Content uploaded by Jean-Louis Gariepy
Author content
All content in this area was uploaded by Jean-Louis Gariepy on Oct 27, 2014
Content may be subject to copyright.
Developmental ftychology
1988,
Vol. 24, No. 6, 815-823Copyright 1988 by (he American Psychological Association, Inc.
O012-1649/88/SO0.75
Social Networks and Aggressive Behavior:
Peer Support or Peer Rejection?
Robert
B.
Cairns, Beverley D. Cairns, Holly J. Neckerman, Scott
D.
Gest, and Jean-Louis Gariepy
University of North Carolina at Chapel Hill
Studied social networks and aggressive behavior in school in 2 cohorts of
boys
and girls in the 4th
and
7th grades
(N
=
695).
Measures of social networks yielded convergent
findings.
Highly
aggressive
subjects (both boys and girls) did not differ from matched control subjects in terms of social cluster
membership or in being isolated or rejected within the social network. Peer cluster analysis and
reciprocal "best friend"
selections
indicated that
aggressive
subjects tended
to
affiliate with aggressive
peers.
Even
though
highly aggressive
children and adolescents
were less
popular than control subjects
in the social network at large, they were equally often identified
as
being nuclear members of social
clusters. Aggressive subjects did not differ from matched control subjects in the number of times
they
were
named by
peers as
"best friend," nor did the
two
groups differ in the probability of having
friendship choices reciprocated
by
peers.
It has been broadly assumed that aggressive children are em-
bedded in a familial matrix whereby negative actions support
the consolidation of further hurtful, aggressive actions (Patter-
son, 1982; see Parke & Slaby, 1983, for a review). The finding
of reciprocities in dyadic aggressive interchanges and coercive
families
has
supported
the
idea that
aggressive
children
are
both
the architects and the victims of their actions (Hall & Cairns,
1984;
Patterson, 1982). Although the coercive social model has
been most clearly elaborated for family interactions, it seems
reasonable to expect that the same processes occur in social
networks beyond the family (Bronfenbrenner, 1979).
In particular, peer social clusters may also provide mutual
support for aggressive behaviors as new social units emerge in
adolescence (Cairns, 1979; Cairns, Neckerman, & Cairns, in
press).
To the extent that adolescents participate in the design
of their social environments, they may be expected to affiliate
with peers who are similar to themselves in salient life-style di-
mensions, including the propensity to act out toward others.
Once in the close network of relationships, reciprocal processes
should bring about even higher
levels
of similarity in aggressive
behavior. The patterns of affiliation may be consolidated by so-
cial choice (selective acceptance by aggressive peers) or social
default (selective exclusion by nonaggressive peers). Accord-
ingly, "coercive clusters" in adolescence may (a) present multi-
ple opportunities for aggressive reciprocation and escalation
and (b) support a value structure that promotes aggressive ac-
tions
toward other persons.
Following these considerations, the dual aims of this study
This research was supported by grants from the Spencer Foundation
and the National Institute of Child Health and Human Development
(R01 23301).
We thank Tamara R. Flinchum and Lynda Ferguson for their assis-
tance in several aspects of this investigation.
Correspondence concerning this article should be addressed to Rob-
ert
B.
Cairns, University of North Carolina at Chapel Hill, Department
of Psychology, Davie Hall
013A,
Chapel
Hill,
North Carolina 27514.
were
to clarify the
roles
that highly aggressive children and ado-
lescents play
in peer
social
networks and to understand
the
func-
tions that networks of peers play in the support of aggressive
patterns. Attempts to study social relationships of aggressive
children indicate that they instigate high
levels
of reciprocity in
aggressive interactions. Within limits, aggression begets aggres-
sion in a reciprocal, escalating pattern (e.g., Hall & Cairns,
l984;Raush,
1965;
Toch,
1969).
The mutual support of aggres-
sive behaviors seems to be a factor in the dynamics of various
coercive groups, including gangs of bullies (Olweus, 1979) and
delinquents
(e.g.,
Cohen,
1955;
Giordano, Cernkovich,
&
Pugh,
1986).
It should be noted that a different picture has been described
in recent studies of children's social status. Subjects identified
as "rejected" (as judged by peer ratings/nominations of likabil-
ity and unlikability) have typically been found to be more ag-
gressive than nonrejected subjects, and early identification as
"rejected" is predictive of subsequent problems with aggressive
behavior (Asher & Dodge, 1986; Coie & Dodge, 1983). Thus,
it would appear that aggressive acts are correlated with social
rejection and that rejection is correlated with the continuation
of aggressive behavior. At a broader
level,
the association of
re-
jection and aggression has been seen as consistent with the em-
phasis of sociologists Hirschi (1969) and Yablonsky (1962) on
the essential social disaffiliation and disengagement of delin-
quent and aggressive youths.
There is some evidence from studies of children, however, to
indicate that dislike and popularity can coexist in the same indi-
vidual and that assertive-aggressive behavior does not necessar-
ily preclude popularity (Coie, Dodge, & Coppotelli, 1982). A
subgroup of subjects labeled
controversial—presumably
be-
cause they obtained, simultaneously, higher-than-average scores
on peer popularity and peer dislike—showed a blend of antiso-
cial behavior and peer acceptance. This subgroup is of special
interest because controversial subjects (who do not show up in
large numbers) can sometimes be viewed as leaders in the peer
group (Coie et
al.,
1982, p. 568).
815
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
816CAIRNS, CAIRNS, NECKERMAN, GEST, GARIEPY
Following assumptions discussed elsewhere (Cairns et al, in
press;
Kandel, 1978), the present research addressed three pro-
posals on the relationship between aggressive behaviors and so-
cial affiliations. First, highly aggressive subjects were expected
to be members of definable social clusters and not to differ from
control subjects on this index of affiliation. Because of their
abusive and coercive behaviors, aggressive subjects may be gen-
erally less popular than control subjects in the social network
as a whole. On the other hand, aggressive adolescents may be
accepted by other persons in the social cluster with whom they
are identified. Hence, aggressive adolescents would be as likely
as
control subjects to have reciprocated "best friend" selections.
A second proposal concerned the extent to which there is so-
cial support by peers for aggressive behaviors. Aggressive sub-
jects may be expected to affiliate with aggressive peers. Across
social clusters, there should be high levels of similarity between
individuals of the
same
cluster with respect
to
aggressive expres-
sion (i.e., "homophily"). This expectation follows from the re-
ciprocal and contagious nature of aggressive behaviors in social
groups as
well as
the social choice and social default factors that
serve to define entry into groups and help consolidate their
structures.
A
third set of
issues
concerned gender differences and whether
the patterns of group membership in highly aggressive
girls were
different from those in highly aggressive boys. Most relevant
evidence indicates that the standards for the acceptability of
as-
sertive-aggressive behavior differ for boys and girls
(e.g.,
Cairns,
Cairns, Neckerman, Ferguson, & Gariepy, 1988; Savin-Wil-
liams,
1979). The different normative standards, in turn, may
account for the difference in the sheer numbers of highly
aggres-
sive girls relative to highly aggressive boys. These social stan-
dards may also promote differences in the extent to which ag-
gressive
patterns are permitted to become a basis for differential
peer group association. On the basis of the foregoing, it was
expected that both
boys
and
girls
would
show
patterns of aggres-
sive
homophily, but that
girls
would be more vulnerable to rejec-
tion and social ostracism because of overt aggressive behavior.
Studies of aggressive subjects and their social groups have
suffered from a gender bias. For example, investigations of re-
ciprocal relationships of violent subjects and studies of "gang"
behavior have been traditionally limited to males (e.g., Cohen,
1955,
Toch,
1969).
There have been few instances where aggres-
sive reciprocal relationships have been studied in girls (or in
boys'
relationships with girls) or in female "gangs" (Giordano
et al., 1986). This bias is of
some
importance because the evi-
dence that is available on reciprocal relationships involving
girls indicates that the
sex
of the "other" does
make
a difference.
Different patterns of reciprocal action have been obtained in
male-female interactions relative to male-male ones
(e.g.,
Bar-
rett, 1979). The difference reflects the operation of a dual stan-
dard, in that girls are not supposed to be attacked by boys, al-
though girls may attack both boys and girls. In one of
the
few
systematic studies of adolescent delinquent gangs of
girls,
Gior-
dano (1978) found within-group support for deviant norms not
unlike that found in males. More recently, Magnusson (1987)
has shown that affiliations with older peers play a powerful fac-
tor in accounting for the deviancy of early maturing Swedish
teenage girls.
Shortcomings in techniques available for social network anal-
ysis have made
it difficult
to
describe
the
supportive
role
of peers
at any age (e.g., Dunphy, 1963; Hartup, 1983; Moreno, 1934).
Procedures that describe the social status of children often do
not provide information about the nature of the social groups
in which children participate
(e.g.,
Coie
&
Dodge,
1983;
Peery,
1979).
Beyond categorization of children as "popular"
"rejected," "isolated," "average," or "controversial," classical
sociometry would seem to require information about (a) the
identities of individuals within a given social cluster and (b) the
number of social clusters within a microsocial network (Bron-
fenbrenner, 1944a,
1944b;
Moreno,
1934).
To this
end,
the
pres-
ent study introduced a procedure designed to capture informa-
tion about the nature, status, and composition of social net-
works within schools.
Method
Subjects
A total of 695 subjects (364 girls and 331 boys) were recruited from
seven public schools in two cohorts. Cohort
1
consisted of 220 fourth-
grade subjects (116 girls and 104 boys; A/age = 10.2 years, SD = .57)
from four elementary schools. Cohort 2 consisted of 475 seventh-grade
subjects (248 girls and 227 boys; Mage = 13.4 years, SD = .58) from
three middle schools. The mean family socioeconomic status on the
Duncan scale (Featherman revision) was 30.2 (SD = 17.1) in Cohort 1
and 31.6 (SD
=
17,8) in Cohort
2,
and the full range of occupations was
represented in the samples (range from 88 to 7; i.e., chair of medical
center department, attorney, regional sales manager, small-business
owner, truck driver, domestic worker, unemployed farm worker, etc.).
Twenty-five percent of
the
subjects
were
minority status (predominantly
Black).
The seven
schools were
located in
two
counties:
one,
a
suburban
metropolitan area, and the other, a rural county (as classified by the
1980 U.S. census). There were no restrictions on inclusion other than
consent: all children in the designated grade (fourth or seventh) in each
school were included in the study if
(a)
the children wished to partici-
pate and (b) they and their parent or legal guardian signed a statement
of informed consent.
The
participation rate ranged from
89%
in
the
last
junior high school assessed (132 of 149) to
50%
in the
first
junior high
school (83 of 166), with an overall participation of
70%
(695 of
994).
The
70%
of the children
who
consented to be subjects and the
30%
who
did not were compared in terms of ethnic status, sex, and probability
of being nominated
as highly
aggressive.
No systematic differences
were
obtained on any of these dimensions.
Within the larger sample, 40 subjects, 20 girls and 20 boys, were
judged by teachers, counselors, and principals to be highly aggressive
(i.e.,
there were 20 highly aggressive subjects in each cohort). In order
for an individual to be selected, he or she had to be nominated by two
school personnel (teacher, counselor, or principal) who were closely ac-
quainted
with the
subject.
An
additional group of
40
nonaggressive
con-
trol subjects was identified and matched individually on the basis of
sex, race, classroom attended, physical size, socioeconomic status, and
chronological age.1 Priority was given to the matching variables in the
! The matching was successful on all variables (e.g., no significant
differences were obtained on classroom, race, physical size, socioeco-
nomic status,
age),
with one exception. The aggressive-control subjects
in the fourth grade did not differ
in
age
(10.1
vs.
10.3
years in aggressive
and control girls and 10.6
vs.
10.5 years in aggressive and control boys,
respectively). However, the aggressive-control subjects in the seventh
grade differed (13.6 vs. 13.0 years in aggressive and control girls and
14.1
vs.
13.3
in
the aggressive
and control
boys,
respectively).
To
correct
for any effects attributable to the age discrepancy in the older sample,
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
SOCIAL SUPPORT OR SOCIAL REJECTION?817
order of
listing.
In order for children to qualify for possible inclusion
in the matched-control group, they could not have received a school
nomination for
being
highly
aggressive.
As a
check
on
the validity of the
school nominations, pair-wise observations of each aggressive-control
pair were conducted over a
4-day
period, with extensive observations
daily
over
two
contexts.
These observations (not reported in this article)
indicated that the aggressive subjects and nonaggressive controls
differed markedly in observed aggressive interchanges in the fourth and
seventh grades.
Measures
Multiple assessment procedures
were
used in order to obtain conver-
gent evidence on the primary hypotheses {following the research strat-
egy outlined in Cairns, 1986, and in Cairns
&
Cairns, 1988). The mea-
sures included in this report
were
as
follows: peer reports of the individ-
ual's role in the school social networks; a social cognition interview on
recent conflicts; Interpersonal Competence Scale-teacher (ICS-T) tests,
which yielded factor scores on aggression, popularity, and academic
competence factors; Interpersonal Competence Scale-self (ICS-S) tests,
which yielded self-perception factor scores on the same factors; peer
nominations for conflict
instigation;
and
peer
affiliations,
as
determined
by "best friend" choices. These procedures and the measures that they
generate have been described elsewhere (e.g., Cairns & Cairns, 1984;
Cairns, Perrin,
&
Cairns,
1985).
The measures included
the
following:
Social network
assessment.
An individual, tape-recorded interview
was conducted with each subject. A semistructured protocol was fol-
lowed by the interviewer. The social networks in which subjects were
involved were plotted on the basis of information obtained from sub-
jects and peers in four sections of the interview.
In one section, subjects and peers were asked, "Now tell me about
your
(school,
class).
Are there
some
people who hang around together
a
lot?"
Follow-up probes elicited specific information about
the
perceived
makeup of
the
clusters. As indicated in Cairns et
al.
(1985), high levels
of intersubject agreement were obtained in the identification of cluster
membership. In a second part, subjects were asked about any persons
who were "not members of any groups." The persons named by the
subject, including themselves, were counted as having been nominated
for social isolation. In a third section, subjects were asked to identify
their "best friends." For peer-friendship analyses, all choices of each
child were listed. By comparing friendship choices across subjects, it
was possible to determine which friendships were reciprocated and
which were not. In a fourth part, subjects were asked to nominate per-
sons (both male and female) who gave them "trouble" or "bothered"
them. These nominations were followed by a request to provide a de-
tailed and concrete account of a recent conflict with a peer (both male
and female).
Social
cluster
identification.
In a preliminary study in this series,
Cairns, Perrin, and Cairns (1985) used a decision rule procedure in
order to identify the peer clusters in a social network and the relations
among persons within each cluster. In the decision rule method, each
respondent generated a "social map" for persons in their school grade.
Because there
was
typically
a
high level of agreement across informants,
it was possible to combine information across respondents and build a
"composite social map" of the groups that existed in the classroom. It
was a
decision
rule
procedure because arbitrary standards
were
adopted
(a) to order the cases into groups, (b)
to
judge
whether borderline cases
belonged to one or more social clusters, and (c) to determine which
all analyses were conducted with and without age as a covariate. In no
instance did covariance analyses (controlling for age) yield outcomes
that
were
significantly different than the
ANOVAS.
persons were central to the cluster (nuclear members) and which were
on the edge of membership (peripheral members).
The present quantitative procedure evolved from the decision rule
method.
It was
introduced
in
order
to
develop
a broadly applicable tech-
nique for describing relations between persons and the structure of
so-
cial networks, with minimal reliance on intuitive
judgments.
(A sepa-
rate article, Cairns, Kindermann, & Gariepy, 1986, describes the
method and alternative quantitative procedures.) In
brief,
four succes-
sive matrices were constructed for each classroom in order to arrive at
the latent structure of the classroom
networks.
First,
a raw
recall matrix
was constructed from the free recall of social groups by all subjects
(male and female) in the classroom: each subject-respondent indicated
which persons in the school belonged to which groups (see Table
1
in
Cairns et
al., 1985,
for one recall matrix where each column
refers
to a
different respondent and the children-to-be-clustered are listed down
the rows).
Second, each
raw
recall
matrix
was
transformed to
a
cluster co-occur-
rence matrix (i.e., a symmetric matrix that summarized the frequency
with which each person was named to the same group as each other
person in the school and where the cells indicate the number of times
two persons "co-occurred" in the same cluster). The rows of the co-
occurrence matrix consisted of
all
children-to-be-clustered (including
the subjects themselves), and the columns of the matrix were the same
as the
rows.
The entries summarized the number of occasions that Per-
son / has been identified by respondents to be in the same social group
as Person
j.
The diagonal of the co-occurrence matrix contained the
number of
occasions
that a given person was named to any group (i.e.,
subjects are considered to be members of all groups to which they have
been named).
Third,
a
correlation matrix
was
generated by intercorrelating the col-
umns of
the
arrays of
the
co-occurrence matrix. Each correlation re-
flected the level of correspondence between the cluster membership
scores of Subject A with those of Subject B. Transitivity typically was
found in groups of individuals. That
is,
if Subject A obtained a signifi-
cant positive correlation in patterns of co-occurrence with both Subject
B
and Subject
C,
then Subject
B
typically obtained a significant positive
correlation with Subject C. This empirical transitivity facilitated the
assignment of individuals to preliminary social clusters (i.e., persons
who were judged
by
their
peers
to "hang around together").
Fourth, a Lambda-X (LX)
LISREL
matrix
was
constructed on the ba-
sis of the preliminary cluster descriptions, where each individual-to-
be-clustered was treated as an "observed" variable, and the "latent"
variables
were
clusters of persons. The aim of the
LISREL was
to clarify
the social structure in those classes where the social clusters were not
clearly defined, often because individuals may appear simultaneously
in two or more
clusters.
In establishing the initial LX matrix, each per-
son was declared free for a
single
latent variable (cluster) and fixed for
all other latent variables
(clusters).
In the
final
step,
LISREL VI
estimates
of optimal factor loadings on the LX matrix
were
determined (Joreskog
& Sorbom, 1984). When adjustments to the LX model were required,
they
usually involved relaxing the
model
so
that
a
subject could simulta-
neously appear in two latent variables
(i.e.,
to become members of two
social clusters at the same time). Decisions about relaxing the model
or relocating subjects were made on substantive, a priori grounds by
inspection of
the
raw recall matrix. The goodness-of-fit measure from
the
LISREL
solution was consulted as a guide to determine whether the
fit was substantially improved by the adjustment.2 These "latent vari-
2 The
LISREL
Goodness-of-Fit Index (GFI), root mean square resid-
ual
(RMR),
and other parameter estimates were used primarily for de-
scription in the present application. The properties of the correlational
matrices derived from the co-occurrence matrices preclude standard
LISREL
inferential interpretations. Nonetheless, the
LISREL VI
program
provides
a
guide
for comparing
the
effects of relocating individuals from
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
818CAIRNS, CAIRNS, NECKERMAN, GEST, GARIEPY
ables"
defined the social network in each extended classroom ("ex-
tended" because the method permitted persons in the school other than
those in the classroom to be cluster members).
After
the
social clusters in each classroom
were
identified, the relative
centrality of each cluster and of
each
member of
the
separate clusters
was
determined.
The
index
to
centrality
was
simply
the
number of times
that a given person was named to a cluster. Using the average of the two
persons in the cluster
who
received the highest number of nominations,
the rank of
the
cluster was determined {i.e., high-, medium-, and low-
salient clusters). Similarly, nomination frequency was used to deter-
mine the status of individuals within their
clusters:
nuclear, secondary,
or peripheral. Clusters (or persons) in the upper
30%
rank of nomina-
tions were considered to be high salient (or nuclear rank), those in the
lowest
30%
were considered to be low salient (or peripheral rank), and
those in the mid-range 40% were considered to be medium salient (or
secondary rank). In sum, the method is a quantitative technique that
yields information about (a) the social clusters within each classroom,
(b) the identity of persons who are members of each cluster and their
relations to each other, (c) the relative centrality of each cluster, (d) the
relative rank of persons within the clusters, and (e) the extent of dual
cluster memberships.
For replication and reliability purposes, the entire network analysis
was
repeated for all designated classrooms
using
the
decision rule analy-
sis
described in Cairns et
al.
(1985). The decision rule method of initial
cluster estimation is relatively easy to
compute:
It preserves the identity
of the other members of
the
social network and captures much of the
salient information about classroom relationships. (It does not involve
the construction of the co-occurrence, correlational, and
LX
matrices.)
In the present data set, the two analyses—quantitative and decision
rule—yielded virtually identical results and identical summary conclu-
sions.
The quantitative method, although more laborious and expensive
in computer time, has an advantage in objectivity and freedom from
subjective
decisions
about cluster
membership.
It
also
provides a quanti-
tative guide for deciding whether the solution
is
improved
by
permitting
joint cluster membership for particular
persons.
Accordingly, the results
of the quantitative method are reported here.3
Peer
nominations
for
conflict
instigation.
In the individual inter-
view, subjects were asked the questions, "Has anybody bothered you
recently or caused you any trouble? Or made you
mad?"
Depending on
the sex of the individuals who were identified in the first question, the
child was asked about the opposite sex and whether any boys or girls
had bothered them or caused them trouble. In follow-up probes, the
names of the other persons were established. This information
was
tape
recorded, transcribed, tabulated, and coded with the aid of class and
one cluster to another
as
well
as the effects of joint cluster membership.
For purposes of description, the mean GFI across
63
classroom analyses
(32 classrooms for boys and
31
classrooms for
girls)
was relatively high
(.92), and the mean
RMR
was low (0.14), indicating that the
LISREL
vi
models of networks provided a close fit to the observed correlational
matrices. It is of interest to observe that no grade differences were ob-
tained in these measures of network definition, but
girls
obtained a sig-
nificantly higher mean GFI: for boys M
=
.90 and for girls M
=
.94,
F(
1,
29) =
9.00,
p <
.01.
Similarly, there
was
a significant
sex
difference
in the average
RMR,
SO
that expected/observed differences were smaller
for
girls
than for
boys:
for
boys
M
=
.
16
and for
girls
M
=
.
12, F\
1,29) =
6.50, p <
.05.
The network analyses for the 31 boys were individually
matched (by classroom) to the network analyses for the 31 girls (i.e.,
the single all-male classroom was eliminated for the purposes of this
comparison). This sex difference in the definition and clarity of
social
networks
is
consonant with the
idea
that
girls have more easily
identified
social clusters and group membership than boys (Gilligan, 1982). It is
also consistent with the more frequent use of social ostracism
by
girls.
school enrollment lists (there was more than 99% agreement by inde-
pendent judges). The number of occasions that a given subject was
named by peers as having caused a conflict was summarized over all
male and female respondents in order to obtain a peer conflict nomina-
tion score. Because entire schools were sampled in the fourth and sev-
enth grades, it
was
possible to compute the z
score
for the peer nomina-
tions of each subject
(i.e.,
every subject
was
compared with others of the
same sex in the schools in which they were enrolled). Conflict nomina-
tion scores
were
available from female
peers,
male
peers,
and summary
female and male scores.
Interpersonal Competence
Scale-Teacher.
The subject's teacher (or
coteachers) completed the Interpersonal Competence Scale (Cairns &
Cairns, 19
84).
The
ICS-T consists
of
15
items
that pertain to aggressive-
ness,
popularity, affiliation, and academic competence. Each item re-
quires
the
respondent
to
describe the subject on
a
7-point
scale.
A
factor
analysis (varimax rotation) of
the
ICS-T items indicates that three dis-
tinct factors emerge with high levels of communality in all age-sex lev-
els.
The three item clusters were aggression, popularity, and academic
competence.
A LISREL
measurement model indicated an excellent
fit
of
the items to these three factors. For example, at the seventh-grade level,
the hypothesized structural equation with three latent variables yielded
a Goodness-of-Fit Index of
.98,
The chi-square with
17
degrees of
free-
dom was 13.83 (p = 0.67), confirming the goodness-of-fit
analysis.
This
LISREL
solution
was
representative of
those
conducted at the other
grade
levels,
indicating that three separate factors (aggressiveness, popularity,
and academic competence) may be reliably identified in the ICS-T
across this
age
range.
Of special importance to this research was the ICS-T factor of "ag-
gressiveness"
(which consisted of
three
items;
namely, "gets
in
trouble
at
school," "fights a lot," and "always
argues").
The concurrent interrater
reliabilities for the aggressive factor scores in assessments were r(35) «
.82,
and r(26) = .78 (Fisher's Z-averaged r = .81). These two-person
interrater reliabilities compare favorably with those reported in other
studies (e.g., Olweus, 1979) and prior reliabilities reported using the
present scale (Cairns & Cairns, 1984). The construct validity of this
aggressive factor score and its linkage to other external measures of
ag-
gressive behavior has been established in companion studies (Cairns &
Cairns, 1984,1988). Comparable reliabilities
were
obtained for
the pop-
ularity and academic factors. The items were reseated on the
7-point
scale so that the
lower
the
ICS-T factor
score,
the higher the social desir-
ability. That
is,
for the aggressive factor, high factor
scores
reflected high
aggressiveness; for the popularity factor, low factor scores reflected high
popularity.
Interpersonal
Competence
Scale-Self.
Each
year
subjects completed
a self-descriptive test form of the Interpersonal Competence Scale
(ICS-
S).
A booklet was prepared with the items printed on separate pages.
Other than the inclusion of distractor items and booklet format, the
subjects' tests (ICS-S) were identical with those completed by adults
(ICS-T). The scales themselves had been developed through extensive
pilot
testing.
The vocabulary and instructions were within the range of
fourth-grade students. If necessary for comprehension, the interviewer
read the item aloud. The same items
were
used across all
years,
and the
year-to-year stabilities are reliable (Cairns et
al.,
1988).
Procedure
Subjects
were
interviewed and tested in the school that they attended.
Confidentiality was assured, and subjects were told that they could de-
cline to
answer
any
question
or
withdraw at any
point.
At the
conclusion
of the interview, they were given the choice of a school-related item
obtained from one of the universities in the region (e.g., notebook, pen,
3 A detailed description of the method and rationale will be made
available on request (Cairns, Kindermann,
&
Gariepy, 1986),
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
SOCIAL SUPPORT OR SOCIAL REJECTION?819
pencil). Teachers were given minimal instructions on rating, except to
use the full range of the scale if appropriate. Society for Research in
Child Development ethical standards
were
followed in all aspects of the
investigation.
Results
The research outcomes are summarized in three
sections.
In
the first part, behavioral differences and similarities among the
high-aggressive and the matched control groups are described.
The second part focuses on aggressive children and their roles
in the social network. The third part analyzes the role of peer
social structures in the promotion and regulation of aggressive
patterns. Gender similarities and differences are discussed in
each section.
Characteristics of Aggressive and Matched
Control Subjects
Subjects who had been placed in the high-aggressive group
differed from those in the matched control group on several be-
havioral dimensions related to aggression and popularity. As
shown in Table 1, subjects in the aggressive condition scored
higher than those in the control condition on all measures of
aggression used. The aggression-control differences were highly
reliable in the ICS-T (teacher) assessments of aggression,
7*1(1,
36) =
34.68,
p <
.001,
in peer nominations for aggressive con-
flicts, t\\t 36) =
29.62,
p <
.001,
and in the ICS-S (self) ratings
of
aggression,
F{\, 36) = 13.34, p <
.001.
Parallel effects were
obtained in both grades and in both girls and
boys.
For the ICS
aggressive factor measures, only the main effects of condition
(aggression vs. nonaggression) were reliable; no other main
effects or the interactions reached statistical significance. In the
case of the summary peer
nominations,
there
was also a
reliable
Table 1
Characteristics
of the
Aggressive
and
Nonaggressive Groups
as
a Function
of
Grade
and Sex
Table 2
Popularity
of
Aggressive
and
Nonaggressive Groups
as a
Function
of
Grade
and Sex
Subjects/
group
Boys
Aggressive
4th grade
7th grade
Control
4th grade
7th grade
Girls
Aggressive
4th grade
7th grade
Control
4th grade
7th grade
ICS-T
(Teacher)
M
5.83
5.17
3.50
3.80
5,07
5.50
2.60
3.03
SD
1.34
0.67
2.29
1.75
0.71
1.33
1.49
1.75
Measures of aggression
Conflict
(Peer)
M
4.10
1.90
1.20
0.70
2.80
1.90
0.80
0.50
SD
2.56
1.91
1.40
1.25
1.81
2.56
1.03
0.71
ICS-S
(Self)
M
3.95
4.11
3.73
3.47
4.53
4.03
3.23
2.97
SD
0.78
1.15
1.34
1.04
1.13
1.01
0.99
0.96
Subjects/
group
Boys
Aggressive
4th grade
7th grade
Control
4th grade
7th grade
Girls
Aggressive
4th grade
7th grade
Control
4th grade
7th grade
Measures of popularity
ICS-T (Teacher)
M
4.12
4.23
2.73
3.73
4.32
4.53
3.03
3.13
SD
1.73
1.46
1.59
1.23
1.38
1.41
1.09
0.69
ICS-S (Self)
M
2.42
3.45
2.38
3.33
2.80
2.80
2.30
2.35
SD
1.00
1.08
0.80
1.31
1.65
0.86
0.87
0.74
Note.
ICS = Interpersonal Competence Scale.
gender effect, in that boys
were
nominated for causing conflicts
more often than
girls,
F(\f 36) = 4.87,p < .05.4
Aggressive subjects were also less popular than peers in the
control group, as evaluated using the ICS-T popularity factor
scores (Table 2). The main effect of condition (aggression vs.
control) was highly reliable, F{\, 36) = 21.08, p < .001. No
interactions were statistically reliable, nor
were
the main effects
of grade or gender. Aggressive subjects were also more fre-
quently disliked
by
peers,
as
inferred from nominations of peers
who saw themselves as being bothered and bullied by them (see
Table 1, "Conflict: peer" column). These outcomes are consis-
tent with previous
findings
that aggressive subjects are likely to
be seen as generally unlikable and unpopular (e.g., Coie &
Dodge, 1983). Self-ratings on popularity, however, yielded par-
allel outcomes in the two groups. Aggressive subjects rated
themselves to be as popular as control subjects in the ICS-S as-
sessment. Moreover, there were no differences between the ag-
gressive subjects and matched-control subjects in the number
of occasions that peers named them
as
best friend (see below).
Aggression, Social Roles, and Rejection
Two analyses of the social roles of highly aggressive and con-
trol subjects were permitted by the data: peer-defined social
clusters and peer judgments of isolation/rejection from the so-
cial network.
Social cluster analysis. The social clusters in the two
Note.
ICS = Interpersonal Competence Scale.
4 Beyond the main effects attributable to the aggressive-control dis-
tinction, further
analysis
of the
peer
nominations indicated that
(a)
boys
were reliably less likely to nominate girls than vice-versa and (b) there
was a significant decrease in total nominations by girls as they grew
older. In all nomination indices—whether generated by boys or
girls—
the differences between the aggressive- and matched-control groups
were highly significant (p
<
.005).
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
820
CAIRNS, CAIRNS, NECKERMAN, GEST, GARIEPY
different grades were identified by
peers.
Confirming the results
of preceding work, the social clusters identified in the cognitive
social maps were predominantly (a) same-sex in composition
and (b) reliably identified across individuals, even when persons
were not themselves members of the clusters (Dunphy, 1963;
Cairns et al., 1985). The cluster identification method was used
to determine whether aggressive (or control) subjects were full
members of the extent social clusters, peripheral members of
social clusters, or were not named as members of any social
cluster. The results of this classification are shown in Table 3.
For summary purposes, subjects who were nuclear members of
high salient
clusters
were classified as
nuclear,
subjects who were
secondary members of high-salient clusters or nuclear/second-
ary members of medium-salient clusters were classified as
sec-
ondary,
and subjects
who
were in low-ranked clusters or periph-
eral members of any cluster, regardless of
status,
were classified
as
peripheral.
Subjects not named at all were classified as iso-
lated.
No reliable differences were obtained between subjects in the
aggressive and control conditions in type of cluster member-
ship,
using chi-square analyses. Across all age-sex groups, 95%
(38 of 40) of the highly aggressive subjects were members of
some social clusters, as identified by
peers.
Approximately one-
third (12 of
40)
were nuclear members of high-salient groups,
and another
45%
were secondary members in the social network
(i.e.,
secondary members of high-salient groups or nuclear/sec-
ondary members of middle-salient groups). One-fifth of the to-
tal (20%) were peripheral members of
the
social network (i.e.,
members of low-status groups or peripheral "hangers on" to
high-salient groups). Only 5% seemed to be removed from the
social structure. These proportions
were
virtually identical with
those of
the
matched control subjects, except that none of the
control subjects were perceived to be removed entirely from the
social structure (the difference is not reliable). No difference
Table 3
Cluster Membership of
Aggressive
and
Nonaggressive Subjects
as a
Function
of
Grade
and Sex
Grade/group Nuclear Secondary Peripheral Isolate
Table
4
Mean Nominations for Social Isolation
by
Peers
by Sex,
Grade, and Aggressive Group Status
4th
Aggressive
Girls
Boys
Control
Girls
Boys
7th
Aggressive
Girls
Boys
Control
Girls
Boys
Sum
{4th
and 7th)
Aggressive
Control
Total
.10(1/10) .60(6/10) .30(3/10) .00(0/10)
.40(4/10) .40(4/10) .20(2/10) .00(0/10)
.20(2/10) .70(7/10) .10(1/10) .00(0/10)
.20(2/10) .50(5/10) .30(3/10) .00(0/10)
.40(4/10) .40(4/10) .10(1/10) .10(1/10)
.30(3/10) .40(4/10) .20(2/10) .10(1/10)
.50(5/10) .30(3/10) .20(2/10) .00(0/10)
.50(5/10) .30(3/10) .20(2/10) .00(0/10)
.30(12/40) .45(18/40) .20(8/40) .05(2/40)
.35(14/40) .45(18/40) .20(8/40) .00(0/40)
.33(26/80) .45(36/80) .20(16/80) .03(2/80)
Subjects
Boys
4th grade
7th grade
Summed over grade
Girls
4th grade
7th grade
Summed over grade
Aggressive
condition
M
1.20
0.60
0.90
1.80
0.80
1.30
SD
1.32
1.26
1.29
3.08
1.23
2.35
Nonaggressive
condition
M
1.00
0.30
0.65
0.70
1.80
1.25
SD
0.82
0.48
0.67
1.89
2.53
2.23
in category representation between
the
aggressive
and
control
groups
was
statistically significant
as
determined using
chi-
square analyses (Table 3).5
Given that highly aggressive subjects were more likely
to be
nominated by peers for causing conflicts,
it
seems reasonable
to
expect that fewer peers would wish
to be
affiliated with them.
This would lead
to a
reduction
in
the overall size of the clusters
in which aggressive subjects were members. There is some mod-
est support
for
this expectation. Among subjects
who
were
members of social clusters, the mean numbers of persons
in
the
clusters
of
the highly aggressive
and
matched control subjects
were
5.11 and 5.72,
respectively. However, this difference
was
not statistically reliable,
F(\, 34) =
2.84,
p>
.05. Although ag-
gressive girls tended
to be in
smaller clusters than control girls,
none
of
the main effects (sex, grade,
or
risk condition)
or
their
interactions were statistically significant.
Peer judgments
of
isolation.
In
addition
to
describing
the
social structure,
all
subjects were asked
to
specify
any
persons
who
did not
belong
to any
social group (i.e., those
who
were
rejected
or
isolated).
The
number
of
times that each aggressive
subject and his
or
her matched control was named as being out-
side the social network was determined. The means for the sex-
age-aggressive groups
are
shown
in
Table
4. The
peer judg-
ments
of
social isolation indicated
no
reliable effects, either
as
main factors
or in
interaction.
Peer judgments of isolation may be used to identify particular
isolated subjects.
A
conservative estimate
of
isolation
may be
obtained
if
a given subject was explicitly identified
by at
least
4
of his
or
her peers
as
having no social group. Overall, 10%
(4 of
Note. Numbers in parentheses represent the number of subjects (out of
total number) rated as being in the various clusters.
5 The clusters in Table
3
cumulated information from all respondents
in
the
study, including
the
subjects themselves.
An
argument
can be
made that the individual's own self-assignments should be omitted be-
cause of possible self-enhancement distortions, particularly
among
sub-
jects
who
are peripheral
(Cairns,
Neckerman,
&
Cairns,
in
press).
When
the clusters were recomputed without the subjects' self-assignments,
no
reliable changes were observed.
As
expected, more persons became cat-
egorized as
isolated when
the additional respondent (i.e., the self) was
eliminated
(5
of 40 in the combined aggressive group and
1
of 40 in the
combined control group; 8% overall
in
combined groups). The differ-
ences were not statistically significant using chi-square analyses, either
overall or in specific age-condition comparisons.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
SOCIAL SUPPORT
OR
SOCIAL REJECTION?
821
40)
of the
highly aggressive subjects
and 8% (3 of 40) of the
nonaggressive control subjects qualified
for
this conservative
judgment of isolation.
The
difference was
not
reliable.
Peer Social Organization and Aggression
The question addressed in the next set of analyses concerned
whether highly aggressive children and adolescents get together
to form cliques. Do aggressive children tend to hang around
together? This matter was investigated by intraclass correla-
tional analyses of the social clusters identified through the quan-
titative network analysis. To determine if the clusters them-
selves differed in mean levels of aggressive expression, the ag-
gressive factor scores
(i.e.,
ICS-T) of the nuclear members of
each group
were
determined.6 An intraclass correlational anal-
ysis of the similarity of ICS-T scores of the nuclear members of
each social cluster permitted a determination of within-cluster
similarity on aggressive
scores.
Intraclass coefficients
were
com-
puted separately for
the two
grades and
two
sexes
for
the
clusters
in all
classes.
All
levels
of clusters
(high,
medium, and
low)
were
included in the analysis.
The intraclass coefficients using the 1CS-T aggressive factors
indicated that nuclear members of the male clusters in the
fourth grade were highly similar in terms of their ratings of
ag-
gression: the intraclass correlation was .75, F\22, 29) = 8.57,
p <
.001.
For fourth-grade
girls,
however, the intraclass correla-
tion for aggression was not significant, r'
—
.03, F{26, 39) ~
1.07, p
>
.
10.
The same analysis was completed for the seventh
grade. In the early adolescent cohort, the intraclass correlation
for aggression was significant both for
males,
r' ~ .43, F(48,
78) =
2.97,
p<
.001,
and for females, f
=
.37,
F(54,97) = 2.60,
p<.001.
''Best
Friend"
Analysis
An alternative technique
for
analyzing social networks
in-
volved the use of the "best friend" information. From each sub-
ject's protocol, persons
who
were named
as
best friends were
determined.
By
cross-reference,
it was
possible
to
determine
who reciprocated
the
selection (i.e., whether
the
best friend
himself or herself named
the
subject as best friend). This analy-
sis
was
completed separately
for the
best friend selections
of
boys
and
girls
in the
two cohorts.
The
mean scores
on the
ICS-
T aggressive factor were determined
for
reciprocated
and non-
reciprocated friendships. The individual's own aggressive factor
scores were correlated separately with
the
mean aggressive
scores of both reciprocated
and
nonreciprocated friends.
The results of this analysis
for the
fourth-grade cohort indi-
cate
a
high relationship
in the
aggressive factor scores of males,
but only if the best friend choices were reciprocated (Table
5).
That
is, the
correlation between
the
ICS-T aggressive factor
scores
for
boys whose friendship choices were mutual
was .61
(p < .01), whereas the aggressive scores of nonreciprocated best
friend choices were
not
reliably correlated,
r -
.12,
p >
.10.
If
one
did not
have information about friendship reciprocation
and
all
best friends named
by the
subject were considered
(as
would
be the
case
if
the friendship choices
of
the best friends
had
not
been determined), similarity
on the
aggressive factor
scores would remain statistically significant
for
fourth-grade
Table
5
Relationships Between the Subject's ICS-T Aggressive Factor
Score and Mean ICS- T Scores
of'
'Best Friends''
Friendship reciprocity
Subjects
Boys
4th grade
7th grade
Girls
4th grade
7th grade
Reciprocal
.61*
.63*
.07
.51*
Nonreciprocal
.12
.40*
.19
.12
Sum
.41*
.49*
.08
.34*
Note.
ICS-T = Interpersonal Competence Scale-Teacher.
*/? <
.01
(product-moment correlation).
males,
r =
.41,
p <
.01.
The
seventh-grade cohort boys repli-
cated
the
findings from
the
younger
boys:
r
=
.63,
p <
.01,
for
the reciprocated pairs.
In the
older male group, however, even
the nonreciprocated best friends
had a
reliable similarity with
the subject
in
terms of ICS-T aggression factor scores,
r
=
.40,
p
<
.01.
For
girls
in the
fourth-grade cohort, there were
low
levels
of
similarity between
the
mean scores
of
best friends
on
the ICS-T aggressive factor, regardless of whether the friendship
selections were reciprocated (Table 5).
In the
seventh-grade
co-
hort, however,
the
best friends of girls were similar
in
aggressive
factor
scores.
Again,
the
relationship was strongest if the female
best friend choices were reciprocated.
The
data
are
consistent
with
the
cluster analyses,
in
that girls
in
early adolescence tend
to affiliate with other girls who
are
similar on aggressive expres-
sion.
Finally, subjects
in the
highly aggressive group
did not
differ
from
the
subjects
in the
matched-control group
in
terms
of
number
of
reciprocated choices
as
best friend.
The
same
per-
centage (43%) of highly aggressive subjects
as
matched control
subjects received reciprocated best friend choices.
The
groups
did
not
differ
in the
number
of
times that they were selected
by peers
as a
best friend
nor in the
number
of
times that their
friendship choices were reciprocated.
Discussion
Highly aggressive girls
and
boys were usually solid members
of peer clusters
in the
fourth
and
seventh grades
and
they typi-
cally
had a
network of friends.
The
finding that aggressive ado-
lescents have lower levels of general popularity
and
likability—
findings that were replicated
in
this report—may serve
to ob-
scure
the
"concealed competencies"7 that permit these persons
to survive
in
particular social contexts. Being popular with
the
6 The ICS-T measure was used because of its demonstrated stability
and relationship to subsequent problem behaviors. Other "external"
measures, including peer nominations, yield parallel results.
7 The term
concealed competence
has been used by Norman Gar-
mezy to refer to unrecognized and perhaps unrealized skills possessed
by persons, especially individuals who have been deemed delinquent,
deviant, retarded, or socially incompetent (private communication,
May 15,1988).
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
822CAIRNS, CAIRNS, NECKERMAN, GEST, GARIEPY
group as a whole may not be the only goal for adolescents, and
failure to achieve broad-based popularity should not be taken
as evidence of wholesale social rejection. Furthermore, aggres-
sion and correlated life-styles appear to provide a significant
basis for peer affiliations.
These findings are consistent with the observations of Gior-
dano (e.g., Giordano et al., 1986) on the cohesion of deviant
groups. The results are also in accord with earlier
findings
that
indicated that some "controversial" aggressive children are
both rejected and accepted by peers (Coie et al,, 1982; Coie &
Dodge, 1983). In the present study, aggressive children in both
cohorts and in both sexes may be disliked by some classmates
for legitimate reasons (bullying, ridiculing, or
victimizing).
But
dislike by certain peers is not equivalent to social rejection or
isolation from the entire social structure. Highly
aggressive
ado-
lescents may alienate many peers, but the relationships that
they establish with some peers seem no less meaningful than
do those of nonaggressive adolescents. In any case, the highly
aggressive subjects in this study were typically members of
so-
cial clusters, and they were as likely as matched control subjects
to have a coterie of reciprocal "best friends."
A dark side to coercive cliques in adolescence requires com-
ment because of its relevance to societal rejection. In the course
of development, coalitions of aggressive adolescents might be
expected to come into conflict with adults as well as peers and
to threaten the existing order. In the competition for hegemony
in the school context, aggressive coalitions of students can dev-
astate the authority of adults. Failures by the persons in charge
of the school to abort the formation of such groups or to regu-
late them could be an abdication of responsibility. Hence, coali-
tions among aggressive adolescents demand the attention of
teachers and principals. Implicit and explicit rejection of ag-
gressive adolescents by school personnel should, in the
long
run,
catalyze the attitudes of
peers
who have been victimized. Over
time,
members of coercive peer groups should be more likely to
drop out of
school,
or to be forced out through suspension and
expulsion.
Turning to sex similarities and differences, it is noteworthy
that aggressive patterns—and correlated behaviors—provided
a basis for social cohesion and commonalities in friendships for
both boys and
girls.
At
the
outset of the study, it seemed reason-
able to expect that anticipatory sex differences would arise be-
cause of the sanctions against direct physical aggression in girls.
Romantic and sexual interests become of greater importance
for girls than boys (e.g., Cairns et al., 1988), and the normative
criteria for feminine attractiveness ordinarily do not include
aggressiveness (Gilligan, 1982). Accordingly, fourth-grade girls
showed no propensity to sort out their friends on dimensions
correlated with aggressive behavior. On the other hand, the
present data clearly show that, by adolescence, aggressive ho-
mophily is almost as strong for girls as it is for
boys.
These find-
ings do not stand
alone.
Both Magnusson (1987) and Giordano
et al. (1986) found that deviant teenage girls tend to affiliate
with other girls who match their acting-out behaviors. By ad-
olescence, the external constraints with respect to aggressive
affiliation may be relaxed for females, possibly because such
groups of girls are perceived to be less threatening than gangs
of adolescent
males.
Alternatively, affiliations among aggressive
adolescent girls may be caused by their exclusion (i.e., ostra-
cism) from other, more acceptable
groups.
All this
is
to empha-
size that social clusters are not merely the creation of children
and adolescents: The broader social community of teachers, ad-
ministrators, and parents probably contributes to the forma-
tion and dissolution of social clusters at all ages.
A comment on methodology
is
in order. Although the quanti-
tative method adopted here for analyzing social networks re-
quires multiple steps, most of the information on cluster mem-
bership can be abducted by a brief inspection of the raw recall
matrix. There is considerable agreement among children and
adolescents in their perceptions of whom was associated with
which cluster. The key to network analyses—whether quantita-
tive,
geometric (Moreno, 1934), rule-guided (Cairns, et al.,
1985),
or abductive (Dunphy, 1963)—is to capture the interre-
lations among persons as they exist in groups and subgroups.
These solutions should be convergent, as the present data sug-
gest. Methods that focus on the popularity of individuals as op-
posed to the dynamics of networks may capture different
sources of variance (Cairns, 1983). Specifically, recent methods
for identifying "social status" on the basis of pooled ratings of
how peers
like,
dislike,
or ignore
the
subject typically
reveal
little
about
the
person's placement in a network of relationships
(e.g.,
Asher & Dodge, 1986). One contribution of the present study
has been to describe a quantitative procedure that can be
broadly applied to networks in classrooms and other social set-
tings.
This work was completed on a representative sample of non-
urban youths in diverse living circumstances in the 1980s; it is
doubtless limited in generality by cultural and temporal con-
straints. The problems addressed, however, are universal. By
certain standards (e.g., socioeconomic status, sex distribution,
racial representation), this sample is not unlike that repre-
sented in the national population. Moreover, generality is sug-
gested
by
key similarities between our results and those of previ-
ous researchers once the difference in methodology and con-
structs is acknowledged (e.g., Asher & Dodge, 1986; Coie &
Dodge, 1983). The present outcomes are also consistent with
recent developmental (Parks & Slaby, 1983) and sociological
(Giordano et al., 1986) perspectives on reciprocities in aggres-
sive behavior. Aggressive adolescents may be unpopular in the
larger social community of peers and adults, yet they can be
accepted by and closely linked to the particular subgroups of
peers.
Further analysis of this phenomenon should be useful in
closing the gap between the sociological focus on delinquent
"gangs"
and the psychological emphasis on normal peer groups
(Hartup, 1983).
References
Asher, S. R., & Dodge, K. A. (1986). Identifying children who are re-
jected
by
their
peers.
Developmental
Psychology,
22,
444-449.
Barrett, D.
E.
(1979)
A
naturalistic study of sex differences in children's
aggression.
Merrill-Palmer
Quarterly,
25, 193-207.
Bronfenbrenner, U. (1944a). A constant frame of reference for socio-
metric research.
Sociometry,
6,363-397,
Bronfenbrenner, U. (1944b). A constant frame of reference for socio-
metric research: Part 2. Experiment and inference.
Sociometry,
7,
40-75.
Bronfenbrenner,
U.
(1979).
The
ecology of
human
development:
Experi-
ments
by natural
design.
Harvard University Press.
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
SOCIAL SUPPORT OR SOCIAL REJECTION?823
Cairns, R. B. (1979). Social
development:
The
origins
and
plasticity
of
social
interchanges.
San Francisco: Freeman.
Cairns, R.
B.
(1983). Sociometry, psychometry, and social structure: A
commentary on six recent studies of popular, rejected, and neglected
children.
Merrill-Palmer
Quarterly,
29,
429-438.
Cairns, R. B. (1986). Phenomena lost: Issues in the study of develop-
ment. In J. Valsiner (Ed.), The
individual subject
and
scientific
psy-
chology
(pp.
97-112).
New
York:
Plenum.
Cairns, R. B.,
&
Cairns,
B.
D. (1984). Predicting aggressive patterns in
girls and
boys:
A
developmental
study.
Aggressive
Behavior,
11,
227-
242.
Cairns, R. B., & Cairns, B. D. (1988). The sociogenesis of
the
self.
In
N.
Bolger, A. Caspi, G. Downey, & M. McClintock (Eds.),
Persons
in a
social context
(pp. 181-202). New
York:
Cambridge University
Press.
Cairns, R. B., Cairns,
B.
D., Neckerman, H. J., Ferguson, L. L.,
&
Ga-
riepy, J.-L. (in press).
Growth
and
aggression:
1.
Childhood
to
early
adolescence.
Developmental
Psychology.
Cairns, R. B., Kindermann, T, & Gariepy, L. (1986).
Cognitive
socio-
metry:
How
to
identify peer clusters through composite social
maps.
Chapel Hill: University of North Carolina, Social Development Lab-
oratory.
Cairns, R. B., Neckerman, H. J., & Cairns,
B.
D. (in press). Social net-
works and shadows of
synchrony.
In G. R. Adams, T. P. Gullota, &
R. Montemayor (Eds.),
Advances
in
adolescent
development.
Beverly
Hills,
CA:
Sage.
Cairns, R.
B.,
Perrin, J. E.,
&
Cairns,
B.
D. (1985). Social structure and
social cognition in early adolescence: Affiliative patterns.
Journal of
Early
Adolescence,
5,
339-355.
Cohen,
A.
K.
(1955).
Delinquent
boys:
The
culture
of the
gang.
Glencoe,
IL:
Free Press of Glencoe.
Coie,
J.,
&
Dodge,
K.
A.
(1983). Continuities and changes in children's
social
status:
A
five-year
longitudinal
study.
Merrill-Palmer
Quarterly,
29,261-282.
Coie,
J. D., Dodge, K. A., & Coppotelli, H. (1982). Dimensions and
types
of social
status:
A cross-age
perspective.
Developmental
Psychol-
ogy,
18,
557-570.
Dunphy, D. C. (1963). The social structure of urban adolescent peer
groups.
Sociometry,
26,
230-246.
Gilligan,
C.
(1982). In a
different
voice.
Cambridge,
MA:
Harvard Uni-
versity Press.
Giordano, P. C. (1978). Research note: Girls, guys, and gangs: The
changing social context of female delinquency. The
Journal
of Crimi-
nal Law
and
Criminology,
69,
126-132.
Giordano,
P.
C, Cernkovich, S. A., & Pugh, M. D. (1986). Friendship
and
delinquency.
American Journal
of
Sociology,
91, 1170-1201.
Hall, W. M., & Cairns, R. B. (1984). Aggressive behavior in children:
An outcome of
modeling
or reciprocity?
Developmental
Psychology,
20, 739-745.
Hartup, W. W. (1983). Peer groups. In P. H. Mussen (Series Ed.) and
E. M. Hetherington
(Vol.
Ed.),
Handbook
of child
psychology:
Vol.
4.
Socialization,
personality,
and social development
(4th ed., pp. 103-
196).
New
York:
Wiley.
Hirschi,
T.
(1969).
Causes
of
delinquency.
Berkeley: University of Cali-
fornia Press.
Jbreskog, K. G.,
&
Sorbom,
D.
(1984).
LISREL
VI:
Analysis
of structural
relationships
by
the
method of maximum
likelihood.
Mooresville.IN:
Scientific Software.
Kandel,
D.
B.
(1978). Homophily, selection, and socialization in adoles-
cent friendships.
American Journal
of
Sociology,
84, 427-436.
Magnusson, D. (1987). Individual
development
from an
interactional
perspective.
Hillsdale, NJ: Erlbaum.
Moreno, J.
L.
(1934).
Who
shall survive? A
new approach to
the problem
of human
interrelations.
Washington,
DC:
Nervous and Mental Dis-
ease Publishing.
Olweus, D. (1979). Stability of aggressive reaction patterns in
males:
A
review. Psychological
Bulletin,
86,
852-875.
Parke, R. D., & Slaby, R. G. (1983). The development of aggression. In
P.
H. Mussen (Series Ed.) and E. M. Hetherington
(Vol.
Ed.),
Hand-
book
of child
psychology:
Vol.
4.
Socialization,
personality,
and social
development
(4th ed., pp. 547-642).
New
\brk: Wiley,
Patterson,
G.
R. (1982). Coercive family
systems.
Eugene,
OR:
Castalia.
Peery,
J.
C.
(1979).
Popular, amiable, isolated, rejected:
A
reconceptual-
izatian of sociometric status in preschool children. Child
Develop-
ment,
50,
1231-1234.
Raush, H. L. (1965). Interaction sequences.
Journal
of
Personality and
Social
Psychology,
2,487-499.
Savin-Williams, R.
C.
(1979).
Dominance hierarchies
in
groups of early
adolescents.
Child
Development,
50,
923-935.
Toch, H. (1969).
Violent
men.
Chicago: Aldine.
Yablonsky, L. (1962).
The violent
gang.
New \brk: Irvington.
Received October 14, 1987
Revision received May 24, 1988
Accepted June 7, 1988 •
This document is copyrighted by the American Psychological Association or one of its allied publishers.
This article is intended solely for the personal use of the individual user and is not to be disseminated broadly.
A preview of this full-text is provided by American Psychological Association.
Content available from Developmental Psychology
This content is subject to copyright. Terms and conditions apply.
