Journal of Clinical Child and Adolescent Psychology
2005, Vol. 34, No. 1, 62–73
Copyright © 2005 by
Lawrence Erlbaum Associates, Inc.
Predicting Early Adolescent Gang Involvement
From Middle School Adaptation
Thomas J. Dishion, Sarah E. Nelson, and Miwa Yasui
Child and Family Center, University of Oregon
This study examined the role of adaptation in the first year of middle school (Grade 6,
age 11) to affiliation with gangs by the last year of middle school (Grade 8, age 13).
The sample consisted of 714 European American (EA) and African American (AA)
boys and girls. Specifically, academic grades, reports of antisocial behavior, and peer
relations in 6th grade were used to predict multiple measures of gang involvement by
8th grade. The multiple measures of gang involvement included self-, peer, teacher,
and counselor reports. Unexpectedly, self-report measures of gang involvement did
not correlate highly with peer and school staff reports. The results, however, were
similar for other and self-report measures of gang involvement. Mean level analyses
revealed statistically reliable differences in 8th-grade gang involvement as a function
of the youth gender and ethnicity. Structural equation prediction models revealed that
peer nominations of rejection, acceptance, academic failure, and antisocial behavior
were predictive of gang involvement for most youth. These findings suggest that the
youth level of problem behavior and the school ecology (e.g., peer rejection, school
failure) require attention in the design of interventions to prevent the formation of
gangs among high-risk young adolescents.
terventions involving peers may actually lead to escalations in problem behavior (Dishion, McCord, &
Poulin, 1999).
The most dramatic example of the potential deleterious effects of peers on behavior is the phenomenon of
gangs. Gang involvement in adolescence can be seen
as the extreme end of a progression of deviant peer involvement. As discussed by Cairns, Cadwallader, Estell, and Neckerman (1997), gang involvement does
not materialize out of thin air. It is often part of a developmental trajectory in which children with a history
of antisocial behavior form peer groups organized in
deviant behavior. When youth spend a good deal
of time with deviant peers, and clearly identify with the
deviant nature of the group, influence is at its greatest
(Kiesner, Cadinu, Poulin, & Bucci, 2002; Osgood,
Wilson, Bachman, O’Malley, & Johnston, 1996). The
work by Thornberry (1998) is noteworthy in demonstrating that gang involvement actually amplifies
problem behavior among youth who were problematic
previously and involved with deviant peers. Longitudinal analyses reveal that even when controlling for
previous deviant peer involvement and delinquent behavior, an adolescent’s membership in a street gang
is associated with marked increases in serious delinquent behavior (Battin, Hill, Abbott, Catalano, &
Hawkins, 1998; Thornberry, Krohn, Lizotte, & ChardWierschem, 1993).
Programs designed to reduce gang involvement and
gang-related criminal activities, once the gangs are es-
Some argue that peer group socialization has the
most powerful environmental effect on young adolescent social and emotional development (Harris, 1995).
The support for this hypothesis perhaps is warranted
most when considering the role of peers in the development of antisocial behavior (Dishion & Patterson, in
press; Hartup, 1996). Without a doubt, a young adolescent’s tendency to spend time with peers who engage
in deviant behaviors is the strongest correlate of delinquency, substance use, and other forms of problem
behavior (Biglan, Brennan, Foster, & Holder, 2004;
Elliott, Huizinga, & Ageton, 1985). The relation goes
beyond a self-report bias: Direct observations of adolescents with their friends reveal that support for deviant talk within friendships predicts increases in adolescent problem behavior (Dishion, Eddy, Haas, Li,
& Spracklen, 1997; Dishion, Spracklen, Andrews, &
Patterson, 1996; Patterson, Dishion, & Yoerger, 2000).
Moreover, intervention research reveals that, in some
conditions, randomly assigning high-risk youth to inThis project was supported by Grants DA07031, DA13773, and
DA16110 to Thomas J. Dishion from the National Institute on Drug
Abuse at the National Institutes of Health.
We are deeply grateful for the hard work of the Project Alliance staff,
study families, and participating schools, without whom this study
would not be possible. Thanks to Ann Simas for editing and graphic
preparation on this article.
Requests for reprints should be sent to Thomas J. Dishion, Child and
Family Center, University of Oregon, 195 West 12th Avenue, Eugene, OR 97401–3408. E-mail: [email protected]
62
EARLY ADOLESCENT GANG INVOLVEMENT
tablished, are often unsuccessful and sometimes are
counterproductive (Klein, 1993). For this reason, it is
important to both understand and prevent early adolescent gang involvement. Studying the development and
ecology of problem outcomes can provide clues to intervention targets and risk factors (Cicchetti & Toth,
1992; Dishion & Patterson, 1999; Patterson, Reid, &
Dishion, 1992). Given the importance of the publicschool ecology for peer group formation, it is within
such settings that preventive interventions can be localized. In this research, we focused on understanding the
patterns of school adaptation that predicted movement
into gangs by the last year of middle school among a
multiethnic, urban sample of youth.
It is important to consider deviant peer and gang
involvement within a broader framework describing
the development of antisocial and problem behavior
in adolescence (e.g., Cadwallader & Cairns, 2002).
An overview of an ecological perspective on gang involvement is provided in Figure 1. This model draws
heavily from an ecological perspective on the development of child and adolescent problem behavior
(Dishion, French, & Patterson, 1995; Dishion &
Patterson, in press; Patterson et al., 1992). As noted
in Figure 1, a number of variables that are not studied
in this report can affect the development of antisocial
behavior and eventual involvement in deviant peers
and gangs. For example, the general assumption is
that the family context can be influential in the development of antisocial behavior and, indirectly, to the
child’s peer environment (Dishion, 1990). Conditions
such as parent psychopathology, substance use, and
marital distress can disrupt parenting practices. This
disruption of parenting can affect the development of
antisocial behavior and deviant peer involvement. In
addition, communities and schools can be influential
in promoting peer dynamics conducive to gang involvement and antisocial behavior (Crone & Horner,
2003; D. Gottfredson, 1990). Poverty can disrupt
families and change developmental trajectories
(McLoyd, 1990), and neighborhoods can promote de-
Figure 1. An ecological framework for the development of early
adolescent gang involvement.
viant peer involvement and problem behavior
(Gorman–Smith & Tolan, 1998; Simcha-Fagan &
Scwhartz, 1986).
Of most interest, however, is adaptation to school
and the emergence of gang involvement in early adolescence. The simplest explanation for the emergence
of deviant peer groups is that antisocial youth “shop”
for other children who are inclined similarly, in a
process often referred to as homophily (Cairns,
Cairns, Neckerman, Gest, & Gariépy 1988; Dishion
et al., 1995; Lahey, Gordon, Loeber, StouthamerLoeber, & Farrington, 1999). Several studies now
confirm the mutual attraction of antisocial youth into
social networks and friendships (Hartup, 1996; Kandel, 1986; Poulin & Boivin, 2000; Vitaro, Tremblay,
Kerr, Pagani, & Bukowski, 1997). The important issue is whether any school experiences amplify the
natural tendency for high-risk youth to coalesce into
groups.
There are data in support of the hypothesis that
school failure in the domains of academics and peer
relationships increases the likelihood of young adolescents seeking the company of deviant peers (Craig,
Vitaro, Gagnon, & Tremblay, 2002; Patterson &
Dishion, 1985; Patterson et al., 1992). Dishion, Patterson, Stoolmiller, and Skinner (1991) examined
whether peer rejection (as measured by peer nominations) and academic failure at age 9 to 10 accounted for
deviant peer involvement at age 11 to 12 in boys. Controlling for the boys’ antisocial behavior, both peer rejection and academic failure accounted for variation in
deviant peer involvement. On the other hand, focusing
on their longitudinal sample of New Zealand adolescent youth, Fergusson and Horwood (1999) did not
find a direct effect of academic achievement and peer
rejection on later deviant peer affiliation, once controlling for youth conduct problems and parent and family
factors. The measure of peer rejection, however, was
based on parent and teacher reports, which are correlated only moderately with peer nomination methods
of peer rejection.
One of the major issues to confront us in the study
of the development and ecology of gang involvement is
that gangs often are predicated on the ethnicity and
gender of their members (Maxson & Klein, 1995) and
often are more prevalent among male adolescents
within the minority culture (Lahey et al., 1999).
Sampson and Laub (1994) made the same point when
considering the role of community ecology in explaining the high rates of delinquency among Irish youth in
a longitudinal sample of youth and families conducted
in the 1930s.
Establishing the link between school failure, problem behavior, and increased likelihood of future gang
involvement is important for two reasons. Theoretically, such findings are consistent with the general
proposition that aggregation among deviant peers is a
63
DISHION, NELSON, YASUI
functional adaptation for high-risk youth (Dishion,
Poulin, & Medici Skaggs, 2000; Dishion, Nelson, &
Bullock, 2004). Youth who do poorly in publicschool settings adapt by forming groups that uphold a
“deviant” value system that has its own short- and
long-term benefits. Second, identification of school
adaptation in addition to general levels of problem
behavior leads the way to expanding the focus of prevention of delinquency, violence, and other forms of
problem behavior to target specific school experiences that may give rise to the proximal cause,
namely, aggregation into deviant peer groups and
gangs.
To establish the importance of school adaptation in
gang formation, it is important to control for antisocial
behavior. There has yet to be a longitudinal analysis of
gang involvement in early adolescence that has integrated sociometric assessments, academic grades, and
antisocial behavior in the first year of middle school as
predictors of gang involvement by the last year of middle school. This research strategy is potentially useful
for the design and evaluation of intervention programs
to reduce gang involvement within a public-school
environment.
To date, one of the shortcomings of the research on
gang involvement is that investigators often rely exclusively on brief, self-report measures. Although self-reported gang involvement is a reasonable place to start,
it is important to consider the possibility of a multiagent, multimethod measure. One advantage of employing multiple measurements of gang involvement is
that they allow evaluation of the construct viability, as
well as the potential for disentangling measurement
bias from patterns of findings (Bank, Dishion, Skinner,
& Patterson, 1990).
This study focuses on sixth-grade peer relationships, academic adaptation, peer relations, and antisocial behavior in relation to early-adolescent gang involvement. This community sample was drawn from
an ethnically diverse metropolitan area, representing
all sixth-grade students in targeted middle schools (N =
999). In the sixth grade, we assessed the students’
sociometric status, antisocial behavior, and academic
grades. In the eighth grade, we conducted a multiagent,
multimethod assessment of gang involvement, based
on teacher, school counselor, peer, and self-reports.
Specifically we hypothesized that:
1. Peer nominations of peer relations and records
of academic grades in the sixth grade would predict
gang involvement in the eighth grade, after controlling
for sixth-grade problem behavior.
2. The model for early gang involvement would be
essentially the same for European American (EA) and
African American (AA) boys and girls. Mean-level differences in early adolescent gang involvement would be
64
Table 1. Demographics for Project Alliance Cohorts 1 and
2 European American and African American Subsample
European
American
African
American
Gender
%
n
%
n
Total
N
Male
Female
Total
31.8
27.4
59.2
227
196
423
22.0
18.8
40.8
157
134
291
53.8
46.2
100.0
384
330
714
explained by mean-level differences in risk factors associated with early gang involvement.
Methods
Participants
Participants included 714 adolescents and their
families, recruited in sixth grade from three middle
schools within an ethnically diverse metropolitan community in the Northwest region of the United States.
This subsample, self-identified as either EA or AA,
was taken from a larger community sample. All sixthgrade students were approached for participation. The
sample of 714 represents a 95% recruitment rate for
Cohort 1 and an 83% recruitment rate for Cohort 2;
85% of Cohort 1 and 86% of Cohort 2 participated in
all assessment waves (from Grades 6 to 9). Table 1
summarizes the demographic characteristics of the
sample with respect to gender and ethnicity. Perusal of
the table reveals that roughly half the sample was female (46.2%), 40.8% self-identified as AA, and 59.2%
self-identified as EA.
Assessment Procedures
Each year, as part of a longitudinal intervention
study,1 student surveys were conducted primarily in
the school context using an instrument developed and
reported by colleagues at Oregon Research Institute
(Irvine, Biglan, Smolkowski, & Ary, 1999). If students
moved out of their original schools, we followed them
to their new location. Students were paid $20 for completing each assessment wave.
Peer nominations were collected in the participating sixth-grade classrooms as part of the
school-based assessment measures. Students chose
from a roster of classmates participating in the study:
1The sample analyzed in this article was part of a randomized intervention study testing the effectiveness of the Adolescent Transitions Program in preventing and reducing adolescent problem behavior and substance use. For information on the intervention and its
outcomes, see Dishion, Bullock, et al. (2002), Dishion and Kavanagh
(2003), or Dishion, Kavanagh, Schneiger, Nelson, and Kaufman
(2002).
EARLY ADOLESCENT GANG INVOLVEMENT
appear in Table 2. The gang construct and its reliability
appear in Table 3.
people they liked, people they disliked, and people
who best fit some behavioral descriptions (e.g., starts
fights, hangs around with kids who get in trouble).
They were allowed to choose as many same-grade
classmates as they wished for each question. Similarly, students selected crowd nominations from a
subsection of a roster that included a randomly selected group of participating classmates. They assigned each classmate to a particular crowd (e.g.,
crowd that is involved in music, crowd that is involved in gangs).
Academic grades for subjects also were collected
as part of the school-based assessment measures in
the sixth-grade classrooms. The public-school system
computerized academic grades, from which grades
were recorded for this sample.
Each year, teachers completed a screening measure
referred to as the Teacher Risk Screening Index (see
Dishion & Kavanagh, 2003) to identify youth at risk
for problem behavior in middle school. Behaviors included in the measure were: stubborn, dislikes school,
doesn’t feel guilty, disliked by other students, associates with troublemakers, smokes or suspected of smoking, spends time with smokers, prefers older youth, argues a lot, and hums or makes noise in class. School
counselors completed a measure similar to the crowd
nominations, assigning students listed on a roster of
names into particular crowds.
Antisocial behavior (age 11, sixth grade). Antisocial behavior was measured averaging across eight
items from a yearly self-report survey that assessed
dimensions of (a) truancy (e.g., “Lied to parents about
whereabouts or who you were with”); (b) property
crime (e.g., “Purposely damaged or tried to damage
property”); and (c) aggressive behavior (e.g., “Intentionally hit or threatened someone at school”). Reliability of the antisocial behavior scale at sixth grade
indicated an acceptable alpha of .74.
Peer relationships (age 11, sixth grade). Peer relationships were assessed using the peer nominations
of liked (accepted) and disliked (rejected) classmates.
Counts of these nominations for each youth provided
scores of how liked and how disliked a youth was
by school peers. Due to different classroom sizes,
each youth received a proportion score, dividing
the liked and disliked nominations received by the
number of classmates. These proportion scores then
were standardized to create rejection and acceptance
indicators.
Data Reduction
Academic success (age 11, sixth grade).
Academic success was measured by each adolescent’s
grade point average (GPA) at Grade 6, which was reported on a 4-point scale ranging from 1 (D or lower)
to 4 (A– to A).
Overview For this article, we created an eighthgrade gang involvement construct using items from
several different measures and respondents. We also
analyzed self-report indicators of antisocial behavior,
peer reports of popularity, and school records from
sixth grade as potential predictors of the gang construct. Means and standard deviations of all measures
Gang involvement (age 13, eighth grade).
To
meet the goal of improving the measurement of gang
involvement from a simple self-report measure, gang
involvement was measured using five separate indexes. Self-reported gang involvement was derived
from one question in the yearly self-report survey:
Table 2. Means and Standard Deviations of Gang Measures
Male
Female
EA
Measure (Scale
Sixth-Grade Adjustment
GPA (0–4)
Disliked*
Liked*
Antisocial (1–5)
Eighth-Grade Gang Involvement
Crowd*
Peer*
Counselor Ratings (1–5)
Teacher Ratings (1–5)
Self-Report (1–5)
AA
EA
AA
Overall
n
M
SD
M
SD
M
SD
M
SD
M
SD
703
695
695
702
2.87
.10
–.19
1.34
.88
1.01
.95
.45
2.10
.23
.28
1.59
.72
.92
1.20
.63
3.22
–.05
–.05
1.19
.78
.99
.85
.29
2.42
–.09
.17
1.45
.78
.10
.98
.52
2.71
.05
.02
1.37
.90
.99
1.01
.49
450
450
462
569
575
1.25
.04
1.10
1.49
1.18
2.50
.06
.33
.76
.76
2.23
.08
1.46
1.93
1.58
2.79
.80
.45
1.11
1.13
.65
.02
1.05
1.34
1.19
1.90
.03
.29
.72
.77
1.50
.04
1.26
1.70
1.33
2.78
.04
.44
1.00
.89
1.30
.04
1.19
1.57
1.29
2.51
.06
.42
.90
.88
Note: EA = European American; AA = African American; GPA = grade point average; * = standardized.
65
DISHION, NELSON, YASUI
Table 3. Gang Construct Reliability, Eighth Grade
Item–Total
Correlations
Gang Construct With
Self-Report (n = 439)*
Self-Report
Counselor Rating
Teacher Rating
Crowd Nominations
Peer Nominations
Gang Construct Without
Self-Report (n = 446)*
Counselor Rating
Teacher Rating
Crowd Nominations
Peer Nominations
Standardized α,
.68
.23
.43
.34
.43
.53
.72
.44
.34
.41
.53
Note: *n varied due to attrition and missing data.
“In the last month, how many times have you spent
time with gang members as friends?” School counselors provided ratings on a 5-point scale for each student about whether the adolescent was perceived to
be a part of gang-involved crowds (i.e., “To what extent is [name] part of the crowd who likes gangs?”).
Peers provided two different indexes of gang involvement. Crowd nominations supplied a measure of
whether each youth was perceived as part of a gang
(each classmate rated, “To what extent does [name]
hang out with gang members?”). Due to different
classroom sizes, a proportion score was calculated
based on number of nominations of gang involvement
and number of classmates. Peers also nominated
classmates whom they perceived to be involved with
deviant peers. Counts of nominations received by
each youth were again reduced to a proportion score
based on number of classmates. In addition, teachers
rated how involved they perceived each student to be
with deviant peers.
These five indicators were standardized and combined to create a multiagent, multimethod construct of
gang involvement. Reliability of the gang construct
was marginally acceptable, with an alpha of .68 (see
Table 3). Because self-report did not correlate well
with the other items, we also created a construct without the self-report measure and compared it to a gang
score based only on self-report. When only the four
other-report measures were used, reliability increased
to an alpha of .72.
Outcome Analyses
The analyses in this study tested which sixth-grade
risk factors predicted gang involvement in eighth grade
and how those predictors differed by gender and ethnicity. First, we examined how gender and ethnicity affected sixth-grade adjustment (peer relations, problem
66
behavior, and academics) using multivariate analysis
of variance. Next, we did the same with the impact of
gender and ethnicity on the gang variables.2
We used structural equation modeling to create
and test a model predicting eighth-grade gang involvement from sixth-grade adjustment and compared the fit of that model for AA and EA girls and
boys. Because self-report did not correlate well with
the other gang variables, we ran the structural equation models on both the other-report gang involvement construct and on the self-report gang involvement measure. We found the best fit for the predictive
model, then used the Mplus program (Muthén &
Muthén, 2000) multisample analysis technique to
compare models with path coefficients constrained to
be equal across ethnicity and gender, releasing constraints on these parameters one at a time to test for
differences.
Results
Sixth-Grade Adjustment
Figure 2 provides scores on sixth-grade peer relations, antisocial behavior, and academic performance
for EA and AA boys and girls. The correlation matrix
(see Table 4) showed that these indexes of adjustment
were related, so we employed multivariate analysis of
variance to test the effects of gender and ethnicity on
the combination of sixth-grade adjustment variables.
Pillai’s criterion was used to determine if the multivariate differences across the groups were statistically reliable.
Ethnicity was associated reliably with the set of dependent variables measured in sixth grade, V = .25,
F(4, 672) = 55.56, p < .001. GPA contributed the most
to these observed differences, loading highly on the
discriminant function (standardized discriminant function coefficient α = .91). Antisocial behavior and
liked-most nominations also contributed to the difference, but liked-least nominations did not differentiate
between groups. EAs had higher GPAs, lower self-reports of antisocial behavior, and fewer liked-most
nominations than AAs.
2Because half the sample participated in an intervention between
sixth and eighth grades as part of the larger longitudinal study, we
tested for intervention effects on gang adjustment. Preliminary analyses indicated that adolescents in the intervention condition reported
less gang involvement, F(1, 573) = 11.57, p < .01. However, differences in treatment condition did not affect the differences in level of
gang involvement for gender or ethnicity, nor did it affect the correlations between sixth-grade adjustment and later gang involvement. Because intervention did not influence either the overall patterns of gang involvement by gender and ethnicity or the predictive
model, we included both intervention and control adolescents in
these analyses.
EARLY ADOLESCENT GANG INVOLVEMENT
Figure 2. Sixth-grade adjustment.
Table 4. Measures: Zero-Order Correlation Matrix
Grade 8
Grade 6
GPA
Liked
Dislike
Antisocial
Crowd Nomination
Peer Nomination
Teacher Ratings
Counselor Ratings
Self-report
Nomination
Ratings
GPA
Liked
Disliked
Antisocial
Crowd
Peer
Counselor
Teacher
Self
1.00
–.20**
.09
–.38**
–.28**
–.32**
–.36**
–.37**
–.21**
1.00
.02
.11*
.13**
.32**
.11*
.18**
.07
1.00
.12*
–.09
.35**
.08
.03
.03
1.00
.23**
.34**
.25**
.22**
.43**
1.00
.44**
.38**
.30**
.21**
1.00
.46**
.42**
.18**
1.00
.35**
.14**
1.00
.11*
1.00
Note: GPA = grade point average. Listwise deletion; n = 423.
*p < .05. **p < .01.
Gender also was associated with sixth-grade adjustment, V = .06, F(4, 672) = 9.80, p < .001. Again, GPA
did the most to distinguish between the groups (α =
.72). Antisocial behavior and liked-least nomination
also contributed to the discriminant function, but
liked-most nominations did not. Girls had higher
GPAs, lower self-reported antisocial behavior, and
fewer liked-least nominations than boys. No significant statistical interaction occurred between gender
and ethnicity on sixth-grade adjustment at the multivariate or univariate level.
Eighth-Grade Gang Involvement
Figure 3 shows gang involvement scores (both
self-report and other report) for AA and EA boys and
girls. Overall, as seen in Figure 3, AA boys scored
highest on gang involvement, followed by AA girls. To
test the effects of ethnicity and gender on all aspects of
gang involvement, we again ran multivariate analyses
of variance, this time on the individual indicators of the
gang construct.3 Using listwise deletion, there was
considerable missing data (n = 439).4
3We decomposed the gang construct into its separate indicators
for this analysis because of the possibility that specific reports (e.g.,
teacher assessments) might be more important than others in distinguishing between groups. This could be due to respondent bias or a
number of other factors.
4One reason for missing data is that many students moved to new
schools after the sixth-grade assessment. Participants with missing
data by eighth grade did differ from those without missing data on
sixth-grade adjustment, V = .10, F(4, 674) = 18.23, p < .001. Although this is not relevant to our analysis, it is important to note.
67
DISHION, NELSON, YASUI
Figure 3. Eighth-grade gang involvement.
Indicators of eighth-grade involvement significantly distinguished between AAs and EAs, V = .14,
F(5, 431) = 14.27, p < .001. All the indicators contributed to the distinction, but counselor ratings and peer
nominations loaded most highly on the discriminant
function. Crowd nominations showed the least loading
on the discriminant function. AAs scored higher on all
indicators of gang involvement than did EAs.
Gender also had a significant effect on gang involvement, V = .09, F(5, 431) = 8.30, p < .001. Peer
nominations mainly accounted for this difference,
self-report did not load at all on the discriminant
function, and the other indicators contributed moderately. Boys scored higher than girls on all indicators. Again, the statistical interaction between gender
and ethnicity on gang involvement was not statistically reliable.
Prediction of Eighth-Grade Gang
Involvement
The correlations shown in Table 4 among the indicators for gang involvement also were run separately
by ethnicity and gender. A Fisher’s Z transformation
was computed to evaluate whether observed differences in the zero-order correlations across subgroups
were statistically reliable. None of these comparisons
were statistically different.
68
Because EA and AA boys and girls appeared to differ in their levels of gang involvement and their
sixth-grade adjustment, we used a structural equation
modeling framework to test whether prediction of
eighth-grade gang involvement differed among these
groups. We ran the model on the full sample to test fit,
then ran a set of multigroup models, testing how the parameters varied by ethnicity and gender. Because of the
low convergence between self- and other-report measures of gang involvement, these models were run on
both the other-report gang involvement construct and
the self-report gang involvement measure.
Specifically, as seen in Figure 4, we set up the
model to predict the eighth-grade gang construct (represented by a single indicator—in the first set of analyses, the average of the other-report gang involvement
variables; in the second set, the self-report indicator)
from sixth-grade antisocial behavior, liked-most and
liked-least nominations, and GPA. All variables were
standardized to run the model. Missing data were imputed using the Mplus expectation maximization approach (Muthén & Muthén, 2000).
We ran the model on the overall sample, setting the
intercepts at zero and allowing all other indexes to be
estimated freely. Next, we ran a multigroup model
across all four groups, allowing measurement
parameters (intercepts and variances) and predictor
covariances to vary across groups, constraining all
EARLY ADOLESCENT GANG INVOLVEMENT
Figure 4. Prediction of other-report and self-report gang involvement: Freely estimated model.
Table 5. Gang Involvement: Parameter Constraints for European American and African American Girls and Boys
Loading Allowed to Vary
Other report
On GPA
On antisocial
On like
On dislike
Self-report
On GPA
On antisocial
On like
On dislike
Improvement From Constrained
Model (χ2 Change)
Loading
MEA
FEA
MAA
FAA
.20*
.20**
.57*
.32*
χ2(3) = 1.91, p = ns
χ2(3) = 5.22, p = ns
χ2(3) = 4.31, p = ns
χ2(3) = 4.03, p = ns
χ2(3) = 4.34, p = ns
χ2(3) = 13.78, p < .05
χ2(3) = 4.07, p = ns
χ2(3) = 5.01, p = ns
Note: MEA = male European American; FEA = female European American; MAA = male African American; FAA = female African American.
Grade point average.
*p < .05.
paths from predictors to gang involvement to be equal
across groups. We allowed the measurement parameters of the model to vary because they were not addressed specifically by our hypothesis and relaxing the
constraints improved the model. Finally, we released
the equality constraint on one predictor loading parameter at a time (e.g., allowing the relation between GPA
and gang involvement to be different for each group)
and tested both the change in model fit and the loading
of the predictor for each group.
Other report of gang involvement For other reports of gang involvement, the overall model fit the data
well. As seen in Figure 4, GPA, antisocial behavior, be-
ing liked, and being disliked all predicted later gang involvement. GPA was the strongest predictor (grades
correlated negatively with the gang construct), followed
by antisocial behavior (positive correlation with gang
involvement) and being disliked and being liked (both
positive correlations with gang involvement).
The multigroup model, with paths from predictors to
gang involvement constrained to be equal across groups,
fit the data marginally well, χ2(15) = 33.48, p < .01, root
mean square error of approximation = .08, comparative
fit index = .96. When the equality constraint on each predictor path was relaxed independently, none of the
resulting models were an improvement in fit beyond the
constrained model (see Table 5).
69
DISHION, NELSON, YASUI
Self-report of gang involvement. For self-report gang involvement, the overall model also fit the
data well. As seen in Figure 4, only being disliked and
antisocial behavior predicted self-report of later gang
involvement. Antisocial behavior was the strongest
predictor (correlating positively with the gang measure), followed by being disliked (also a positive correlation with gang involvement).
The multigroup model, with paths from predictors
to gang involvement constrained to be equal across
groups, fit the data marginally well, χ2(15) = 28.79, p <
.05, root mean square error of approximation = .07,
comparative fit index = .97. When the equality constraint on each predictor path was relaxed independently, the relaxation of the path from antisocial behavior to self-reported gang involvement improved the
model fit significantly beyond the constrained model,
χ2 change (15) = 13.78, p < .01. Antisocial behavior
was a much stronger (positive) predictor of self-reported gang involvement for AA boys than for any
other group (see Table 5).
Discussion
This study provided a rare opportunity to examine
the role of school adaptation in the early emergence of
gang involvement in adolescence. As expected, youth
antisocial behavior predicted deviant peer affiliation
and gang involvement. This is not a surprise and is predicted by most models of adolescent problem behavior,
including control theory (M. R. Gottfredson & Hirschi,
1990). However, once controlling for antisocial behavior, there was clear evidence that “failure” in public
middle school also was associated with the emergence
of gang involvement. Of specific concern was the relatively robust prediction of gang involvement for peer
rejection and academic failure. This finding replicates
and extends our previous work on predicting deviant
peer involvement in early adolescence (Dishion et al.,
1991).
The picture is complex, however, as both peer rejection and acceptance predict gang involvement 2 years
later, at least when the latter construct is comprised of
teacher, peer, and school-staff ratings. This unexpected
finding suggests that a more fine-grained analysis of
peer social networks is in order, because it may well be
that some antisocial youth are both liked and disliked by
their various cliques across the network of sixth-grade
students. For example, Stormshak et al. (1999) found
that aggressive youth actually were well-liked in classrooms with other aggressive students. It seems that
youth on the path to gang involvement are respected,
liked, and disliked by many of their peers in school.
These data do support the hypothesis that school environments with high rates of failure and peer acrimony will serve as breeding grounds for deviant peer
70
groups, gang involvement, and, consequently, high
rates of serious problem behavior. If negative experiences in school are prognostic of gang involvement,
the critical question becomes, what is the mechanism?
Over the past 10 years, we have studied a process referred to as the “deviant friendship process” that may
be relevant to understanding the mechanism linking
failure to deviant peer aggregation. The idea is relatively straightforward: Through a process of selective
reinforcements (i.e., “shopping”), young adolescents
end up interacting with individuals who support their
attitudes, values, and behaviors (Dishion, 1990; Patterson et al., 1992).
Poor grades and peer disdain are punishing experiences, indeed. The salience of peer reinforcement is
amplified in youth with such experiences, and marginalized youth soon find each other. Topics of discussion that emerge as interesting for these youth are
negative discussions of adults, teachers, prosocial
peers, principals, and deviant behavior. Attention,
laughter, and eventually peer respect reinforce and
maintain increasingly more serious forms of problem
behavior and a deep psychological sense of identification with deviance (Dishion, Nelson, Winter, &
Bullock, 2004).
By way of extension, we expect that certain boys
and girls in middle school increasingly will select
friendships that support deviance, which, in turn, will
be associated with increases in deviance, identification, involvement, and affiliation with deviant peer
groups. Over the course of middle school, deviant peer
groups will begin to coalesce. However, the more seriously deviant peer groups, those that identify as a gang,
may not reside in the school setting. In fact, there is a
need to conduct more careful peer relations research
that establishes the geography of social networks as
they emerge over time. An emerging area of study that
seems particularly relevant to this issue is the finding
that sibling problem behavior is as predictive of problematic outcomes as are deviant peer groups (Brook,
Brook, & Whiteman, 1999; Patterson, 1984, 1986;
Stormshak, Comeau, & Shepard, 2004). In fact, careful
analyses of family dynamics suggest that siblings collude to undermine parents’ efforts to monitor and manage their young adolescents’ behavior (Bullock &
Dishion, 2002).
The complexity of defining a peer group independent of other social groups may speak to the difficulty of
defining a coherent gang construct. Klein (1993) discussed the conceptual complexities of defining a gang.
The lack of convergence between our self-reported
measure of gang involvement and those derived from
peers, teachers, and school staff speak to this issue.
This is particularly problematic for research on gang
involvement, as much of the literature is based on
self-report. We suggest that the self-report measure be
expanded to include more items to increase its reliabil-
EARLY ADOLESCENT GANG INVOLVEMENT
ity. One way to do this is to include attitudes toward
gangs, and not simply whether the youth is involved in
a gang. The work by Kiesner et al. (2002) underscored
the importance of identification with a deviant peer
group in the influence process.
The richness of the sample allowed us to consider
whether the development and ecology of early gang involvement varied by gender and ethnicity. Although
the adolescents differed by gender and ethnicity on
measures of sixth-grade adjustment and eighth-grade
gang involvement, we did not find major differences in
the pattern of prediction of gang involvement when we
used a conservative approach to examining model invariance. The conservative approach is one in which
we allow the measurement of each of the constructs to
vary by group but constrain the structural coefficients.
Using this approach, the only statistically reliable difference we found was in the prediction of gang involvement from earlier antisocial behavior, which did
vary modestly by subgroup.
Limitations exist to this research in addressing the
hypothesis under consideration. First, the developmental age span is narrow, especially when considering
gang involvement. Most studies of gang involvement
suggest that gang involvement begins around age 13
and increases up to age 18 (Lahey et al., 1999; Thornberry et al., 1993). Certainly the dynamics of early
gang involvement may be different from those of gang
involvement later in adolescence. For one thing, we
can be certain that many youth involved in gangs in
later adolescence are less sensitive to school experiences simply because they no longer attend school.
Second, much of the research on adolescent problem behavior has used deviant peer group involvement
as predictors of adolescent problem behavior. However, an ecological understanding of problem behavior
may result from considering the developmental experiences that result in the formation of deviant peer
groups, in general, and organized gangs, in particular.
Research by Thornberry et al. (1993) and events over
the past decade related to school shootings reveal that
such deviant peer networks can have dramatic and
tragic effects for all concerned.
We propose that future research continues to clarify
the conditions conducive to gang involvement. We
suggest an ecological framework in providing a broad
perspective on the etiology and course of antisocial behavior and deviant peer clustering. A cornerstone on
the ecological perspective is the central role of families
as both a resource and a vulnerability factor. As a resource, the evidence suggests that parents can be mobilized as a countervailing force to the deviant peer group
in early adolescence. For example, our analyses of the
impact of our family-centered intervention revealed
statistically reliable reductions in self-reported deviant
peer involvement (Dishion, Bullock, & Granic, 2002)
and self-reported gang involvement. However, other
reports of gang involvement seemingly were unaffected by our family interventions (for details, see
Dishion & Kavanagh, 2003).
If the findings linking academic failure and peer rejection to early gang involvement replicate, it would be
useful to consider an integration between a family-centered intervention and those that focus on the school as
a system (see Crone & Horner, 2003; D. C. Gottfredson, G. D. Gottfredson, & Hybl, 1993; Sugai, Horner, & Sprague, 1999). Improving the organization,
structure, pedagogy, and behavior management practices may, in fact, make a difference with respect to the
peer dynamics within a public-school environment.
Such improvements, however, require an investment of
resources in public education environments to enhance
the ability of school professionals to address the
changing needs of students and families.
References
Bank, L., Dishion, T. J., Skinner, M., & Patterson, G. R. (1990).
Method variance in structural equation modeling: Living with
“Glop.” In G. R. Patterson (Ed.), Depression and aggression in
family interaction (pp. 247–279). Hillsdale, NJ: Lawrence Erlbaum Associates, Inc.
Battin, S. R., Hill, K. G., Abbott, R. D., Catalano, R. F., & Hawkins,
J. D. (1998). The contribution of gang membership to delinquency beyond delinquent friends. Criminology, 36, 93–112.
Biglan, A., Brennan, P. A., Foster, S. L., & Holder, H. D. (2004).
Helping adolescents at risk: Prevention of problem behaviors.
New York: Guilford.
Brook, J. S., Brook, D. W., & Whiteman, M. (1999). Older sibling
correlates of younger sibling drug use in the context of parent–child relations. Genetic, Social, and General Psychology
Monographs, 125, 451–468.
Bullock, B. M., & Dishion, T. J. (2002). Sibling collusion and problem
behavior in early adolescence: Toward a process model for family
mutuality. Journal of Abnormal Child Psychology, 30,143–153.
Cadwallader, T. W., & Cairns, R. B. (2002). Developmental influences and gang awareness among African-American inner city
youth. Social Development, 11, 245–265.
Cairns, R. B., Cadwallader, T. W., Estell, D., & Neckerman, H.
(1997). Groups to gangs: Developmental and criminological
perspectives and relevance to prevention. In D. M. Stoff, J.
Breiling, & J. D. Mader (Eds.), Handbook of antisocial behavior (pp. 194–205). New York: Wiley.
Cairns, R. B., Cairns, B. D., Neckerman, H. J., Gest, S. D., &
Gariépy, J. L. (1988). Social networks and aggressive behavior:
Peer support or peer rejection. Developmental Psychology, 24,
815–823.
Cicchetti, D., & Toth, S. L. (1992). The role of developmental and
theory in prevention and intervention. Development and Psychopathology, 4, 489–493.
Craig, W., Vitaro, F., Gagnon, C., & Tremblay, R. E. (2002). The
road to gang membership: Characteristics of male gang and
nongang members from ages 10 to 14. Social Development, 11,
53–68.
Crone, D. A., & Horner, R. H. (2003). Building positive behavior
support systems in schools: Functional behavioral assessment.
New York: Guilford.
Dishion, T. J. (1990). Peer context of child and adolescent troublesome behavior. In P. Leone (Ed.), Understanding troubled and
troublesome youth (pp. 128–153). Beverly Hills, CA: Sage.
71
DISHION, NELSON, YASUI
Dishion, T. J., Bullock, B. M., & Granic, I. (2002). Pragmatism in
modeling peer influence: Dynamics, outcomes, and change
processes. In D. Cicchetti & S. Hinshaw (Eds.), How prevention
intervention studies in the field of developmental psychopathology can inform development theories and models [Special
Issue]. Development and Psychopathology, 14, 969–981.
Dishion, T. J., Eddy, J. M., Haas, E., Li, F., & Spracklen, K. (1997).
Friendships and violent behavior during adolescence. Social
Development, 6, 207–223.
Dishion, T. J., French, D. C., & Patterson, G. R. (1995). The development and ecology of antisocial behavior. In D. Cicchetti & D. J.
Cohen (Eds.), Developmental Psychopathology: Vol. 2. Risk,
disorder, and adaptation (pp. 421–471). New York: Wiley.
Dishion, T. J., & Kavanagh, K. (2003). Intervening in adolescent
problem behavior: A family-centered approach. New York:
Guilford.
Dishion, T. J., Kavanagh, K., Schneiger, A., Nelson, S., & Kaufman,
N. (2002). Preventing early adolescent substance use: A family-centered strategy for the public middle-school ecology. In
R. L. Spoth, K. Kavanagh., & T. J. Dishion (Eds.), Universal
family-centered prevention strategies: Current findings and
critical issues for public health impact [Special Issue]. Prevention Science, 3, 191–201.
Dishion, T. J., McCord, J., & Poulin, F. (1999). When interventions
harm: Peer groups and problem behavior. American Psychologist, 54, 755–764.
Dishion, T. J., Nelson, S. E., & Bullock, B. M. (2004). Premature adolescent autonomy: Amplification of the risk trajectory to
young adult problem behavior. In J. Kiesner & M. Kerr (Eds.),
Peer and family processes in the development of antisocial and
aggressive behavior [Special Issue]. Journal of Research in Adolescence, 27, 515–530.
Dishion, T. J., Nelson, S. E., Winter, C. E., & Bullock, B. M. (2004).
Friendship interactions in the development and course of antisocial behavior: Entropy as a dynamic index of self-organized
deviance. Journal of Abnormal Psychology, 32, 651–663.
Dishion, T. J., & Patterson, G. R. (in press). The development and
ecology of antisocial behavior. In D. Cicchetti & D. J. Cohen
(Eds.), Developmental psychopathology: Vol. 3. Risk, disorder,
and adaptation. New York: Wiley.
Dishion, T. J., & Patterson, G. R. (1999). Model-building in developmental psychopathology: A pragmatic approach to understanding and intervention. Journal of Clinical Child Psychology, 28,
502–512.
Dishion, T. J., Patterson, G. R., Stoolmiller, M., & Skinner, M. S.
(1991). Family, school, and behavioral antecedents to early adolescent involvement with antisocial peers. Developmental
Psychology, 27, 172–180.
Dishion, T. J., Poulin, F., & Medici Skaggs, N. (2000). The ecology
of premature autonomy in adolescence: Biological and social
influences. In K. A. Kerns, J. M. Contreras, & A. M. Neal–Barnett (Eds.), Family and peers: Linking two social worlds (pp.
27–45). Westport, CT: Praeger.
Dishion, T. J., Spracklen, K. M., Andrews, D. M., & Patterson, G. R.
(1996). Deviancy training in male adolescent friendships. Behavior Therapy, 27, 373–390.
Elliott, D. S., Huizinga, D., & Ageton, S. S. (1985). Explaining delinquency and drug use. Beverly Hills, CA: Sage.
Fergusson, D. M., & Horwood, L. J. (1999). Prospective childhood
predictors of deviant peer affiliations in adolescence. Journal of
Child Psychology and Psychiatry, 40, 581–592.
Gorman–Smith, D., & Tolan, P. (1998). The role of exposure to community violence and developmental problems among inner city
youth. Development and Psychopathology, 10, 101–116.
Gottfredson, D. (1990). Changing school structures to benefit highrisk youth. In P. Leone (Ed.), Understanding troubled and troubling youth: Multidisciplinary perspectives (pp. 123–135).
Newbury Park, CA: Sage.
72
Gottfredson, D. C., Gottfredson, G. D., & Hybl, L. G. (1993). Managing adolescent behavior: A multiyear, multischool study.
American Educational Research Journal, 30, 179–215.
Gottfredson, M. R., & Hirschi, T. (1990). A general theory of crime.
Stanford, CA: Stanford University Press.
Harris, J. R. (1995). Where is the child’s environment? A group socialization theory of development. Psychological Review, 102,
458–489.
Hartup, W. W. (1996). The company they keep: Friendships and their
developmental significance. Child Development, 67, 1–13.
Irvine, A. B., Biglan, A., Smolkowski, K., & Ary, D. V. (1999). The
value of the Parenting Scale for measuring the discipline practices of parents of middle school children. Behavior Research
and Therapy, 37, 127–142.
Kandel, D. B. (1986). Process of peer influence on adolescence. In
R. K. Silbereisen (Ed.), Development as action in context (pp.
33–52). Berlin: Springer-Verlag.
Kiesner, J., Cadinu, M. R., Poulin, F., & Bucci, M. (2002). Group
identification in early adolescence: Its moderator effects on
peer influence and its relationship with peer status. Child Development, 73, 196–208.
Klein, M. W. (1993) Attempting gang control by suppression: The
misuse of deterrence principles. Studies on Crime and Crime
Prevention, 2, 88–111.
Lahey, B. B., Gordon, R. A., Loeber, R., Stouthamer–Loeber, M., &
Farrington, D. P. (1999). Boys who join gangs: A prospective
study of predictors of first gang entry. Journal of Abnormal
Child Psychology, 27, 261–276.
Maxson, C. L., & Klein, M. W. (1995) Investigating gang structures.
Journal of Gang Research, 3, 33–40.
McLoyd, V. C. (1990). The impact of economic hardship on Black
families and children: Psychological distress, parenting, and
socioemotional development. Child Development, 61, 311–346.
Muthén, B., & Muthén, L. K. (2000). Integrating person-centered
and variable-centered analysis: Growth mixture modeling with
latent trajectory classes. Alcoholism: Clinical and Experimental Research, 24, 882–891.
Osgood, D. W., Wilson, J. K., Bachman, J. G., O’Malley, P. M., &
Johnston, L. D. (1996). Routine activities and individual deviant behavior. American Sociological Review, 61, 635–655.
Patterson, G. R. (1984). Siblings: Fellow travelers in a coercive system. In R. J. Blanchard & D. C. Blanchard (Eds.), Advances in
the study of aggression (Vol. 1, pp. 173–215). New York:
Academic.
Patterson, G. R., & Dishion, T. J. (1985). Contributions of families
and peers to delinquency. Criminology, 23, 63–79.
Patterson, G. R., Dishion, T. J., & Yoerger, K. (2000). Adolescent
growth in new forms of problem behavior: Macro- and micro-peer dynamics. Prevention Science, 1, 3–13.
Patterson, G. R., Reid, J. B., & Dishion, T. J. (1992). A social learning approach: IV. Antisocial boys. Eugene, OR: Castalia.
Poulin, F., & Boivin, M. (2000). The formation and development of
friendship in childhood: The role of proactive and reactive aggression. Developmental Psychology, 36, 233–240.
Sampson, R. J., & Laub, J. H. (1994). Urban poverty and the family
context of delinquency: A new look at structure and process in a
classic study. Child Development, 65, 523–540.
Simcha-Fagan, O., & Schwartz, J. E. (1986). Neighborhood and delinquency: An assessment of contextual effects. Criminology,
24, 667–703.
Stormshak, E. A., Bierman, K. L., Bruschi, C., Dodge, K. A., Coie, J.
D., & the Conduct Problems Prevention Research Group. (1999).
The relation between behavior problems and peer preference in
different classroom context. Child Development, 70, 169–182.
Stormshak, E. A., Comeau, C. A., & Shepard, S. A. (2004). The relative contribution of sibling deviance and peer deviance in the
prediction of substance use across middle childhood. Journal of
Abnormal Child Psychology, 32, 635–649.
EARLY ADOLESCENT GANG INVOLVEMENT
Sugai, G., Horner, R. H., & Sprague, J. R. (1999). Functional assessment-based behavior support planning: Research to practice to
research. Behavior Disorders, 24, 253–257.
Thornberry, T. P. (1998). Membership in youth gangs and involvement in serious and violent offending. In R. Loeber & D. P.
Farrington (Eds.), Serious and violent juvenile offenders: Risk
factors and successful interventions (pp. 147–166). Newbury
Park, CA: Sage.
Thornberry, T. P., Krohn, M. D., Lizotte, A. J., & Chard–Wierschem,
D. (1993). The role of juvenile gangs in facilitating delinquent
behavior. Journal of Research in Crime and Delinquency, 30,
55–87.
Vitaro, F., Tremblay, R. E., Kerr, M., Pagani, L., & Bukowski, W. M.
(1997). Disruptiveness, friends’ characteristics, and delinquency in early adolescence: A test of two competing models of development. Child Development, 68, 676–689.
Received August 14, 2003
Accepted July 12, 2004
73