grouped into one of three antisocial behavior patterns based on self- report of these nine items at earlier assessment waves. Reliability analyses indicated that internal consistency of this scale was strong for each of the earlier waves of assessment: T1 (6th grade; α=0.83), T2 (7th grade; α=0.84), and T3 (8th grade; α=0.77). Information contained in juvenile court records of youth until age 16 was also used in an analysis to support the ecological validity of the a priori antisocial behavior grouping procedure. For this analysis, a sum score of number of documented arrests for misdemeanors and felonies by age 16 was used. Antisocial behavior at T7 was measured via self-report on the Adult Self Report questionnaire (Achenbach and Rescorla 2003). Items measuring T6 antisocial behavior were not collected at T7. In order to maximize content similarity across T6 and T7, the DSM-IV Antisocial Behavior scale score from the ASR was used. This scale consists of 20 items assessing youth engagement in antisocial behaviors (e.g. “I break rules at work or elsewhere,” “I don’t feel guilty after doing something I shouldn’t,” “I get in many fights”), with items measured on a 3-point scale (0=not true, 1=sometimes true, 2=often true). High internal consistency for this scale was observed in the current sample (α=0.81). Expectably, distributions of both T6 and T7 antisocial behavior were positively skewed. Analysis To test the multivariate relation between self-regulation, peer deviance and antisocial behavior, a 5-step hierarchical regression analysis was employed to determine the degree to which self-regulation influences the later development of antisocial personality problems. Step one included the covariates treatment condition (whether they received treatment or not), gender, and ethnicity. Step two added T6 antisocial score to account for previous antisocial behavior levels. Step three entered peer deviance, a known predictor variable, to the regression equation. Step four entered the self-regulation variable, and step five entered the self-regulation by peer deviance interaction term. The predictor variables were centered to maximize interpret- J Abnorm Child Psychol (2008) 36:273–284 277277 ability and to minimize potential problems with multi- collinearity among the predictor variables and their higher order terms as a result of scaling (Aiken and West 1991). Results Bivariate correlations among demographic variables, self- regulation, deviant peer affiliation, and T6 and T7 antisocial behavior are presented in Table 1. Intervention status was not significantly associated with any other variable in the analysis. Gender was significantly associated with deviant peer affiliation and self-reported antisocial behavior at T6 and T7. Males reported significantly more involvement with a deviant peer group and more frequent engagement in antisocial behavior at both time points. Ethnicity was recoded to create a dichotomous variable representing whether the adolescent identified him or herself as European American or as non-European American. Non- European American youth reported significantly more involvement with peers who engage in deviant behavior than youth who self-identified as European American. Of primary interest for the empirical question posed by this study were the associations between deviant peer affiliation, antisocial behavior at wave 6 and wave 7, and self-regulation. As shown in Table 1, the intercorrelations among these variables were all statistically significant in the expected directions. Youth who spent more time with deviant peers reported more engagement in antisocial behaviors at both wave 6 and wave 7 relative to youth who spent less time in a deviant peer group, and youth who affiliated more with deviant peers were lower in self- regulation. Expectably, levels of antisocial behavior at wave 6 and wave 7 were significantly associated with one another, with higher levels of antisocial behavior at wave 6 associated with higher levels of antisocial behavior at wave 7. Youth who were rated higher in self-regulation reported less engagement in antisocial behavior at both time points relative to youth rated lower in self-regulation. Self-Regulation and Developmental Status of the Adolescent Three a priori groups of youth were formed based on their reported patterns of antisocial behavior. Youth were classified as manifesting an early-onset persistent develop- mental pattern of antisocial behavior if they scored above the sample median on the antisocial behavior index at T1 (6th grade), at either T2 (7th grade) or T3 (8th grade), and again at T6 (11th grade). Youth who were at or below the sample median on the antisocial behavior index at T1, T2, and T3, but were above the sample median at T6 were classified as reporting late-onset antisocial behavior. A third subgroup of youth was identified that did not manifest elevated antisocial behavior at any of the four assessment points. Only youth who provided data at each of the four time points were eligible for inclusion in these groups. The a priori decision rules resulted in a group of 66 youth (54.5% male, 43.9% European American) who reported early-onset antisocial behavior that persisted through T6 (approximately age 17), a group of 70 youth (54.3% male, 71.4% European American) who reported their first elevated level of antisocial behavior at T6, and a group of 215 youth (38.6% male, 54.9% European American) who reported levels of antisocial behavior at or below the sample median at all 4 assessment points. As a simple test of external validity of the a priori …