You are on page 1of 14

AGGRESSIVE BEHAVIOR

Volume 36, pages 127–140 (2010)

Testing the Developmental Distinctiveness of Male


Proactive and Reactive Aggression With a Nested
Longitudinal Experimental Intervention
Edward D. Barker1, Frank Vitaro2, Eric Lacourse2, Nathalie M. G. Fontaine3, Rene Carbonneau1, and
Richard E. Tremblay2
1
Department of Psychology, Center for the Prevention of Youth Behavior Problems, University of Alabama, Tuscaloosa, Alabama
2
University of Montreal, Groupe de recherche sur l’inadaptation psychosociale chez l’enfant, Montréal, Québec, Canada
3
Department of Criminal Justice, Indiana University, Bloomington, Indiana

: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :
An experimental preventive intervention nested into a longitudinal study was used to test the developmental distinctiveness of
proactive and reactive aggression. The randomized multimodal preventive intervention targeted a subsample of boys rated disruptive
by their teachers. These boys were initially part of a sample of 895 boys, followed from kindergarten to 17 years of age.
Semiparametric analyses of developmental trajectories for self-reported proactive and reactive aggression (between 13 and 17 years of
age) indicated three trajectories for each type of aggression that varied in size and shape (Low, Moderate, and High Peaking). Intent-
to-treat comparisons between the boys in the prevention group and the control group confirmed that the preventive intervention
between 7 and 9 years of age, which included parenting skills and social skills training, could impact the development of reactive more
than proactive aggression. The intervention effect identified in reactive aggression was related to a reduction in self-reported coercive
parenting. The importance of these results for the distinction between subtypes of aggressive behaviors and the value of longitudinal-
experimental studies from early childhood onward is discussed. Aggr. Behav. 36:127–140, 2010. r 2010 Wiley-Liss, Inc.

: : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : : :

Keywords: proactive; reactive; developmental trajectories; intervention; mediation

Over the past two decades, researchers have empha- in reaction to harsh, threatening, or unpredic-
sized the distinction between reactive aggression (RA) table environment or cold and abusive parenting.
and proactive aggression (PA) [Dodge and Coie, 1987] Conversely, PA is believed to thrive in environments
on the basis of function and motivation. RA can be that support the use of aggression as a means to
traced to the frustration–anger theories of aggression achieve a goal, such as criminogenic families or gangs
[Berkowitz, 1962, 1989; Dollard et al., 1939], where [Akers, 1998; Vitaro et al., 2006b]. There are, to
aggressive behavior can be a response to provocation, our knowledge, four studies that support the idea that
frustration, or threat, and is usually accompanied by PA and RA develop from distinct experiences. For
the expression of anger. The construct of PA, on the RA, incarcerated youth show retrospective histories
other hand, is more in line with the social learning of physical abuse and harsh parental discipline,
model of aggression and/or criminal behavior
[Bandura, 1973], where aggression is an acquired Grant sponsor: Québec’s FQRSC; Grant sponsor: FRSQ; Grant
behavior governed by external reinforcement contin- sponsor: Canada’s Social Sciences and Humanities Research
Council; Grant sponsor: Canadian Institutes for Health Research;
gencies. According to this notion, PA requires neither Grant sponsor: U.S. National Science Foundation; Grant number:
anger nor provocation, but is driven by the anticipa- SES-9911370; Grant sponsor: U.S. National Institute of Mental
tion of reward [Dodge and Coie, 1987]. Health; Grant number: MH65611-01A2
Correspondence to: Edward D. Barker, Department of Psychology,
Center for the Prevention of Youth Behavior Problems, University of
Alabama, Tuscaloosa, AL 35401. E-mail: ted.barker@ua.edu
DEVELOPMENTAL MODEL OF PA AND RA
Received 5 August 2009; Revised 5 November 2009; Accepted 14
PA and RA are hypothesized to originate November 2009
from different social experiences [Agnew, 1992; Published online 5 January 2010 in Wiley InterScience (www.
Dodge, 1991]. Specifically, RA is believed to develop interscience.wiley.com). DOI: 10.1002/ab.20337

r 2010 Wiley-Liss, Inc.


128 Barker et al.

whereas PA-only and PA–RA mixed aggressive teacher reports developed by Dodge and Coie
individuals do not [Dodge et al., 1997]. In two [1987]. Although the two types of aggression are
longitudinal studies, harsh parenting predicted RA highly related (r 5 .70) [Kempes et al., 2005; Vitaro
but not PA [Vitaro et al., 2006a] and negative child and Brendgen, 2005], variable-centered studies have
emotionality predicted RA but not PA [Xu et al., achieved reliable and valid measurement and con-
2009]. Finally, the development of PA, but not of RA, struct validity [Dodge and Coie, 1987; Poulin and
has been retrospectively identified to be initiated by Boivin, 2000a], even if the two types of aggressive
exposure to aggressive role models in the family that behaviors are often manifested by the same indivi-
value the use of aggression to resolve conflict or to duals [Barker et al., 2006; Crick and Dodge, 1996;
advance personal interests [Connor et al., 2004; Day et al., 1992; Fite et al., 2008; Little et al., 2003;
Dodge et al., 1997]. Pellegrini et al., 1999; Salmivalli and Nieminen,
Other studies have reported distinct correlates and 2002].
consequences for each type of aggression [see Measurement and construct validity of PA and
Kempes et al., 2005; Vitaro and Brendgen, 2005]. RA, however, have typically assessed the distinc-
For example, several studies have shown that RA tiveness of each type of aggression by using cross-
but not PA is related to hostile attribution biases in sectional or pooled data, and studies have also
response to ambiguous provocation situations created a priori heterogeneous groups using thresh-
[Dodge and Coie, 1987; Dodge et al., 1997; Hubbard olds (e.g., 1 SD above the group mean). This
et al., 2001; Nas et al., 2005; Orobio de Castro et al., strategy typically results in a majority of children
2002], negative affect [Fite et al., 2009], attention for being discriminated as non-aggressive, followed by
rejection, ridicule, and failure cues [Schippel et al., mixed PA–RA, RA-only, and PA-only [Vitaro and
2003], problem-solving deficits in difficult social Brendgen, 2005]. Although these studies have
situations [Brown and Kolko, 1999; Dodge, 1991; yielded important information about the prevalence
Price and Dodge, 1989], and impairment in execu- of different types of aggression, they have not
tive functions [Ellis et al., 2009]. elucidated the developmental stability of the PA
In contrast, PA children who appear to value and RA subtypes [Barker et al., 2006].
aggression as an effective means to achieve desired To our knowledge, only one published person-
goals more than other children [Crick and centered study has examined the dual development
Dodge, 1996; Dodge et al., 1997] and affiliate with of PA and RA [Barker et al., 2006]. This study
similarly aggressive peers in deviant cliques [Dodge employed a data analytic technique that allows for
and Coie, 1987; Dodge et al., 1997; Poulin and Boivin, the empirical identification of children following
2000b], show psychopathic tendencies [Nouvion distinct trajectories of each type of aggression as well
et al., 2007] and engage in higher levels of antisocial as the interrelationships among the different trajec-
behavior [Fite et al., 2009]. Many of these correlates tories [Nagin, 2005; Nagin and Tremblay, 2001].
have been replicated cross-nationally [Fossati et al., Contrary to the literature reviewed, in a longitudinal
2009; Xu and Zhang, 2008]. sample of high-risk boys between the ages of 13–17,
The consequences of PA and RA also appear Barker et al. [2006] found that PA and RA
distinct. For instance, RA in preadolescence is related trajectories generally co-occurred (i.e., mixed PA
to greater risk for internalizing disorders, such as and RA) and were unable to identify boys who were
neuroticism and depression in mid-adolescence and high in one type of aggression but low in the other
adulthood [Dodge et al., 1997; Pulkkinen, 1996; (i.e., PA-only and RA-only). Moreover, boys
Vitaro et al., 2002] and the use of physical aggression following high peaking trajectories in both types of
in teen dating experiences [Brendgen et al., 2001, aggression—not just PA—were over twice as likely
2002]. Conversely, preadolescent to early adolescent to have affiliated with deviant peers (between ages
PA has been associated with delinquency, general 13 and 17).
aggressiveness, and violent conduct problems in mid- Because the developmental trajectories of PA and
adolescence and adulthood [Brendgen et al., 2001; RA co-occurred and boys who were high in both
Fite et al., 2008; Pulkkinen, 1996; Vitaro et al., 1998]. types of aggression were more likely than boys in the
other trajectories to engage in delinquent lifestyles
[Barker et al., 2006], it is tempting to conclude that,
METHODOLOGICAL MATTERS IN THE
opposite to Dodge’s model, PA and RA may in fact
MEASUREMENT OF PA AND RA
share a common etiology. Indeed, the conceptual
The majority of studies on RA and PA have and methodological distinctiveness of PA and RA
operationalized these two types of aggression with has highlighted the need for a better understanding

Aggr. Behav.
Testing the Developmental Distinctiveness 129

of the etiology of these two types of aggressive The aim of this study was to test the develop-
behavior. One methodological avenue through mental distinctiveness of PA and RA and their
which the developmental distinctiveness of these putative mediators, by using data from an experi-
two types of aggression may be examined is via mental intervention nested within a prospective
randomized controlled, preventive interventions. longitudinal study [Tremblay et al., 2003]. To test
these hypotheses, we identified a subsample of boys
from the original sample [i.e., Barker et al., 2006;
PREVENTIVE INTERVENTIONS TO TEST THE explained in detail below] and identified distinct
DEVELOPMENTAL DISTINCTIVENESS OF PA developmental trajectories of PA and RA during
AND RA adolescence. We then assessed the impact of the
experimental intervention on the developmental
The developmental model proposed by Dodge
trajectories of PA and RA. The multimodal
implies that a preventive intervention that targets
program targeted improvement in parents’ manage-
putative causal factors, such as coercive parenting
ment skills and their children’s social-cognitive
practices, children’s social-cognitive deficits, chil-
skills. This program has been shown to have a
dren’s social skills, and peer experiences (i.e., peer
significant impact on parental supervision and
rejection) has the potential to modify the develop-
aggressive behavior, such as physical aggression,
ment of RA but not PA. Conversely, based on a
vandalism, theft, and gang membership [Boisjoli
model, such as the Gottfredson and Hirschi’s [1990]
et al., 2007; Lacourse et al., 2002; Vitaro et al., 1999;
self-control model, and as appears to be suggested
Vitaro et al., 2001].
by Barker et al.’s [2006] findings, a preventive
In accordance with Dodge’s model of dual etiology,
intervention may also be expected to equally impact
it is possible that a preventive intervention targeting
PA and RA. In their explanation, Gottfredson and
RA factors, such as parents’ management skills and
Hirschi [1990] state that the propensity to engage
children’s social-cognitive skills, may have an impact
in aggressive and antisocial behaviors—low self-
on the probability of intervention boys to follow the
control—is laid out in the family of origin (e.g.,
low trajectories of RA, but not PA. Conversely, the
ineffective child-rearing, negative reinforcement, low
common etiological factors model inspired from
parental monitoring). This theoretical standpoint
Gottfredson and Hirschi’s [1990] perspective suggests
differs, however, as these researchers claim that low
a generalized impact of the preventive intervention on
self-control can explain all individual differences in
both RA and PA. In the case where a differential
the propensity to engage in all acts of crime and
impact is found, we intend to test whether or not the
deviance (including PA and RA, equally). There-
intervention effects were indirectly affected by factors
fore, an intervention that targets factors associated
hypothesized to be distinct to PA and/or RA. It is
to the development of delinquency, such as parent-
important to note that the role of the indirect effects
ing and social skills, could be expected to equally
is the key: the tenability of a variable that indirectly
impact PA and RA.
affects one type of aggression, such as peer rejection
and the development of RA, is increased if rival
variables that could indirectly effect an intervention
ADVANTAGES OF THE USE OF PREVENTIVE
effect on PA are taken into account in the same
INTERVENTIONS TO TEST THEORETICAL
analysis. This strategy can help to eliminate the
MODELS
possibility that relevant but omitted variables linked
An adequately assessed preventive and corrective to PA and RA are significantly related to the
intervention that targets postulated causal factors is intervention effects [Howe et al., 2002]. In other
not only a pragmatic test of an intervention’s words, testing indirect effects for PA and RA in the
effectiveness, but also offers an excellent opportunity presence of positive RA intervention effects increases
to test causal hypotheses of the development of RA confidence that the estimated effects are owing to
and PA [e.g., Schwartz et al., 1980]. Demonstrating putative RA factors and not to co-occurring PA
that an intervention imbedded in an experimental factors [Bierman et al., 2002; Reynolds et al., 2004].
design modifies a trajectory of aggressive behavior We tested for potential indirect effects of variables
through putative mediators, in a manner that is hypothetically distinct to PA, such as deviant group
compatible to a putative etiological model, can be membership, low parental supervision, and positive
considered a more convincing test of causal theory expectancies toward use of aggression in addition to
compared with retrospective data or even correla- variables hypothetically distinct to RA, such as
tional prospective data. coercive parenting and peer rejection.

Aggr. Behav.
130 Barker et al.

In accordance with Dodge’s dual etiology model intervention. The disruptive boys were randomly
which is supported by a majority of empirical studies assigned to one of the following groups: (a)
in comparison to Gottfredson and Hirschi’s com- intervention group (INT; n 5 69), (b) no treatment
mon etiological factors perspective, we expected: (1) control group (CONT; n 5 58); and (c) observa-
boys who participated in the preventive intervention tional group (OBS; n 5 123). One hundred and
would be more likely to follow desisting and/or low seventy-two of the 250 families who were rando-
levels of RA, (2) intervention effects would not be mized actually participated in the intervention
identified for PA, and (3) the intervention effects— (INT 5 46; CONT 5 42; OBS 5 84). The remaining
the probabilities of following the desisting and/or non-completers, i.e., the boys and their families who
low trajectories of RA—would be mediated by RA did not participate in the intervention (n 5 78), were
and not PA factors. included in the analyses to allow for intent-to-treat
analyses (ITT; explained below). Owing to the
missing data on variables of interest, the final
METHOD intervention sample was 198 for the mediational
analyses (INT 5 54; CONT 5 50; OBS 5 94). The
Sample
excluded boys did not significantly differ from the
Behavior ratings of male students, mean age 6.1 included boys on any of the variables implemented
years; SD 5 0.32, were obtained from 87% of the in this study.
kindergarten teachers from 53 schools in lower The OBS group was included in the experimental
socioeconomic areas in Montréal (Québec, Canada) design to observe parent–child interactions in both
at the end of the 1984 school year. A total of 1,161 the lab and at home. Over a 6-year period, every
boys were rated. The sample was reduced to 1,037 second year the OBS boys participated in the
participants by creating a homogenous sample of following activities: (a) they spent one half of a
French-speaking children, whose parents were born day with their families in the university laboratories
in Canada, and by eliminating subjects who refused to participate in a series of tests and observation
to participate or could not be traced. The sample sessions; (b) families were visited during evenings for
was further reduced to 895 boys, because a observations in the home setting; (c) each boy was
characteristic of the boys who received random observed at school for half a day on four occasions;
assignment in the intervention was that their parents and (d) each boy spent a whole day in the university
had an educational level of 14 years or less. To laboratories during the summer. In contrast, the
create a matched (i.e., homogenous) comparison CONT boys were only followed through question-
sample of boys who were not randomly assigned in naires during the intervention period. They were
the intervention—but answered yearly question- then met once a year at school to fill in ques-
naires for PA, RA, and the mediator variables— tionnaires that included the self-reported PA and
the sample was restricted to only youth whose RA scales.
parents had an educational level of 14 years or less Analyses revealed that the boys in the OBS
(i.e., 895 boys). and CONT groups did not significantly differ in
The boys were assessed by their kindergarten means on any variable measured at pre- or post-
teacher using the Social Behavior Questionnaire test. Therefore, it was decided to collapse the
[SBQ; Tremblay et al., 1991]. The SBQ contained 29 CONT and OBS boys into one group (i.e., CONT
items grouped into four components: disruptive (ten group) to increase statistical power. The rest of
items), anxious (five items), inattentive (four items), the larger sample, representing participants who
and prosocial (ten items). The disruptiveness scale obtained scores below the 70th percentile in
(a 5 .93) was subdivided into three a priori cate- disruptive behaviors (n 5 645), was kept in this
gories of behavior [Loeber et al., 1989]: fighting (3 study to serve as a ‘‘Low Risk’’ comparison
items; a 5 .88), oppositional behavior (5 items; group. The final sample used in this study con-
a 5 .84), and hyperactivity (2 items; a 5 .89) was sisted of 895 boys (INT 5 69; CONT 5 181; Low
used to identify at-risk children. Risk 5 645).
From the total sample, boys with a score above Prevention program. The prevention program
the 70th percentile (n 5 250) on the disruptiveness was implemented over a 2-year period from ages 7 to 9.
scale were considered to be at risk for later antisocial The program included two main components
behavior and dropping out of school [Tremblay (i.e., social skills training with the children and
et al., 1992; White et al., 1990]. These 250 boys and improvement of parenting skills) that were believed
their families were selected to participate in the at that time to be most likely to alter the boys’

Aggr. Behav.
Testing the Developmental Distinctiveness 131

disruptive behaviors [Kazdin, 1987]. It was expected taught how to manage family crises through
that they would become less disruptive if they problem solving and how to use negotiation
learned alternate and more appropriate social strategies in everyday situations.
behaviors through social skills training [Milan and Parents participated in an average of 17.4 sessions
Kolko, 1985]. Improvement of parenting skills (i.e., (SD 5 13.2; median 5 15). The maximum was 47.
use of positive reinforcement contingencies and Six families participated in only two training
sustained supervision) was also used as a strategy sessions. For most of the families, the number of
to reduce disruptive behaviors at home, and facil- training sessions required depended on how well the
itate the generalization and consolidation of the therapist believed the parents had mastered the
skills learned by the children at school. targeted skills. For 14 families, however, the training
Social and problem-solving skills training were ended prematurely because the parents were un-
conducted at school in small groups. A total of motivated. The boys from these families were
19 sessions were planned (ten in the first year and nine nevertheless kept in the INT group.
in the second year). On average, the boys attended Implementation assessment. At the end of
14.5 sessions (SD 5 4.7). Four trained professionals each child or parent session, the professionals
(two child-care workers, one social worker, and one responsible for the program application indicated
psychologist) conducted the sessions. In each group, whether the session had taken place and the
there were four or six teacher-nominated prosocial percentage of content delivered during the session
boys and one or two target boys. Including the relative to a pre-planned standardized content.
prosocial boys in the sessions served two purposes. More than 85% of the children who participated
First, they were positive models and reinforcement in the intervention attended at least two-thirds of the
agents. Second, their presence allowed the target social skills training sessions. For the parents who
children to participate without being stigmatized by participated in the intervention, the number of
classmates. The school-based training sessions took sessions varied greatly. Despite variation, more than
place once every other week between November and 75% of the parents covered at least two-thirds of the
April for two consecutive years. Each session lasted content and objectives of the parent training
approximately 45 min. Verbal instructions, positive component. In addition, child sessions were video-
reinforcement, modeling, and behavioral rehearsal taped and parent sessions were audiotaped; these
were used to teach the specific skills to the target boys. tapes were used by the program coordinator to give
Parent training was adapted from the program weekly feedback to each of the professionals and
developed by the Oregon Social Learning Center maintain the standardization of the program.
[Patterson et al., 1975]. The same four professionals,
who conducted the social and problem-solving
Measures
sessions at school, conducted the parent training
sessions in the boys’ homes. However, to stimulate Annual assessments of PA and RA (ages
teamwork among the professionals, each family had 13–17)
different professionals for parent training and for Self-reported PA and RA. In keeping with the
social skills training. Parents were first taught to existing studies on PA and RA, the boys completed
recognize, observe, and record their children’s three RA and three PA items used by Dodge and Coie
problem behaviors. Next, they were taught to define [1987]. These items were translated to French.
appropriate behaviors and to set clear objectives for Following Little [2003], self-reports were used instead
their child. Third, they learned how to use verbal of teacher ratings. Self-reports in adolescence are
and material reinforcement in a systematic and often preferred to teacher ratings for two reasons.
contingent manner to favor the child’s acquisition of First, self-reports show significant association to
appropriate behaviors. Parents also learned to official court records of aggressive acts [Loeber and
punish inappropriate behavior systematically but Farrington, 2000]. Second, adolescents are unsuper-
moderately with short time-out periods. vised by teachers and parents during the times in
Response–cost strategies involving the use of which they are most likely to be involved in aggressive
naturally occurring consequences for inappropriate and delinquent behaviors [Education, 2000].
behavior were also used (i.e., if the child broke The three proactive items were: ‘‘I have encour-
something that did not belong to him, he had to aged other kids to attack someone I do not like’’;
replace it). Parents were encouraged to supervise ‘‘I have used physical strength (or I have threatened
their children’s schoolwork and monitor their child’s to do so) to dominate other kids’’; and ‘‘I have
behavior outside the home. Finally, parents were threatened or attacked others to obtain what

Aggr. Behav.
132 Barker et al.

I want.’’ The three reactive items were: ‘‘I have Delinquent attitudes. When subjects were 12
accused others of having started the fight and I have years old, they received the Jesness Inventory
said it’s never my fault’’; ’’When another kid picks [Jesness, 1983], a standardized measure for assessing
on me or threatens me, I have become angry easily personality and delinquent profiles. We selected five
and have hit the other one’’; and ’’When someone items we thought would most closely associate to
hits me I get angry and start fighting.’’ The PA, or delinquent attitudes. These items were:
questionnaire was administered at 13, 14, 15, 16, ‘‘Most people can be cheated,’’ ‘‘It is Ok to trick
and 17 years of age. If a participant had dropped out someone,’’ ‘‘Winning a fight is enjoyable,’’ ‘‘The
of school (n 5 112 by the age of 17), the question- way to solve a problem is to cheat,’’ and ‘‘It is Ok to
naire was administered at his home. a for PA steal from the rich’’ (a 5 .59). Delinquent attitudes
and RA ranged from .60 to .77, and .65 to .70, at age 11 and 12 were summed to provide a single
respectively. index for the mediation analysis.
Assessments of the putative mediators (ages Analyses. Analyses of the prevention program
11 and 12) were conducted with respect to an ITT strategy
Peer-rated popularity. For a proxy of peer [Fishman, 2000; Kendall et al., 1999]. The key point
relationships for RA, i.e., peer rejection, we examined of ITT is to use the randomized individuals in the
perceived popularity of the boys as assessed by their groups to which they were assigned, regardless of
peers. The Pupil Evaluation Inventory [PEI; Pekarik subsequent refusal to participate or attrition. The
et al., 1976] was used to gather classmates’ assess- primary advantage of this approach is the max-
ments when the boys were 11 and 12 years old. The imization of the internal validity of the intervention;
PEI contains 34 short behavior descriptions grouped ITT assures equivalence of groups generated by the
into three scales: aggressiveness disturbance (20 random assignment of participants.
items), social withdrawal (9 items), and likeability The analyses consisted of three separate steps.
(5 items). Classmates nominated up to four boys in First, we identified the best fitting trajectory models
the classroom who best fit each behavior descriptor. for PA and RA. Second, we compared the trajec-
A code number was assigned to each boy in the tories followed by the INT group, the CONT group,
classroom and these were presented in a roster format and the Low Risk boys. Third, where intervention
to the children. PEI assessment took place near the effects were identified, we examined whether or not
end of the school year. The likeability scale was used these effects were indirectly related to factors
to assess peer ratings of each boy’s popularity. hypothesized to distinctively relate to RA in the
Specifically, the total number of received nominations presence of PA factors. Owing to the fact that we
for this scale was calculated for each participant and expected directional intervention effects, we present
z-standardized within classroom to account for probabilities associated to one-way t-tests and 90%
differences in classroom size. confidence intervals (CIs) in the results section.
Affiliation with deviant peers. As part of the To identify the trajectories, we used a group-based
annual assessments at ages 11–12, participants were method described in Nagin [2005] and Nagin and
asked, ‘‘During the past 12 months, were you a Tremblay [2001]. Group-based trajectory models are
member of a gang or a group that did reprehensible widely used in developmental psychopathology and
acts?’’ Use of a single dichotomous item has a long criminology [Piquero, 2007]. A finite mixture of
tradition in criminological studies of gangs and Poisson distributions was used to identify distinctive
delinquent effects [Howell et al., 2000; Lacourse clusters of individual trajectories within the sample.
et al., 2003; Thornberry et al., 1993; Warr, 1993]. Bayesian Information Criterion (BIC) was the basis for
Coercive parenting and parental supervision. selecting the optimal model (i.e., the number of
When subjects were 11 and 12 years old, they reported trajectories for PA and RA) [D’Unger et al., 1998].
on coercive parenting (2 items, a 5 .62) and parental An important output of the group-based method is the
supervision (2 items, a 5 .76). The supervision ques- posterior probability that each individual will follow
tions included: ‘‘Do your parents know where you are the trajectories identified in the optimal model. By way
when you go out?’’ and ‘‘Do your parents know of the maximum assignment rule, the boys are assigned
whom you hang around with?’’ The coercive items to the trajectory for which they have the highest
included: ‘‘Do your parents punish you by hitting probability of following. The posterior probabilities of
you?’’ and ‘‘Do your parents punish you by calling following the different PA and RA trajectories serve as
you bad names?’’ These items were summed between the outcomes in the analyses presented below.
ages 11 and 12 to provide a single index of coercive Following the lead of Lacourse et al. [2002], we
parenting and parental supervision. used t-tests to test for significant differences in

Aggr. Behav.
Testing the Developmental Distinctiveness 133

probabilities of following a specific trajectory con- program ceased (age nine); therefore, the direct effect
ditional on the treatment conditions (i.e., CONT of the current intervention on the trajectories may be
group, INT group, and Low Risk group). To attenuated [see, e.g., Shrout and Bolger, 2002].
examine the extent to which the intervention is Accordingly, we relaxed the assumption that the
associated with a reduction in coercive parenting X-Y path must be significant, but evaluated this
and an increase in positive peer relations, which in path by way of the magnitude of the coefficient
turn increased the likelihood for the intervention [Shrout and Bolger, 2002], and the presence of a
boys to follow low and/or desisting trajectories of significant indirect pathway as measured by the Sobel
RA, we tested for indirect effects. test [MacKinnon et al., 2002; Sobel, 1982].
As explained in the Measures section, the inter- To correct for potential skew in the distributions of
mediate variables consist of factors (ages 11 and 12) the indirect pathways (the small sample size), we used
hypothesized to relate to PA (affiliation with deviant 1,000 bootstrap resamples with replacement- and
peers, parental supervision, delinquent attitudes) bias-corrected CIs [see, e.g., Preacher and Hayes,
and those believed to relate to RA (peer-rated 2004]. Bootstrapping is a non-parametric approach
popularity, coercive parenting). Hence, for the to effect–size estimation and hypothesis testing, and
purpose of this study, we would expect the INT makes no assumptions about the shape of the
boys to have lower rates of unpopularity and distributions of the variables or the sample distribu-
coercive parenting, compared with the CONT boys, tion of the statistic [Efron and Tibshirani, 1993]. All
but no differences between the INT and CONT on analyses were performed in SAS [SAS Institute,
the putative mediators for PA. 2001]. Missing data were handled through a single
We tested for indirect effects rather than mediation imputation strategy using SAS PROC MI.
[e.g., Baron and Kenny, 1986; see MacKinnon et al.,
2002; Preacher and Hayes, 2004; Shrout and Bolger,
2002]. Each approach examines the extent to which RESULTS
an intermediary variable (M) may explain the
Descriptive Results
relationship between a predictor (X) and an outcome
(Y). In an explanation of mediation, Baron and Table I contains means and standard deviations of
Kenny [1986] stated that the first step is to establish PA and RA, and the mediator variables for INT,
that there is an effect to be mediated (i.e., a significant CONT, and Low Risk boys. In general, PA and RA
relationship between X and Y). In our case, however, appear to increase and peak at about 14 years of age
the measurement of the PA and RA (age 13) and desist thereafter. For the mediators, the INT and
trajectories began four years after the intervention CONT boys were lower than the Low Risk boys in

TABLE I. Means and Standard Deviations

Variable INT M (SD) CONT M (SD) Low risk M (SD) Significant differences

Proactive aggression
Age 13 0.59 (1.19) 0.55 (1.06) 0.53 (1.00) ns
Age 14 0.73 (1.38) 0.72 (1.27) 0.54 (1.09) C, INT4LR
Age 15 0.68 (1.80) 0.90 (1.56) 0.50 (1.14) C4INT4LR
Age 16 0.37 (0.98) 0.72 (1.29) 0.33 (0.80) C4INT, LR
Age 17 0.30 (0.61) 0.54 (1.23) 0.30 (0.88) C4INT, LR
Reactive aggression
Age 13 0.98 (1.52) 1.29 (1.42) 0.99 (1.34) C4INT, LR
Age 14 1.42 (1.88) 1.18 (1.45) 1.01 (1.37) INT4C4LR
Age 15 0.94 (1.71) 1.32 (1.72) 0.84 (1.40) C4INT4LR
Age 16 0.47 (1.01) .95 (1.59) 0.61 (1.06) C4LR4INT
Age 17 0.50 (1.23) .85 (1.32) 0.50 (1.05) C4INT, LR
Mediators (ages 11, 12)
Peer-rated popularity .72 (1.38) .69 (1.33) .02 (1.16) LR4C, INT
Coercive parenting 8.09 (3.24) 8.77 (3.04) 8.02 (2.70) C, INT4LR
Deviant peer affiliation 2.75 (1.37) 2.99 (1.39) 2.56 (1.01) C, INT4LR
Parental supervision 13.81 (2.53) 13.43 (2.48) 14.81 (2.21) LR4C, INT
Delinquent attitude 1.55 (1.01) 1.75 (1.34) 1.57 (1.00) C4INT, LR

Mean difference at a Po.05; ns, no significant differences between groups; INT, intervention group; C, control group; LR, Low Risk
comparison.

Aggr. Behav.
134 Barker et al.

peer-rated popularity and parental supervision; other- improve with additional groups. Figure 1 presents
wise, the INT and Low Risk boys self-reported similar the shape and size of the three respective trajectories
levels of coercive parenting, deviant peer affiliation, for PA and RA. Not surprisingly, these trajectories
and delinquent attitude. Table II contains the are highly similar to those originally reported by
correlations of the age 11 and 12 mediators to PA Barker et al. [2006]. For PA, 54.4% engaged in
and RA by year for the INT and CONT boys. The stable and ‘‘Low’’ levels, 33.6% were stable and
first two columns contain the mediators hypothesized ‘‘Moderate,’’ and the remainder ‘‘Peaked’’ in PA at
to relate to RA—peer-rated popularity and coercive 15 years of age. For RA, 33.9% were ‘‘Low,’’ 52%
parenting. Notice that peer-rated popularity was not were ‘‘Moderate,’’ and 15% ‘‘Peaked’’ at 15 years of
positively related to RA for the INT boys or the age, and thereafter desisted.
CONT boys. Coercive parenting was significantly
correlated to RA more so than PA for the INT boys, Step 2: Intervention Effects on PA and RA
whereas the CONT boys showed the opposite
pattern—coercive parenting was related to PA more Figure 2 contains the mean probabilities for the
than RA. Deviant peer affiliation, parental super- INT, CONT, and Low Risk boys to follow the PA
vision, and delinquent attitudes, factors believed to be and RA trajectories. Group differences were sig-
unique to PA, are located in the last three columns in nificant only for Low RA (t(196) 5 2.49, po.01;
Table II. PA and RA, however, are significantly Cohen’s d 5 0.38), but not for PA trajectories. These
related to these three variables. Nevertheless, as would results have supported our first and second hypoth-
be expected, parental supervision is inversely related eses: (1) boys who participated in the intervention
to PA and RA, whereas deviant peer affiliation and were more likely to follow low levels of RA, and (2)
delinquent attitudes are positively related. intervention effects were not identified for PA.

Step 1: Trajectories of PA and RA Step 3: Potential Mediators of Intervention Effects


For both PA and RA, three group models were To examine the potential mediators of the inter-
chosen as the optimal model. The BIC scores did not vention effect on the Low RA trajectory, we

TABLE II. Correlations Between PA and RA and Mediators for the Intervention and Control Boys

Behavior Peer-rated popularity Coercive parenting Deviant affiliation Parental supervision Delinquent attitudes

Intervention
Proactive aggression
Age 13 .05 .22 .02 .24 .33
Age 14 .10 .29 .27 .33 .34
Age 15 .06 .20 .42 .24 .24
Age 16 .17 .21 .34 .15 .56
Age 17 .05 .06 .30 .06 .24
Reactive aggression
Age 13 .06 .31 .18 .28 .29
Age 14 .08 .39 .39 .30 .28
Age 15 .01 .31 .38 .36 .34
Age 16 .15 .15 .59 .18 .42
Age 17 .06 .05 .50 .12 .37
Control
Proactive aggression
Age 13 .13 .17 .20 .32 .24
Age 14 .01 .07 .21 .26 .33
Age 15 .01 .14 .21 .22 .24
Age 16 .10 .23 .23 .24 .27
Age 17 .14 .18 .40 .12 .44
Reactive aggression
Age 13 .17 .12 .15 .25 .37
Age 14 .02 .15 .34 .19 .36
Age 15 .07 .01 .27 .22 .26
Age 16 .03 .19 .30 .27 .21
Age 17 .10 .12 .23 .06 .20
Po.05; Po.01; Po.001.

Aggr. Behav.
Testing the Developmental Distinctiveness 135

3 0.6
2.5
0.5
2

1.5

Mean Probability
0.4
1 Intervention
0.5 0.3 Control
0 Low Risk
13 14 15 16 17 0.2
PA
Low (54.4%) Moderate (33.6%) Peaking (12.0%) 0.1

0
3.5
Low Moderate Peaking
3
2.5 PA
2 0.6
1.5
1 0.5
0.5
0 Mean Probability 0.4
13 14 15 16 17 Intervention
RA 0.3
Control
Low (33.9%) Moderate (51.5%) Peaking (14.6%) Low Risk
0.2
Fig. 1. Trajectories of PA and RA.
0.1

compared the INT and CONT boys (1 5 INT, 0


0 5 CONT) and excluded the Low Risk boys; these Low Moderate Peaking
individuals did not participate in the intervention RA
and their inclusion would contaminate the relation-
Fig. 2. Mean Probabilities of Trajectory Group Membership for PA
ships between the intervention, the outcome (i.e., the
and RA.
probability of following the Low RA group), and
the mediators.
A probability, when used as a dependent vari- this trajectory. Coercive parenting is the sole
able (i.e., Y), often does not satisfy the assump- variable with significant a and b paths. The third
tion of a linear relationship with the predictor and fourth parts of Table III contain the total effects
variables (i.e., X–Y) necessary for ordinary least (Path c) and the direct effects of the Intervention on
square regression analysis [Tabachnick and Fidell, RA (Path c0 ). Notice the total effect is not
2001]. To ensure that the outcome—the posterior significant, but the magnitude of the X-Y path
probability of following Low RA—met the assump- (0.12) can be considered a low effect size [Cohen,
tion of linearity, we logit-transformed this prob- 1988] and is potentially important [e.g., McCartney
ability (ln(prLowRA)/(1prLowRA)) and tested for and Rosenthal, 2000]. The Sobel test for the total
non-linearity. Non-linearity was not identified. indirect effect (c–c0 ) was significant (see Table III).
The first part of Table III contains the standar- A reduction in coercive parenting (b 5 .049; 90%
dized effects of the intervention on the age 11 and 12 CIs 5 0.010–0.219) increased the probability of
variables (Path a). Notice that the only significant following the Low RA trajectory for the INT boys.
effect was for coercive parenting. The INT boys self-
reported lower levels of coercive parenting com-
pared with the CONT boys. The second part of
DISCUSSION
Table III contains the direct effects of the age 11 and
12 variables on the probability of following the Low The aim of this research was to use a preventive
RA trajectory (Path b). Increasing levels of deviant experiment nested in a longitudinal study to test
peer affiliation, parental coercion, and delinquent developmental and mediation hypotheses about PA
attitudes predicted lower probabilities of following and RA. We reestimated group-based trajectories

Aggr. Behav.
136 Barker et al.

TABLE III. Indirect Effects of the Intervention on Low RA

Effects of intervention on indirect variables a path b t-Value 90% CI: Lower 90% CI: Upper Prob-t (one-way)

Popularity .145 0.910 .117 .407 .182


Gang affiliation .040 0.242 .311 .231 .404
Parental coercion .336 2.062 .605 .067 .020
Parental supervision .133 0.835 .131 .397 .202
Delinquent attitudes .093 0.595 .350 .164 .277
Effects of indirect variables on Low RA
b path
Popularity .090 1.337 .021 .201 .092
Gang affiliation .208 3.017 .322 .094 .002
Parental coercion .145 2.234 .252 .038 .014
Parental supervision .056 0.808 .060 .172 .210
Delinquent attitude .331 4.588 .450 .212 .000
Total effects of intervention on Low RA
c path .121 0.730 .151 .393 .233
Direct effects of the intervention on Low
RA, controlling for mediators
c’ path .012 0.085 .236 .260 .466
Total indirect effects of intervention on
Low RA
c– c0 path .108 3.085 .050 .166 .005

from an earlier study [i.e., Barker et al., 2006], with a Having shown that disruptive kindergarten boys
sample that allowed for concrete comparisons of a who received the intervention were more likely to be
subgroup of targeted disruptive boys ran- low in RA throughout adolescence, we then tested
domly assigned to an INT or CONT condition to whether or not the intervention effects would be
the rest of the sample of lower-disruptive boys indirectly affected by social and behavioral factors
used as a Low Risk comparison group. We used an in early adolescence. Factors hypothesized to be
ITT approach to maximize the internal validity unique to the development of RA (peer relations,
of these comparisons. Similar to Barker et al. coercive parenting) and PA (deviant peer affiliation,
[2006], we found that both PA and RA were best parental supervision, delinquent attitudes) at age
described by three trajectories. The majority of boys 11 and 12 were assessed. In analysis of program
followed a Low stable PA group, followed by effectiveness, an identified intermediary variable can
moderate stable and high peaking groups. For RA, increase the confidence of its putative causal
on the other hand, the boys followed moderate contribution, provided its rival intermediaries are
desisting patterns, followed by low stable and high eliminated [Howe et al., 2002]. We found that the
peaking groups. probability of following the Low RA trajectory was
To test our first hypothesis, we tested whether or indirectly related to a reduction in coercive parent-
not a preventive intervention targeting factors, ing for the INT boys. Factors usually related to PA
believed to be causal in the development of RA, did not mediate the intervention effects to RA. This
would positively deflect the development of the RA finding provides support for Dodge’s [1997] devel-
trajectory, between 13 and 17 years of age. Our opmental model, and supports earlier research
results showed that compared with the CONT boys relating coercive parenting to RA, but not PA, by
for the INT boys, the intervention (age 9) managed a way of retrospective links [Connor et al., 2004;
small effect size (.12) on the probability to follow the Dodge et al., 1997] and prospective links [Vitaro
Low RA trajectory (beginning at age 13). Despite et al., 2006a; Xu et al., 2009].
the distant temporal relation, this small effect was Our finding that a prevention program that
nevertheless sufficient to make the INT boys primarily targets factors believed to associate with
virtually indistinguishable from the Low Risk RA did not also affect PA may support and extend
counterparts with respect to RA. No significant an intervention by Phillips and Lochman [2003]. In a
intervention effects were found for PA (second short-term follow-up, these researchers assessed the
hypothesis). It is important to note; however, a effectiveness of two behavior manipulations—a
trend for intervention effects was clearly evident in reactive anger management group and a proactive-
PA, and will be revisited below. instrumental group—on 10–12 year old boys.

Aggr. Behav.
Testing the Developmental Distinctiveness 137

Results indicated that the proactive-instrumental the child, the sex of the parent, and the ethnic and
manipulation significantly lowered aggressive re- socioeconomic background of the family [Keenan
sponses for both PA and RA, whereas the reactive- and Shaw, 1997; Lamb and Oppenheim,
anger manipulation reduced only RA behaviors. 1989].Therefore, replications are needed with non-
Phillips and Lochman [2003] raised the question of low SES samples, fathers and mothers, female
whether or not the reward component of aggressive children and adolescents, and with diverse ethnic
behavior, thought to be exclusive to PA, may have a groups in rural and urban settings to verify the
universal appeal for aggressive children in general. generalizability of the results. Fourth, a relatively
From this perspective, an intervention, aimed at small group of boys were submitted to the interven-
improvements in parenting and the social skills of tion. This not only resulted in low statistical power,
children, may not have dealt sufficiently with the but also prevented more sophisticated analyses, such
reward component of aggressive behavior that as the joint trajectory technique used by Barker and
appears related to both PA and RA. However, an colleagues [2006] on the larger sample, as well as the
alternate explanation for the lack of an intervention ability to describe whether or not the intervention
effect on PA in this study is one of statistical power. affected the actual shapes of the trajectories (i.e.,
Indeed, because the trend of the results is in the INT boys vs. the CONT boys). Fifth, the scales used
same direction for PA and RA, we may not have to examine proactive and RA were low in internal
had the statistical power to detect all the positive consistency and, therefore, may have affected these
impacts of the intervention on PA, especially results. Finally, there is also good evidence that
because PA is more rare than RA (i.e., lower mean constellations of behaviors analogous to those
levels in the counterpart trajectories), and therefore, measured by Dodge and Coie’s PA and RA items
offers less variability for analysis compared with RA can be traced to early childhood [Hay, 2005; Hay
[e.g., see Vitaro and Brendgen, 2005]. et al., 2000; Tremblay, 2003; Vitaro et al., 2006b]. To
fully understand the development of these two types
of aggression and their transformation by preventive
Limitations
interventions, we need studies that trace the devel-
This study had a number of limitations. First, the opment of these two types of aggression from early
longitudinal data that were used to estimate childhood to adulthood.
trajectories of PA and RA were based on self- Despite these limitations, this study used four
reports, as were four out of the five age 11 and 12 methodological innovations to test the development
mediators. This brings forward shared method of PA and RA in highly aggressive children: (1) a
variance. Second, although we examined five theo- preventive experimental intervention was nested
retically distinct mediators of PA and RA, we could within a longitudinal study, (2) an intent-to-treat
not assess whether or not the intervention effect for approach was utilized to maximize the internal
following the Low RA trajectory was mediated by validity of the protocol of random assignment for
changes in omitted but related variables, such as a the intervention, (3) developmental trajectories
reduction in hostile attribution biases, perhaps the were employed to examine the impact of the
most robust correlate of RA [Dodge et al., 1997; intervention on distinct subgroups of children
Dodge and Coie, 1987; Hubbard et al., 2001; Nas defined by different developmental trajectories in
et al., 2005; Orobio de Castro et al., 2002). In the longitudinal study, and (4) we tested whether or
addition, the item we used to tap deviant peer not identified intervention effects could be mediated
affiliations was measured affiliations to a ‘‘gang’’ or by putative factors distinct to the development of
a ‘‘group’’ of peers, and captures just a share of all RA and PA. An intervention effect was identified in
types of deviant peer affiliations. Third, although RA that was indirectly related to a reduction in self-
the intervention evidenced a significant long-term reported coercive parenting. These results add
impact on the development of RA, it was applied to support to Dodge’s dual model with regard to the
White, French-speaking boys from a low SES urban development of PA and RA, and shed new light on
environment, particularly at risk for the develop- the specific role of parent–child relations. Given that
ment of RA [e.g., Agnew, 1992], given the high levels parent–child relations served a mediating role in the
of coercive or inconsistent parenting in lower SES context of a preventive intervention, its putative
communities [Bornstein et al., 2003; Caughy and causal role is substantially increased in comparison
Franzini, 2005]. Differences in coercive parenting or to past correlational studies, although it cannot be
its impact on children’s behaviors appear to exist definitively ascertained, given the multimodal nature
[Pinderhughes et al., 2000], depending on the sex of of the preventive intervention used in this study.

Aggr. Behav.
138 Barker et al.

Furthermore, preventive experimental interventions Brown EJ, Kolko DJ. 1999. Child victims’ attributions about being
in early childhood are needed to test causal models physically abused: An examination of factors associated with
for the development of aggression, because chronic symptom severity. J Abnorm Child Psych 27:311–322.
Caughy MO, Franzini L. 2005. Neighborhood correlates of cultural
aggressive trajectories start in early childhood differences in perceived effectiveness of parental disciplinary
(Tremblay, in press). tactics. Parent-Sci Pract 5:119–151.
Cohen J. 1988. Statistical Power Analysis for the Behavioral Sciences
(2nd edition). Hillsdale, NJ: Lawrence Erlbaum.
ACKNOWLEDGMENTS Connor DF, Steingard RJ, Cunningham JA, Anderson JJ, Melloni
RH. 2004. Proactive and reactive aggression in referred children
This research has been supported by Québec’s and adolescents. Am J Orthopsychiat 74:129–136.
FQRSC (Tremblay, Vitaro, Lacourse) and FRSQ Crick NR, Dodge KA. 1996. Social information-processing mechan-
isms on reactive and proactive aggression. Child Dev 67(3):
(Tremblay, Vitaro) research funding programs, 993–1002.
Canada’s Social Sciences and Humanities Research Day DM, Bream LA, Pal A. 1992. Proactive and reactive aggression:
Council (Tremblay, Vitaro) and Canadian Institutes An analysis of subtypes based on teacher perceptions. J Clin
for Health Research (Tremblay, Vitaro), the U.S. Child Psychol 21:210–217.
National Science Foundation (SES-9911370), the de Castro BO, Merk W, Koops W, Veerman JW, Bosch JD. 2005.
Emotions in social information processing and their relations
U.S. National Institute of Mental Health (RO1
with reactive and proactive aggression in referred aggressive
MH65611-01A2). boys. J Clin Child Adolesc 34:105–116.
Dodge KA. 1991. The structure and function of reactive and
proactive aggression. In: Pepler DJ, Rubin KH (eds.). The
Development and Treatment of Childhood Aggression. Hillsdale,
NJ: Lawrence Erlbaum Associates, pp 201–219.
REFERENCES
Dodge KA, Coie JD. 1987. Social-information-processing factors in
Agnew R. 1992. Foundation for a general strain theory of crime and reactive and proactive aggression in children’s peer groups. J Pers
delinquency. Criminology 30:47–88. Soc Psychol 53:1146–1158.
Akers RL. 1998. Social Learning and Social Structure: A General Dodge K, Lochman JE, Harnish JD, Bates JE, Pettit GS. 1997.
Theory of Crime and Deviance. Boston: Northeastern University Reactive and proactive aggression in school children and
Press. psychiatrically impaired chronically assaultive youth. J Abnorm
Bandura A. 1973. Aggression: A Social Learning Analysis. Psychol 106:37–51.
New York: Holt. Dollard J, Miller NE, Doob LW, Mowrer OH, Sears RR. 1939.
Barker ED, Tremblay RE, Nagin DS, Vitaro F, Lacourse E. 2006. Frustration and aggression. New Haven: Yale University Freer.
The development of male proactive and reactive aggression in D’Unger A, Land K, McCall P, Nagin D. 1998. How many latent
adolescence. J Child Psychol Psyc 47:783–790. classes of delinquent/criminal careers? Results from mixed
Baron RM, Kenny DA. 1986. The moderator-mediator variable Poisson regression analyses of the London, Philadelphia, and
distinction in social psychological research: Conceptual, strategic, Racine cohorts’ studies. Am J Sociol 103:1593–1630.
and statistical considerations. J Pers Soc Psychol 51:1173–1182. Education U.S.D.o. 2000. Working for Children and Families: Safe and
Berkowitz L. 1962. Aggression: A Social Psychological Analysis. Smart After-School Programs. Washington, DC: Education U.S.D.o.
New York: McGraw-Hill. Efron B, Tibshirani RJ. 1993. An Introduction to the Bootstrap.
Berkowitz L. 1989. Frustration-aggression hypothesis: Examination New York: Chapman & Hall.
and reformulation. Psychol Bull 106:59–73. Ellis ML, Weiss B, Lochman JE. 2009. Executive functions in
Bierman KL, Coie JD, Dodge KA, Greenberg MT, Lochman JE, children: Associations with aggressive behavior and appraisal
McMahon RJ, Pinderhughes EE, 2002. Using the fast track processing. J Abnorm Child Psych. 37(7):945–956.
randomized prevention trial to test the early-starter model of the Fishman DB. 2000. Transcending the efficacy versus effectiveness
development of serious conduct problems. Dev Psychopathol debate: Proposal for a new electronic ‘‘journal of pragmatic case
14:925–943. studies.’’ Prevention and Treatment 3:1–23.
Boisjoli R, Vitaro F, Lacourse E, Barker ED, Tremblay RE. 2007. Fite PJ, Colder CR, Lochman JE, Wells KC. 2008. Developmental
Impact and clinical significance of a preventive intervention for trajectories of proactive and reactive aggression from fifth to
disruptive boys: 15 year follow-up. Br J Psychiatry 191:415–419. ninth grade. J Clin Child Adolesc 37:412–421.
Bornstein M, Hahn C, SuwalskyJ, Haynes OM. 2003. Socio- Fite PJ, Stoppelbein L, Greening L. 2009. Proactive and reactive
economic status, parenting, and child development: The Hol- aggression in a child psychiatric inpatient population. J Clin
lingshead Four-Factor Index of social status and the Socio- Child Adolesc 38:199–205.
economic Index of occupations. In: Bornstein M, Bradley R Fossati A, Raine A, Borroni S, Volpi E, Maffei C, Santalucia I,
(eds.). Socioeconomic Status, Parenting, and Child Development. Bizzozero A. 2009. A cross-cultural study of the psychometric
Mahwah, NJ: Lawrence Erlbaum Associates, pp 29–82. properties of the reactive and proactive aggression questionnaire
Brendgen M, Vitaro F, Tremblay RE, Lavoie F. 2001. Reactive and among Italian nonclinical adolescents. Psychol Assessment
proactive aggression: Predictions to physical violence in different 21:131–135.
contexts and moderating effects of parental monitoring and Gottefredson M, Hirschi T. 1990. A General Theory of Crime.
caregiving behavior. J Abnorm Child Psych 29:293–304. Stanford, CA: Stanford University Press.
Brendgen M, Vitaro F, Tremblay RE, Wanner B. 2002. Parent and Hay D. 2005. The beginnings of aggression in infancy. In: Tremblay
peer effects on delinquency related violence and dating violence: RE, Hartup WW, Archer J (eds.). Developmental Origins of
A test of two mediational processes. Soc Dev 11:225–244. Aaggression. New York: Guilford.

Aggr. Behav.
Testing the Developmental Distinctiveness 139

Hay DF, Castle J, Davies L. 2000. Toddlers’ use of force against Nas CN, Orobio De Castro B, Koops W. 2005. Social information
familiar peers: A precursor of serious aggression? Child Dev processing in delinquent adolescents. Psychol Crime Law 11:363–375.
71:457–467. Nouvion SO, Cherek DR, Lane SD, Lieving LM, Tcheremissine OV.
Howe GW, Reiss D, Yuh J. 2002. Can prevention trials test theories 2007. Human proactive aggression: Association with personality
of etiology? Dev Psychopathol 14:673–694. disorders and psychopathy. Aggr Behav 33:552–562.
Howell JC, Egley A, Gleason DK. 2000. Youth gangs: Definitions Orobio de Castro B, Veerman JW, Koops W, Bosch JD,
and the age-old issue: Article presented at the meeting of the Monshouwer HJ. 2002. Hostile attribution of intent and
American Society of Criminology, San Francisco, CA. aggressive behavior: A meta-analysis. Child Dev 73:916–934.
Hubbard JA, Dodge KA, Cillessen AHN, Coie JD, Schwartz D. Patterson GR, Reid JB, Jones RR, Conger RR. 1975. A Social
2001. The dyadic nature of social information processing in boys’ Learning Approach to Family Intervention: Families With
reactive and proactive aggression. J Pers Soc Psychol 80:268–280. Aggressive Children (Volume 1). Eugene, OR: Castalia.
Jesness CF. 1983. The Jesness Inventory (Revised edition). Palo Alto, Pekarik EG, Prinz RJ, Liebert DE, Weintraub S, Neale JN. 1976. The
CA: Consulting Psychologist Press. pupil evaluation inventory: A sociometric technique for assessing
Kazdin AE. 1987. Treatment of antisocial behavior in children: children’s social behavior. J Abnorm Child Psych 4:83–97.
Current status and future directions. Psychol Bull 102:187–203. Pellegrini AD, Bartini M, Brooks F. 1999. School bullies, victims,
Keenan K, Shaw D. 1997. Developmental and social influences on and aggressive victims: Factors relating to group affiliation and
young girls’ early problem behaviors. Psychol Bull 121:95–113. victimization in early adolescence. JEduc Psychol 91:216–224.
Kempes M, Matthys W, de Vries H, van Engeland H. 2005. Reactive Phillips NC, Lochman JE. 2003. Experimentally manipulated change
and proactive aggression in children—A review of theory, in children’s proactive and reactive aggressive behavior. Aggr
findings and the relevance for child and adolescent psychiatry. Behav 29:215–227.
Eur Child Adoles Psy 14:11–19. Pinderhughes E, Dodge K, Bates J, Pettit G, Zelli A. 2000. Discipline
Kendall PC, Flannery-Schroeder EC, Ford JD. 1999. Therapy responses: Influences of parents’ socioeconomic status, ethnicity,
outcome research methods. In: PC Kendall, JN Butcher, beliefs about parenting, stress, and cognitive-emotional pro-
GN Holmbeck (eds.). Handbook of Research Methods in cesses. J Fam Psychol 14:380–400.
Clinical Psychology (p. 330–363). New York: John Wiley. Piquero AR. 2007. Taking stock of developmental trajectories of
Kenny DA, Kashy DA, Bolger N. 1998. Data analysis in social criminal activity over the life course. In: Liberman A (ed.). The
psychology. In: Gilbert D, Fiske ST, Lindzey G (eds.). Handbook Yield of Longitudinal Studies. New York: Springer.
of Social Psychology (Volume 1). New York: McGraw-Hill, Poulin F, Boivin M. 2000a. Reactive and proactive aggression:
pp 233–265). Evidence of a two-factor model. Psychol Assessment 12:
Lacourse E, Cote S, Nagin DS, Vitaro F, Brendgen M, Tremblay 115–122.
RE. 2002. A longitudinal-experimental approach to testing Poulin F, Boivin M. 2000b. The role of proactive aggression and
theories of antisocial behavior development. Dev Psychopathol reactive aggression in the formation and development of boys’
14:909–924. friendships. Dev Psychol 36:233–240.
Lacourse É, Nagin DS, Tremblay RE, Vitaro F, Claes M. 2003. Preacher KJ, Hayes AF. 2004. SPSS and SAS procedures for
Developmental trajectories of boys’ delinquent group member- estimating indirect effects in simple mediation models. Behav Res
ship and facilitation of violent behaviors during adolescence. Dev Meth Ins C 36:717–731.
Psychopathol 15:183–197. Price JM, Dodge KA. 1989. Reactive and proactive aggression in
Lamb ME, Oppenheim D. 1989. Fatherhood and father-child childhood: Relations to peer status and social context dimen-
relationships. In: Cath SH, Gurwitt A, Gunsberg L (eds.). sions. J Abnorm Child Psych 17:455–471.
Fathers and Their Families. Hillsdale, NJ: The Analytic Press, Pulkkinen L. 1996. Proactive and reactive aggression in early
pp 11–26. adolescence as precursors to anti- and prosocial behavior in
Little TD, Jones SM, Henrich CC, Hawley PH. 2003. Disentangling young adults. Aggr Behav 22:241–257.
the ‘‘whys’’ from the ‘‘whats’’ of aggressive behaviour. Int J Reynolds AJ, Ou S-R, Topitzes JW. 2004. Paths of effects of early
Behav Dev 27:122–133. childhood intervention on educational attainment and delin-
Loeber R, Farrington DP. 2000. Young children who commit crime: quency: A confirmatory analysis of the Chicago child-parent
Epidemiology, developmental origins, risk factors, early inter- centers. Child Dev 75:1299–1328.
ventions, and policy implications. Dev Psychopathol 12:737–762. SAS Institute. 2001. The SAS System for Windows (Version 8. 2)
McCartney K, Rosental. R. 2000. Effect Size Practical importance, [Computer software]. Cary, NC: SAS Institute.
and Socil Policy for children. Child Development 71:173–180. Salmivalli C, Nieminen E. 2002. Proactive and reactive aggression
Loeber R, Tremblay RE, Gagnon C, Charlebois P. 1989. Continuity among school bullies, victims, and bully-victims. Aggr Behav
and desistance in disruptive boys early fighting at school. Dev 28:30–44.
Psychopathol 1:39–50. Schippel PL, Vasey MW, Cravens-Brown LM, Bretveld RA. 2003.
MacKinnon DP, Lockwood CM, Hoffman JM, West SG, Sheets V. Suppressed attention to rejection, ridicule, and failure cues: A
2002. A comparison of methods to test mediation and other unique correlate of reactive but not proactive aggression in
intervening variable effects. Psychol Methods 7:83–104. youth. J Clin Child Adolesc 32:40–55.
Milan AM, Kolko DJ. 1985. Social skills training and complemen- Schwartz D, Flamant R, Lellouch J. 1980. Clinical Trials. New York:
tary strategies in anger control and the treatment of aggressive Academic Press.
behavior. In: L’Abate L, Milan MA (eds.). Handbook of Social Shrout PE, Bolger N. 2002. Mediation in experimental and
Skills Training and Research. New York: Wiley, pp 101–135. nonexperimental studies: New procedures and recommendations.
Nagin DS. 2005. Group-Based Modeling of Development Over the Psychol Methods 7:422–445.
Life Course. Cambridge, MA: Harvard University Press. Smithmyer CM, Hubbard JA, Simons RF. 2000. Proactive and
Nagin DS, Tremblay RE. 2001. Analyzing developmental trajec- reactive aggression in delinquent adolescents: Relations to
tories of distinct but related behaviors: A group-based method. aggression outcome expectancies. J Clin Child Psychol 29:
Psychol Methods 6:18–34. 86–93.

Aggr. Behav.
140 Barker et al.

Sobel ME. 1982. Asymptotic confidence intervals for indirect effects Archer J (eds.). The Developmental Origins of Physical Aggres-
in structural equation models. In: Leinhardt S (ed.). Sociological sion in Humans. New York: Guilford Press.
Methodology. San Fransisco: Jossey-Bass, pp 290–312. Vitaro F, Gendreau PL, Tremblay RE, Oligny P. 1998. Reactive and
Tabachnick BG, Fidell LS. 2001. Using Multivariate Statistics proactive aggression differentially predict later conduct
(4th edition). Boston: Allyn & Bacon. problems. J Child Psychol Psyc 39:1–9.
Thornberry T. 1998. Membership in youth gangs and involvement in Vitaro F, Brendgen M, Pagani L, Tremblay RE, McDuff P. 1999.
serious and violent offending. In: Loeber R, Farrington DP Disruptive behavior, peer association, and conduct disorder:
(eds.). Serious and Violent Juvenile Offenders: Risk Factors and Testing the developmental links through early intervention. Dev
Successful Interventions. Thousand Oaks, CA: Sage Publications, Psychopathol 11:287–304.
pp 147–166. Vitaro F, Brendgen M, Tremblay RE. 2001. Preventive intervention:
Thornberry TP, Krohn MD, Lizotte AJ, Chard-Wierschem D. 1993. Assessing its effects on the trajectories of delinquency and testing
The role of juvenile gangs in facilitating delinquent behavior. for mediational processes. Appl Dev Sci 5:201–213.
J Res Crime Delinq 30:55–87. Vitaro F, Brendgen M, Tremblay RE. 2002. Reactively and
Tremblay RE. 2003. Why socialization fails: The case of chronic proactively aggressive children: Antecedent and subsequent
physical aggression. In: Lahey BB, Moffitt TE, Caspi A (eds.). characteristics. J Child Psychol Psyc 43:495–505.
Causes of Conduct Disorder and Juvenile Delinquency. Vitaro F, Barker ED, Boivin M, Brendgen M, Tremblay RE. 2006a. Do
New York: The Guilford Press. pp 182–224. early difficult temperament and harsh parenting differentially
Tremblay RE. Developmental origins of disruptive behaviors. predict reactive and proactive aggression? Int J Behav Dev 30:12–19.
J Child Psychol Psychiatry, in press. Vitaro F, Brendgen M, Barker ED. 2006b. Subtypes of aggressive
Tremblay RE, McCord J, Boileau H, Charlebois P, Gagnon C, behaviors: A developmental perspective. Int J Behav Dev 30:12–19.
Le Blanc M, Larivee S. 1991. Can disruptive boys be helped to Vitaro F, Boivin M, Tremblay RE. 2007. Peers and violence: a two-
become competent? Psychiatry 54:148–161. sided perspective. In: Flannery DJ, Vazonsyi A, Waldman I
Tremblay RE, Masse B, Perron D, Le Blanc M, Schwartzman AE, (eds.). Cambridge Handbook of Violent Behavior. Cambridge:
Ledingham JE. 1992. Early disruptive behavior, poor school Cambridge University Press.
achievement, delinquent behavior, and delinquent personality: Warr M. 1993. Age, peers, and delinquency. Criminology 34:11–36.
Longitudinal analyses. J Consult Clin Psych 60:64–72. White JL, Moffitt TE, Earls FJ, Robins LN, Silva PA. 1990. How
Tremblay RE, Vitaro F, Nagin D, Pagani L, Seguin JR. 2003. The early can we tell? Predictors of childhood conduct disorder and
Montreal longitudinal and experimental study. In: Thornberry T adolescent delinquency. Criminology 28:507–533.
(ed.). Taking Stock of Delinquency: An Overview of Findings Xu Y, Zhang Z. 2008. Distinguishing proactive and reactive
from Contemporary Longitudinal Studies. New York: Kluwer aggression in Chinese children. J Abnorm Child Psych 36:539–552.
Academic/Plenum, pp 205–254. Xu Y, Farver JAM, Zhang Z. 2009. Temperament, harsh and
Vitaro F, Brendgen M. 2005. Proactive and reactive aggression: A indulgent parenting, and Chinese children’s proactive and
developmental perspective. In: Tremblay R.E., Hartup WW, reactive aggression. Child Dev 80:244–258.

Aggr. Behav.

You might also like