You are on page 1of 15

Child Development, July/August 2007, Volume 78, Number 4, Pages 1240 – 1254

Peer Group Status as a Moderator of Group Influence on Children’s


Deviant, Aggressive, and Prosocial Behavior
Wendy E. Ellis and Lynne Zarbatany
The University of Western Ontario

Group status was examined as a moderator of peer group socialization of deviant, aggressive, and prosocial
behavior. In the fall and 3 months later, preadolescents and early adolescents provided self-reported scores for
deviant behavior and group membership, and peer nominations for overt and relational aggression, prosocial
behavior, and social preference. Using the social cognitive map, 116 groups were identified involving 526 children
(282 girls; M age 5 12.05). Hierarchical linear modeling revealed that high group centrality (visibility) magnified
group socialization of relational aggression, deviant behavior, and prosocial behavior, and low group acceptance
magnified socialization of deviant behavior. Results suggest group influence on behavior is not uniform but
depends on group status, especially group visibility within the larger peer context.

As children approach adolescence, the vast majority based groups, and we adopt this approach in the
of their time is spent interacting with peers (Brown, present research. Ethnographic studies of interaction-
1990; Hartup, 1993). These interactions provide chil- based peer groups have clearly demonstrated their
dren with needed social support and companionship significance in children’s lives (e.g., Adler & Adler,
and help to shape and direct children’s behavior. 1995; Eder, 1985). In recent years, evidence for peer
Extant research on peer socialization has focused group socialization has been established in numerous
primarily on the influence of children’s dyadic friend- behavioral domains, including aggression (Cairns,
ships, demonstrating that friends’ behavior and atti- Cairns, Neckerman, Gest, & Gariepy, 1988; Espelage,
tudes generally become more alike over time (Hartup Holt, & Henkel, 2003), deviant behavior (Kiesner,
& Stevens, 1997). Children interact in many distinct Cadinu, Poulin, & Bucci, 2002; Urberg, Degirmencioglu,
social units, however, and by early adolescence the & Pligrim, 1997), and school engagement/attitudes
majority of peer interaction takes place in group (Kindermann, 1993; Ryan, 2001).
contexts (Crockett, Losoff, & Petersen, 1984). Despite our growing understanding of peer group
Child and adolescent peer groups typically have socialization of behavior, questions remain regarding
been defined in two ways. Peer networks, or cliques, the nuances of peer group influence. For example,
consist of three or more peers who regularly associate information is relatively sparse regarding factors that
and have a common environment and set of norms may mitigate or exacerbate group influence. Some
(Brown, 1990). Crowds, which typically emerge dur- promising preliminary evidence regarding child-
ing adolescence, are usually larger than cliques and related moderating factors, such as identification with
form based on reputation only (e.g., ‘‘the jocks’’ or the peer group (Kiesner et al., 2002), status within the
‘‘the brains’’; Brown, 1990). Within crowds, children peer group (Crosnoe & Needham, 2004), and family
do not necessarily experience a shared climate or even environment (Dishion, Patterson, Stoolmiller, &
spend time together. Thus, research examining peer Skinner, 1991), suggests that the effects of group
group influence has most often studied interaction- participation are not expressed uniformly by group
members. However, the potential moderating influ-
ence of group-related factors has not been assessed. In
This article is based on a doctoral dissertation by the first author
under the direction of the second author. Portions of this research the present study we explore group status as a mod-
were funded by a grant from the Social Sciences and Humanities erator of peer group influence on the behavior of
Research Council of Canada, and by a New Research and Scholarly children in the preadolescent and early adolescent
Initiative and Academic Development Fund Award from The
University of Western Ontario to the second author. The authors years. Group status reflects the extent to which the
wish to thank the school personnel and children who participated in group is in a dominant position in the larger peer
this research, and the many students who helped gather and network (Gest, Graham-Bermann, & Hartup, 2001). We
process the data.
Correspondence concerning this article should be addressed to
Wendy Ellis, King’s University College at The University of Western
Ontario, 266 Epworth Ave., London, Ontario, N6A 2M3. Electronic # 2007 by the Society for Research in Child Development, Inc.
mail may be sent to wendy.ellis@uwo.ca. All rights reserved. 0009-3920/2007/7804-0013
Peer Group Status as a Moderator 1241

justify our selection of this important group variable groups identify with their groups (Ellemers, 1993;
after first describing the peer group socialization pro- Tarrant, 2002) and select their own group members for
cesses that likely contribute to the group status effect. activities more than do members of low-status groups
(Horn, 2006). In-group favoritism coupled with a high
degree of group-motivated behavioral conformity
Processes of Group Influence
likely help maintain positive group identity and pro-
Mechanisms of group influence have not been tect the power that high-status groups provide.
investigated extensively, but active and passive group We follow extant research (Adler & Adler, 1995;
socialization processes have been described by theo- Rodkin, Farmer, Pearl, & Van Acker, 2000) in our
rists. Adler (1969) and others (e.g., Baumeister & primary conceptualization of group status as reflect-
Leary, 1995) have argued persuasively that humans ing visibility and dominance, but we also explore an
have a fundamental need to feel connected with the alternative conceptualization of group status sug-
social world. For children, this need likely stimulates gested by research on status at the individual level
friendship and peer group formation. Proponents of (Parkhurst & Hopmeyer, 1998). Specifically, we con-
social identity theory further suggest that peer group trast the socializing effects of groups that are visible
membership plays a critical role in self-appraisal, and well known (group centrality) with those that are
with members motivated to create and maintain the well liked (social preference). At the individual level,
norms of their group to achieve a positive group only modest overlap has been observed between
identity (Tajfel, 1972; Tarrant, 2002). This self- popular children identified based on these two mod-
socialization process has been posited to explain els of status (Parkhurst & Hopmeyer, 1998), and
conformity in children’s peer groups and the devel- different behavioral correlates have been associated
opment of adolescent risk-taking behavior (Harris, with each (Gest et al., 2001). Thus, visibility does not
1995; Stewart-Knox et al., 2005). guarantee liking, and different skills may be impli-
A second form of group influence casts group cated in achieving these two social outcomes
members as passive recipients of processes such as (Cillessen & Mayeux, 2004). To the extent that visibil-
peer pressure, manipulation, and social reinforcement ity and liking are group characteristics that enhance
applied by other group members to ensure that group- social privilege, we would expect group socialization
valued behaviors are acquired and maintained (e.g., efforts to be emphasized in behavioral domains that
Brown, Clasen, & Eicher, 1986). For example, Dishion, contribute to status maintenance. That is, visible
Spracklen, Andrews, and Patterson (1996) coined the (central) groups should promote behavior that en-
term deviancy training to explain friends’ positive hances group visibility, whereas well-liked (socially
reinforcement of delinquent behavior. In their study, preferred) groups should promote behavior that
delinquent friends reacted to rule-breaking topics by enhances likability. However, given that visible
laughing and displaying positive affect, whereas non- groups have been described as the most powerful
delinquent friends reacted positively to prosocial talk. groups in the peer network (Hawley, 2003), we
Social reinforcement, as well as certain pressure tactics anticipated that the socializing influences of central
may be used to protect group identity and test groups would be greater than the socializing influen-
members’ commitment and loyalty to the group. ces of socially preferred groups. We now describe the
behavioral domains most likely to be promulgated by
group centrality and group social preference.
Peer Group Status
Taken together, the most plausible explanation of
Centrality Status
group socialization involves a combination of indi-
vidual motivation for group belonging and group Peer group social status has been most commonly
members’ shaping of desired behavior. The extent to conceptualized as the degree to which a group occu-
which these processes occur may depend on charac- pies a central location within the larger social network
teristics of children’s peer groups, especially the (e.g., Gest et al., 2001). The visibility of central groups
degree to which the group enjoys status within the might be enhanced by having trendy fashions and
larger peer context. Group social status is a character- gadgets, being physically attractive (for girls), and
istic recognizable and salient to most individuals engaging in noticeable and prestigious activities (e.g.,
(Brown & Lohr, 1987). High-status groups generally team sports; Adler, Kless, & Adler, 1992; LaFontana &
offer greater access to social recognition, relation- Cillessen, 2002; Lease, Musgrove, & Axelrod, 2002).
ships, and resources than low-status groups (Eder, Group visibility also may result from positive (i.e.,
1985). It is not surprising that members of high-status prosocial) or negative (i.e., aggressive, antisocial)
1242 Ellis and Zarbatany

behavior, but often a combination of the two is evident cial strategies combined with positive affect, rather
in central groups (Adler & Adler, 1995; Cairns, than harsh or coercive methods. Given the value of
Gariepy, Kindermann, & Leung, 1991; Rodkin et al., relationship harmony and acceptance in well-liked
2000). Because central groups are trend setters in their groups, we expected prosocial behavior to be incul-
school and highly influential both within and outside cated in their groups. We did not expect a moderating
their peer groups (Adler et al., 1992; Eder, 1985), effect of group social preference for aggressive and
central group members likely are highly motivated to deviant behavior because the latter behaviors do not
maintain their group membership and social posi- contribute to the identity of well-accepted groups
tion. They may be especially inclined to conform to and, in fact, are negatively associated with peer
group norms and particularly harsh in dealing with acceptance (Parkhurst & Hopmeyer, 1998).
group members who fail to adhere to group standards
for behavior (Abrams, Rutland, & Cameron, 2003).
Age Considerations
We expected that peer group socialization of devi-
ant behavior (e.g., smoking cigarettes, stealing) and The extant research on group socialization in late
school misconduct (e.g., cheating on a test, talking in childhood and early adolescence has documented
class) would be magnified in central groups because group influence on behavior in both preadolescence
these behaviors likely contribute to central groups’ and early adolescence (Espelage et al., 2003; Kiesner
visibility and ensure they remain recognized by et al., 2002; Ryan, 2001). Although prosocial behavior
others (Rodkin et al., 2000). Because relational aggres- consistently has been recognized as a key element of
sion (e.g., social manipulation, rumor spreading; individual status across childhood and adolescence
Crick & Grotpeter, 1995) and prosocial behavior are despite changes in its relative importance (LaFontana &
important means through which group boundaries Cillessen, 2002; Xie et al., 2006), other behavioral
are protected and loyalty is maintained, we also correlates of group centrality status vary with age.
anticipated greater change in these behaviors with Specifically, relational aggression and deviancy become
increasing group centrality. In contrast, group cen- increasingly important from preadolescence to early
trality effects were not anticipated for overt aggres- adolescence (Cillessen & Mayeux 2004; LaFontana &
sion (e.g., hitting, threatening, name calling) because Cillessen, 2002; Xie et al., 2006). Thus, we anticipated
reliance on overt aggression as a form of social control that high-status central groups might more strongly
diminishes in favor of more subtle manipulative socialize relational aggression and deviant behavior in
strategies by early adolescence, and research linking early adolescence than in preadolescence.
overt aggression with popularity has been inconsis-
tent (Rose, Swenson, & Waller, 2004; Xie, Li, Boucher,
Gender Considerations
Hutchins, & Cairns, 2006).
Information on gender-differentiated behavior at
the peer group level is scarce; therefore, gender
Social Preference Status
considerations were tentatively based on findings
High status also can be defined based on peer at the individual level. Previous research has often
liking or social preference (i.e., sociometric popular- shown that boys tend to engage in more deviant
ity). Peer group status has not been assessed in terms behavior than do girls, and girls tend to engage in
of peer acceptance even though well-liked children more relational aggression and prosocial behavior
are known to cluster together (Kupersmidt, DeRosier, than do boys in late childhood and early adolescence
& Patterson, 1995). Based on knowledge of the be- (e.g., Crick & Grotpeter, 1995; Lease et al., 2002; Santor,
havioral characteristics of well-liked peers, we ex- Messervey, & Kusumakar, 2000). Despite these pos-
pect groups composed of well-liked children to sible mean differences, deviancy and relational
engage in low rates of deviant behavior and aggres- aggression are socialized in both male and female
sion, and high rates of prosocial behavior and aca- peer groups (Espelage et al., 2003; Kiesner et al.,
demic engagement (Lease et al., 2002). These 2002), and both have been associated with high
behaviors likely contribute to their reputation as status for boys and girls (Cillessen & Mayeux, 2004;
being kind and fun to be around. Hawley, 2003; LaFontana & Cillessen, 2002). Prosocial
Well-liked groups may be in a socially advanta- behavior has also been established as an important
geous position because others enjoy their company element of social acceptance and dominance for both
and wish to join their group. To protect their genders (Hawley, 2003; LaFontana & Cillessen, 2002).
well-liked positions, these groups may use subtle There is some evidence based on interviews with
processes to regulate others’ behavior, such as proso- African American adolescents (Xie et al., 2006) and
Peer Group Status as a Moderator 1243

observations of elementary school students (Adler schools (kindergarten to Grade 8) and one middle
et al., 1992) that deviant behavior might contribute school (Grades 5 to 8) in and around a mid-sized
more to the status of individual boys than girls, but Canadian city. Parental consent was received for all
the extent to which these findings apply to peer participants. Census data on socioeconomic and eth-
groups is unknown. In view of existing research, we nic characteristics of the school neighborhoods re-
had no strong basis for expecting that predictions vealed that the sample was middle to lower-middle
regarding the moderating influence of group status class, with an average yearly income of $32,433, and
on behavioral socialization would vary substantially primarily (91%) Caucasion. The largest minority
across genders. groups are Blacks, Arabs, and West Asians.
Because peer nominations were conducted within
grades (rather than classes) in each school to allow for
Summary and Predictions
cross-class relationships within grades, only grades
Motivations to align with a group and establish that had participation rates of at least 60% were
a positive group identity combined with social rein- included in the study to ensure the validity of the
forcement, peer pressure, and other group processes nomination scores. This resulted in exclusion of the
likely contribute to socialization in children’s peer fifth-, sixth-, and seventh-grade groups from one
groups. Because high-status groups have more bene- school, and the seventh-grade group from another.
fits than low-status groups, pressure from group In total there were 16 nominating groups with partic-
members and motivation to conform likely are stron- ipation rates ranging from 60% to 100% (Mdn 5 77%,
gest in these groups. M 5 74%, SD 5 10.99): 4 fifth-grade, 4 sixth-grade, 3
Although members of high-status central groups seventh-grade, and 5 eighth-grade groups. The begin-
and high-status socially preferred groups both may be ning sample consisted of 711 children (M age 5 12.05
socially advantaged and strive to maintain their years, SD 5 1.17).
positions (Hawley, 1999), they likely use different Although consent was obtained for 711 children, the
strategies to communicate their preferences, with number of participants varied across assessment peri-
central group members relying on a combination of ods primarily because of absenteeism. A total of 665
positive behavior and intimidation and control, and children completed questionnaires at Time 1. From
socially preferred group members primarily using Time 1 to Time 2 (a 3-month interval), 43 participants
relationship-building strategies. were lost for various reasons (26 repeatedly absent, 17
Over a 3-month period we examined the moderat- moved away or dropped out). Independent t tests
ing effects of group centrality status and group social conducted on all relevant variables showed that drop-
preference status on group socialization in three outs were significantly less prosocial, t(658) 5 3.65,
domains: deviance (school misconduct and more p , .01, and engaged in more school misconduct,
serious deviant acts), aggression (overt and rela- t(658) 5 2.17, p , .05, than those who remained.
tional), and prosocial behavior. We generally expected
that group socialization effects would be strongest in
Measures
high-status groups and that group centrality would
produce stronger effects than group social preference. Peer groups. The social cognitive map (SCM) tech-
We also anticipated that the two types of group status nique developed by Cairns et al. (1991) was used to
would produce some nonoverlapping behavioral ef- identify children’s natural social groups. This pro-
fects, with socialization of deviancy, school miscon- cedure is based on the assumption that children can
duct, relational aggression, and prosocial behavior accurately recall the social relationships of children
likely to occur in central groups, and socialization of around them, even those in which they are not directly
prosocial behavior likely to occur in well-liked groups. involved. Children were asked: ‘‘Do you have a group
The moderating effects of group status on deviant (of three or more kids) in your class or grade that you
behavior and relational aggression were expected to hang around with a lot?’’ Regardless of their group
be stronger for older than for younger groups. membership, all children also were asked: ‘‘Are there
other people in your class or grade who hang around
together a lot?’’ They could identify up to three other
Method groups in their class or grade. To ensure the validity of
the procedure, group members could include children
Participants
who were not participants in the study.
Participants were recruited from all fifth-, sixth-, Based on reports from all participants in a nomi-
seventh-, and eighth-grade classes in four elementary nating group, a co-occurrence matrix was computed
1244 Ellis and Zarbatany

containing the frequency with which each child was in other groups resulted in dissolution of the original
nominated as being in the same group as every other group. When only one or two participating children
child. These affiliation profiles for every child were were members of groups with nonparticipating stu-
then correlated with every other child’s profile, and dents, they also were excluded, as we were unable to
groups were clustered according to similarity using ascertain group characteristics for these cases. Thus,
a standard cutoff point for similarity (r . .40). To the final sample included 526 children (244 boys, 282
create nonoverlapping groups, children who were girls; M age 5 12.05, SD 5 1.14) and was distributed
initially affiliated with more than one group were across grades as follows: Grade 5 (n 5 99; 54 boys, 45
assigned membership to the group for which they had girls; M age 5 10.37, SD 5 .30), Grade 6 (n 5 136;
a .50 correlation with at least 50% of the group 67 boys, 69 girls; M age 5 11.42, SD 5 .33), Grade 7 (n 5
members (Cairns et al., 1991). In cases where this 142; 67 boys, 75 girls; M age 5 12.40, SD 5 .42), Grade 8
was true of both group assignments, children were (n 5 149; 56 boys, 93 girls; M age 5 13.37, SD 5 .36).
placed in the group with the strongest correlations We compared the 96 (56 boys, 40 girls) excluded
with other group members. children the 526 group members on all outcome
In total there were 14 nominating groups for the variables (delinquent behavior, aggression, and pro-
group membership assessment: 4 fifth-grade, 4 sixth- social behavior) using a 2 (gender)  4 (grade)  2
grade, 2 seventh-/eighth-grade split, 1 seventh- (group member, no or yes) multivariate analysis of
grade, and 3 eighth-grade groups. All but 1 (see the variance. No significant multivariate or univariate
following) group had participation rates on the group effects emerged involving group membership.
membership question higher than 60%. A strength of Centrality status. In addition to the specification of
the SCM procedure is its strong psychometric prop- group members, the SCM program determines the
erties in groups with low to moderate participant centrality status of each group. The frequency of
rates. Because aggregate responses are used and individual nominations reflects the group’s position
groups are generally publicly agreed on, group pro- or status (Gest et al., 2001). A centrality index is
files based on participation rates of 50% are almost created for each group based on the average of the
indistinguishable from those with participation rates two people with the highest number of nominations
of 100% (Cairns, Leung, Buchanan, & Cairns, 1995). within each group (see Cairns et al., 1995, for further
Furthermore, self-nominated peer groups overlap details). Nuclear groups (high centrality status) are
considerably with SCM aggregate reports, suggesting those that have the greatest number of individual
that even children with single (or few) nominations nominations. Peripheral groups (low centrality sta-
are reliably placed into groups (Cairns et al., 1995). tus) are those with the fewest individual nominations.
Observational data and short-term group stability Secondary groups (average centrality status) are those
assessments also confirm the reliability of the SCM that fall between these two extremes. Thus, group
measure for identifying groups with varying num- centrality is a categorical variable with three levels.
bers of nominations (Cairns et al., 1995; Gest, Farmer, Cairns et al. (1995) found no difference in group
Cairns, & Xie, 2003). stability (retest reliability) of central, secondary, and
Given that group stability could confound the peripheral groups over a 3-week period, suggesting
interpretation of group status effects (i.e., if central that groups with many nominations (central) are no
groups are more stable, socialization effects may be due more accurately identified than groups with few
to longer exposure rather than to status), we included nominations (peripheral).
stability as a covariate in all major analyses. Groups Problem behavior. The 18-item Problem Behavior
were identified at Time 2 using the same procedure. Scale consisted of 13 items measuring school mis-
Group stability scores were created by dividing the conduct (e.g., cheating on an exam, not doing home-
number of members that remained the same across work, being suspended; Jenkins, 1997) and 6 items
periods for each group by the total size of the group at assessing deviant behaviors (i.e., destroying school
Time 1 (M stability 5 56.11%, SD 5 33.44%). property, drinking alcohol, smoking cigarettes, using
Application of the SCM procedure resulted in the illegal drugs, stealing, vandalism; Santor et al., 2000).
loss of 96 children from the sample. These children Participants reported how often they engaged in each
were excluded because they were in a class with very behavior in the past 3 months using a 3-point scale (1
low participation rates on the group membership 5 never, 2 5 sometimes, 3 5 often). Separate scores were
question (n 5 20), they were not nominated as belong- created for school misconduct (13 items, alpha 5 .82)
ing to a group (n 5 20), or they could not be and more serious deviant acts (6 items, alpha 5 .82).
successfully placed in a group (n 5 56). The latter Although self-reported measures of deviant behav-
occurred when placement of multiple-group members ior may sometimes lead to an underestimation of
Peer Group Status as a Moderator 1245

deviant behavior, researchers generally have shown psychological adjustment. The current report repre-
that this method can yield valid results, particularly sents the findings related to group influence
with respect to substance abuse (see Brener, Billy, & on behavioral development. Questionnaires were
Grady, 2003, for a review). Bendixen and Olweus administered in children’s home classrooms 3 times
(1999) used a similar scale of self-reported early during the school year, beginning in October and
adolescent school-related problem behavior and subsequently at 3-month intervals. Only data from
non-school-related antisocial behavior, and demon- the first two waves of data collection are reported
strated strong psychometric properties including here. The number of days between Time 1 and Time 2
stability and convergent validity. Furthermore, our ranged from 88 to 104 (M 5 95.13, SD 5 4.98). The
problem behavior scores were stable over the 3-month questionnaire order remained the same across testing
period (school misconduct: r 5 .66; deviant behavior: waves to ensure any changes over time were not the
r 5 .53). result of order effects. The nomination questionnaires
Overt aggression, relational aggression, prosocial behavior, were placed at the beginning of the questionnaire
and leadership. In a modified version of the Revised booklet, and the remaining questionnaires were
Class Play (Masten, Morrison, & Pellegrini, 1985), ordered to vary positive and negative adjustment
participants were asked to select up to three people measures, as well as response format to minimize
from a list of participating grade-mates who fit eight carry-over effects. Students were led through the
behavioral profiles. Two items assessed overt aggres- questionnaires by a pair of undergraduate or gradu-
sion (starts fights, picks on or teases others), two ate researchers (one male and one female researcher
assessed relational aggression (when mad, ignores or were present for most of the sessions). Instructions
stops talking to person; when mad, gets even by and examples for each measure were read aloud to the
keeping person from being in their group of friends; class and individual questions were read aloud when
Crick & Grotpeter, 1995), two assessed prosocial necessary. Children were encouraged to ask questions
behavior (helps others, kind to others), and two at any time if needed. Each session took approxi-
assessed leadership (good leader, everyone listens mately 1 hr 15 min.
to). Leadership scores were used to validate our
conceptualization of group centrality as reflecting
high power. The number of nominations each child Results
received was averaged across items within each
Descriptive Characteristics of Peer Groups
behavioral cluster and then standardized within
nominating groups to control for group size. At Time A total of 113 nonoverlapping groups (50 male, 43
1 the alpha coefficients for overt aggression, relational female, 20 mixed gender) made up the final sample,
aggression, prosocial behavior, and leadership were including 25 fifth-grade groups, 34 sixth-grade
.88, .74, .83, and .78, respectively. As part of a larger groups, 30 seventh-grade groups, and 24 eighth-
study, children also nominated peers on victimization grade groups. Group size ranged from 3 to 21 mem-
and shyness. bers (M 5 7.56, SD 5 4.55). To check for gender and
Social preference status. Social preference status grade differences in group size, a 3 (group gender)  4
scores were derived from children’s nominations of (grade) analysis of variance was computed. Signifi-
three peers they really enjoyed being with at school cant main effects for gender, F(2, 100) 5 11.32, p , .001,
and three peers they did not care to be with at school. and grade, F(3, 100) 5 7.18, p , .001, were followed up
Like-least nominations were subtracted from like- with Tukey’s post hoc comparisons. The size of
most nominations, and difference scores were stan- mixed-gender groups (M 5 8.33, SE 5 .63) was
dardized within groups (Coie, Dodge, & Coppotelli, significantly larger than female groups (M 5 5.79,
1982). To determine group social preference, chil- SE 5 .54) and male groups (M 5 4.59, SE 5 .47; all ps ,
dren’s social preference scores were averaged within .01). Male and female groups did not differ in size.
their peer group as identified in the SCM procedure Groups in eighth grade (M 5 8.67, SE 5 .66) were
As expected, group members had similar social significantly larger than groups in sixth grade (M 5
preference scores, as indicated by the significant 4.65, SE 5 .64) and fifth grade (M 5 5.44, SE 5 .61; all
intraclass correlation coefficient of .13 (p , .001). ps , .01), but not seventh grade (M 5 6.19, SE 5 .63).
The interaction between group gender and grade was
not significant. We also checked for grade differences
Procedure
in the gender composition of the groups using chi-
This study was part of a larger study on friend and square analysis but found no grade differences in the
group influence on behavioral development and tendency for groups to be same or mixed gender.
1246 Ellis and Zarbatany

Correlations Among Group Variables

0.61***

7
The means, standard deviations, and zero-order
correlations among variables at the group level (group
means) at Time 1 are shown in Table 1. Group central-
ity and group social preference were significantly

0.39***
related but only moderately so (r 5 .36). Consistent

0.22*

6
with past research, social preference was positively
correlated with prosocial behavior and negatively
correlated with problem behavior, whereas group
centrality status was positively correlated with both
types of aggression, school misconduct, and prosocial

0.35***
0.35***
behavior (Lease et al., 2002). As we expected, leader-

0.13

5
ship was significantly related to group centrality (r 5
.57), validating our conceptualization of centrality as
reflecting social power and influence. It is interesting
that leadership was more highly related to group
centrality than to social preference: Fisher r to z trans-

0.28**

0.20*
0.03

0.14

formation, t(97) 5 2.24, p , .05.

4
Hypothesis Testing: Analytic Overview
The study of groups involves nested contexts that
are best examined using hierarchical linear modeling

0.66***
0.45***

0.68***
0.14*
(HLM; Bryk & Raudenbush, 1992). We used a two-

0.13

level HLM model to determine the effect of peer 3
group context at Time 1 on children’s behavior at
Means, Standard Deviations, and Zero-Order Correlations for Group Variables (Group Means) at Time 1

Time 2 and to examine the potential moderating effect


of group status on Time 2 outcomes. Three steps are

0.49***
0.37***
involved in a two-level HLM model. These steps were
0.30**
0.24*
0.16
0.01

repeated for each of the Time 2 outcome variables


2

(deviant behavior, school misconduct, overt aggres-


sion, relational aggression, and prosocial behavior).
First, we estimated a fully unconditional model.
Within- and between-group components were used to
0.36***

0.39***

0.57***

estimate the total variance that lay systematically


0.33**
0.09*

0.19*
0.02

1

between groups on the outcome variable. The pro-


portion of the total variance between groups relative
to the variance within groups is called the intraclass
correlation (ICC). A significant ICC provides the
foundation for rejecting the null hypothesis that mean
2.22 (0.51)
0.01 (0.55)
1.34 (0.20)
1.09 (0.18)
0.06 (0.53)
0.06 (0.55)
0.07 (0.52)
0.07 (0.51)

scores are identical across groups on the variable of


M (SD)

interest and indicates that multilevel analysis is


warranted. The strength of the ICC also indicates
the level of within-group similarity.
Second, we computed a within-group model in
*p , .05. **p , .01. ***p , .001.

which regression equations were used to predict


individual Time 2 behavior as a function of Time 1
Relational aggression

individual behavior within each peer group. This


School misconduct

Prosocial behavior
Deviant behavior
Overt aggression
Social preference
Group centrality

model also assessed between-group differences when


equated for individual levels of behavior at Time 1. In
Leadership

this equation, the intercept was the average individ-


ual score for behavior at Time 2.
Table 1

Third, we used the randomly varying intercept


from the Level 1 analysis as the dependent variable in
1
2
3
4
5
6
7
8
Peer Group Status as a Moderator 1247

the between-group (Level 2) model to determine gender) produced no significant moderating effects
significant predictors of Time 2 average individual involving group grade and gender. Significant
behavior, beyond the contribution of Time 1 behavior. Status  Behavior interactions were followed up by
Two Level 2 models were computed. First, we tested three-way interactions with gender and grade to
the main effect of peer group behavior at Time 1 on determine whether the effects applied equally to all
children’s Time 2 behavior. Second, we examined the groups in the sample. No significant three-way inter-
moderating effect of group status using interaction actions were found. Thus, gender and grade did not
terms for each measure of status and behavior. The moderate any of the major findings.
significance of the interaction terms determined
whether peer group status moderated the relation-
Predicting Deviant Behavior
ship between Time 1 group behavior and Time 2
average individual behavior. Group size and group We first examined peer group effects on children’s
stability were entered as control variables in all Level deviant behavior. In the first step, a fully uncondi-
2 equations; however, these variables did not emerge tional model was estimated. The ICC showed that
as statistically significant predictors in any of the a significant proportion of the variance (16.24%, p ,
Level 2 analyses. Thus, coefficients related to these .001) in deviant behavior was between groups and
variables are not reported. The final model is repre- multilevel analysis was warranted to determine what
sented by the following equation: variables accounted for the between-group differ-
ences, s 5 .0097, v2(90) 5 186.81, p , .001. At Level 1,
b0j ðLevel 1 intercept=Average behavior at Time 2Þ children’s deviant behavior at Time 1 was a significant
predictor of their deviant behavior at Time 2 (t 5 8.77,
5 c00 þ c01 ðgroup stabilityÞ þ c02 ðgroup sizeÞ
p , .001), accounting for 35.51% of the individual level
þ c03 ðgroup behaviorÞ variance. However, significant variation remained in
þ c04 ðgroup social preferenceÞ deviant behavior scores across peer groups, as indi-
þ c05 ðgroup centralityÞ cated by the estimated variance among group means,
s 5 .0018, v2(88) 5 68.17, p , .01. Next, a Level 2 model
þ c06 ðsocial preference  group behaviorÞ
was tested with mean group deviant behavior as
þ c07 ðcentrality  group behaviorÞ a predictor. As a main effect, peer group deviant
þu0j ðgroup errorÞ: behavior at Time 1 was a significant, positive pre-
dictor of children’s deviant behavior at Time 2 (see
The interpretation of interactions in HLM is basi- Table 2, Panel A). Finally, a Level 2 model testing the
cally the same as in regression analysis. As such, moderating effect of peer group status revealed
significant interactions were analyzed according to significant interactions between peer group deviant
the guidelines outlined by Aiken and West (1991). To behavior and group social preference status, and
decipher the overall pattern of each interaction, group centrality status (see Table 3, Panel A). Specif-
variables in the interaction were split into 1 SD above ically, when the group had high levels of deviant
and 1 SD below their means. Simple-slope analyses behavior at Time 1, children in groups with low social
were conducted to determine whether the slopes preference showed an increase in deviant behavior at
representing each relationship were significantly dif- Time 2, b 5 .27, t(86) 5 2.01, p , .05 (see Figure 1), but
ferent from zero at high and low levels. children in groups with high social preference did
Predictors in the Level 1 analysis were grand-mean not, b 5 .17, t(86) 5 1.27, ns. The interaction be-
centered. When using grand-mean centering, the Level tween group centrality and deviant behavior showed
2 model represents the group-level relationship the opposite pattern. When the group was deviant at
between the predictor and the outcome adjusted for Time 1, children in highly central groups experienced
the influence of the Level 1 predictor (Bryk & an increase in deviant behavior at Time 2, b 5 .27,
Raudenbush, 1992). Thus, final effects are predictive t(86) 5 2.56, p , .01 (see Figure 2), whereas children
beyond Time 1 individual behavior. All variables and in less central groups did not, b 5 .22, t(86) 5 1.54,
group-level interactions at Level 2 were group-mean ns. The Level 2 model accounted for 41.42% of
centered. variance between peer groups in deviant behavior.
Children’s age and gender were entered as control
variables at Level 1 but did not alter the pattern of
Predicting School Misconduct
findings for Level 2 models and were excluded from
the final models. At Level 2, cross-level and group- We next examined peer group effects on children’s
level interactions with status (i.e., male, female, mixed school misconduct. The fully unconditional model
1248 Ellis and Zarbatany

Table 2 Table 3
Hierarchical Linear Models Predicting Time 2 Individual Behaviors From Hierarchical Linear Models With Group Status as a Moderator of Time 2
Time 1 Group Behavior Behaviors

Variable Effect SE t test Variable Effect SE t test

A. Between-group model for A. Between-group model for deviant behavior


deviant behavior Average intercept (c00) 1.10 .02 62.37***
Group deviant behavior (c03) 0.02 .09 0.18
Average intercept (c00) 1.11 .02 64.57***
Group social preference (c04) 0.07 .02 3.43**
Group deviant behavior (c01) 0.16 .08 1.99*
Group centrality (c05) 0.03 .01 1.99*
B. Between-group model for Social Preference  Group Deviant 0.55 .16 3.37**
school misconduct Behavior (c06)
Average intercept (c00) 1.37 .01 105.73*** Centrality  Group Deviant 0.23 .12 1.92*
Behavior (c07)
Group school misconduct (c01) 0.31 .07 4.12***
Individual deviant behavior at 0.71 .18 3.67***
C. Between-group model for Time 1 (c10)
relational aggression B. Between-group model for school misconduct
Average intercept (c00) 0.01 .04 0.04 Average intercept (c00) 1.37 .01 95.59***
Group school misconduct (c03) 0.24 .08 2.91***
Group relational aggression (c01) 0.23 .08 2.77**
Group social preference (c04) 0.02 .03 0.67
D. Between-group model for Group centrality (c05) 0.04 .02 1.67
prosocial behavior Social Preference  Group School 0.17 .15 1.10
Average intercept (c00) 0.02 .03 0.43 Misconduct (c06)
Group prosocial behavior (c01) 0.03 .07 0.16 Centrality  Group School 0.22 .12 1.85y
Misconduct (c07)
E. Between-group model for Individual school misconduct at 0.60 .05 11.67***
overt aggression Time 1 (c10)
Average intercept (c00) 0.01 .03 0.19 C. Between-group model for relational aggression
Group overt aggression (c01) 0.15 .08 1.97* Average intercept (c00) 0.04 .05 0.98
Group relational aggression (c03) 0.11 .10 1.16
*p , .05. **p , .01. ***p , .001. Group social preference (c04) 0.08 .10 0.85
Group centrality (c05) 0.11 .08 1.31
Social Preference  Group Relational 0.21 .16 1.28
Aggression (c06)
showed that a significant proportion of the variance Centrality  Group Relational 0.34 .15 2.26*
(22.01%, p , .001) in school misconduct was between Aggression (c07)
groups, s 5 .1445, v2(96) 5 336.65, p , .001. The Level Individual relational aggression at 0.21 .04 9.59***
Time 1 (c10)
1 model showed that children’s school misconduct at D. Between-group model for prosocial behavior
Time 1 was a significant predictor of their behavior at Average intercept (c00) 0.03 .04 0.76
Time 2 (t 5 13.41, p , .001), accounting for 32.29% of Group prosocial behavior (c03) 0.04 .09 0.39
the individual level variance. After controlling for Group social preference (c04) 0.03 .10 0.33
Time 1 behavior, there remained significant variation Group centrality (c05) 0.10 .07 1.53
Social Preference  Group Prosocial 0.03 .18 0.16
in average school misconduct across peer groups, s 5 Behavior (c06)
.0039, v2(88) 5 141.44, p , .01. The first Level 2 model Centrality  Group Prosocial 0.26 .12 2.09*
showed that peer group school misconduct at Time 1 Behavior (c07)
was a significant positive, predictor of children’s Individual prosocial behavior 0.63 .05 10.62***
school misconduct at Time 2 (see Table 2, Panel B). at Time 1 (c10)
E. Between-group model for overt aggression
The next Level 2 model revealed that the interaction Average intercept (c00)
between peer group school misconduct and group Group overt aggression (c03) 0.14 .09 1.18
centrality status approached significance (p 5 .06; Group social preference (c04) 0.03 .07 .38
see Table 3, Panel B). The interaction between group Group centrality (c05) 0.08 .07 1.18
centrality and school misconduct showed that Social Preference  Group Overt 0.22 .14 1.55
Aggression (c06)
children in highly central groups experienced an Centrality  Group Overt 0.04 .14 .18
increase in school misconduct at Time 2, b 5 .39, Aggression (c07)
t(86) 5 3.34, p , .01 (see Figure 3), but children in Individual overt aggression at 0.69 .08 9.08***
less central groups did not, b 5 .09, t(86) 5 .82, ns. Time 1 (c10)
The between-group model accounted for 83.10% of
*p , .05. **p , .01. ***p , .001. yp 5 .06.
variance between peer groups in average school
misconduct.
Peer Group Status as a Moderator 1249

1.20
2.40
Problem Behavior at Time 2

Average Individual School


Average Individual Deviant

Misconduct at Time 2
1.15 2.05

1.10 1.70

1.35
1.05

1.00
1.00 1 SD below 1 SD above
1 SD below 1 SD above
Group School Misconduct at Time 1
Group Deviant Problem Behavior at Time 1 High Group Low Group
High Group Low Group Centrality Centrality
Social Preference Social Preference

Figure 1. Time 1 Group Deviant Behavior  Group Social Prefer- Figure 3. Time 1 Group School Misconduct  Group Centrality
ence predicting Time 2 deviant behavior. predicting Time 2 school misconduct.

Predicting Relational Aggression aggression across peer groups, s 5 .0291, v2(84) 5


105.72, p 5 .05. The first Level 2 model showed that
Next, we examined peer group effects on children’s group relational aggression was a significant positive
relational aggression. In the first step, the ICC showed predictor of children’s relational aggression at Time 2
that a significant proportion of the variance (11.34%, (see Table 2, Panel C). The next Level 2 model pro-
p , .001) in relational aggression was between groups, duced a significant interaction between peer group
s 5 .1100, v2(96) 5 165.00, p , .01. The Level 1 model relational aggression and group centrality (see
showed that children’s relational aggression at Time 1 Table 3, Panel C). Children who were part of highly
was a significant predictor of their relational aggres- relationally aggressive and central peer groups expe-
sion at Time 2 (t 5 10.69, p , 001), accounting for rienced an increase in Time 2 relational aggression,
27.80% of the individual level variance in the out- b 5 .35, t(83) 5 2.48, p , .01 (see Figure 4), whereas
come. After controlling for Time 1 relational aggres- children in relationally aggressive, less central groups
sion, there remained moderate variation in relational did not, b 5 .12, t(83) 5 .80, ns. The between-group
model accounted for 80.95% of the variance between
1.20 peer groups in average relational aggression.

Predicting Prosocial Behavior


Problem Behavior at Time 2

1.15
Average Individual Deviant

We next tested peer group effects on children’s


prosocial behavior. A significant proportion of the
1.10 variance (10.75%, p , .001) was between groups,
s 5 .1057, v2(90) 5 147.12, p , .001. The Level 1 model
revealed that children’s prosocial behavior at Time 1
was a significant predictor of their behavior at Time 2
1.05
(t 5 11.06, p , .001), accounting for 58.80% of the
individual level variance. After controlling for Time 1
scores, there remained significant variation in proso-
1.00
1 SD below 1 SD above cial scores across peer groups, s 5 .0369, v2(89) 5
Group Deviant Problem Behavior at Time 1
141.39, p , .01. The first Level 2 model showed that
peer group prosocial behavior was not a significant
High Group Low Group predictor of the same behavior at Time 2 (see Table 2,
Centrality Centrality
Panel D). However, the second Level 2 model testing
Figure 2. Time 1 Group Deviant Behavior  Group Centrality the interaction terms involving group status produced
predicting Time 2 deviant behavior. a significant interaction between peer group prosocial
1250 Ellis and Zarbatany

0.25
0.25
Average Individual Relational

Average Individual Prosocial


Aggression at Time 2

0.10
0.10

Behavior at Time 2
-0.05 -0.05

-0.20 -0.20

-0.35
-0.35
1 SD below 1 SD above 1 SD below 1 SD above
Group Relational Aggression at Time 1 Group Prosocial Behavior at Time 1

High Group Low Group High Group Low Group


Centrality Centrality Centrality Centrality

Figure 4. Time 1 Group Relational Aggression  Group Centrality Figure 5. Time 1 Group Prosocial Behavior  Group Centrality
predicting Time 2 relational aggression. predicting Time 2 prosocial behavior.

behavior and group centrality status (see Table 3, Panel Discussion


D). In this interaction neither slope was significant,
Consistent with previous work, we have demon-
but the significant interaction itself reflects different
strated that peer groups socialize deviancy, school
effects of group context depending on group centrality.
misconduct, and aggression in late childhood and
Average individual prosocial behavior increased with
early adolescence. More important, we have shown
increasing group centrality and decreased with lower
for the first time that peer groups vary in influence,
group centrality (see Figure 5). The between-group
with high-status groups being more powerful agents
model accounted for 21.58% of the variance between
of socialization than low-status groups. Comparison
peer groups in average prosocial behavior.
of two aspects of group status, centrality (visibility)
and social preference (acceptance), revealed the for-
Predicting Overt Aggression mer as primarily associated with behavior change,
especially for behavior that helped maintain group
Finally, we checked for group effects on children’s
status. Specifically, group centrality intensified group
overt aggression. A significant proportion of the
socialization of deviant behavior, school misconduct,
variance (8.90%, p , .001) in overt aggression was
relational aggression, and prosocial behavior. In
between groups, s 5 .0700, v2(96) 5 114.94, p , .01.
contrast, (low) group acceptance affected only the
The Level 1 model revealed that children’s overt
socialization of deviant behavior. These effects were
aggression at Time 1 was a positive predictor of their
obtained regardless of children’s gender and age.
overt aggression at Time 2 (t 5 9.98, p , .001) and
accounted for 51.63% of the individual level variance.
After controlling for Time 1 scores, there remained
Peer Group Socialization of Deviant Behavior and
moderate variation in mean overt aggression scores
School Misconduct
across peer groups, s 5 .0251, v2(81) 5 101.79, p 5 .06.
The Level 2 model with peer group overt aggression Individuals who have a propensity toward deviant
as a main effect showed that there was a significant behavior may be initially drawn together, but as we
effect of Time 1 group overt aggression on children’s and others have shown (e.g., Kiesner et al., 2002), the
Time 2 aggression (see Table 2, Panel E). The second peer group continues to support and shape their
Level 2 model revealed no significant interactions involvement in antisocial activities. Our study dem-
between group status and group aggression (see onstrates that school-related misconduct (e.g., cheat-
Table 3, Panel E). The final model accounted for ing on a test, skipping class) also is strongly
72.09% of the variance between peer groups in susceptible to group influence, a worrisome finding
average overt aggression. given that this behavior can lead to more serious acts
Peer Group Status as a Moderator 1251

of delinquency (Bryant, Schulenberg, Bachman, 2004). In the future it would be useful to follow central
O’Malley, & Johnston, 2000). Most notable, group and rejected groups over longer periods to confirm
influence on deviant behavior and school misconduct motivation for, as well as manifestation, escalation, and
was magnified in central peer groups. Central groups consequences of, deviant behavior.
are well known by others, and their disruptive
behavior is highly visible to peers (Cillessen &
Peer Group Socialization of Aggression
Mayeux, 2004; Rodkin et al., 2000). Acting out in the
classroom is one way for central groups to attract Consistent with others’ reports (Cairns et al., 1988),
attention. Similarly, experimenting with smoking or we found that children’s overt aggression escalated
drinking, or engaging in vandalism may be consid- when they were members of aggressive peer groups.
ered ‘‘cool’’ or a sign of maturity when central group Given that peer groups’ shared overt aggression is
members exhibit these acts, and other network mem- likely to be goal directed rather than reactive (Poulin
bers may flock to join in. Given that these findings et al., 1997), children in overtly aggressive peer
held regardless of the age or gender configuration of groups may come to value aggression as a means to
the groups, it seems reasonable to conclude that goal attainment and rely on their group alliance
membership in central peer groups may form a sig- for protection and support. We also provide some of
nificant pathway to deviant behavior beginning in the first evidence that peer groups contribute to the
late childhood and possibly earlier. socialization of relational aggression (see also
Children in deviant groups that were low in social Espelage et al., 2003). Manipulative acts involving
preference (i.e., rejected) also showed increases in gossip and social exclusion need involvement from
deviancy. Rejected children generally establish net- most group members. In socially aggressive peer
works with other rejected children (Ladd, 1983), groups, members may participate in relational
alliances that develop from shared negative experi- aggression to guard group boundaries and shield
ences with parents and peers and lead to a trajectory themselves from similar attacks (Adler & Adler,
of antisocial behavior (Dishion et al., 1991). Because 1995).
rejected children are generally isolated from other As we anticipated, group centrality moderated
peers, when they do find acceptance in deviant socialization of relational but not overt aggression.
groups they may strive for behavioral conformity to This may reflect early adolescents’ decreasing reli-
group norms to maintain group membership. ance on overt aggression as a form of social control in
Our finding that group centrality and group social favor of more covert manipulative tactics (Rose et al.,
preference had opposite effects on group socialization 2004). Compared with overt aggression, relational
of deviance suggests that different mechanisms may aggression is a highly skilled means of controlling
underlie group status effects that have long-term social boundaries because children can use this
consequences for adjustment (Rodkin et al., 2000). behavior to damage relationships without directly
Central groups may use deviant behavior as a means implicating themselves. In this manner, children
of enhancing and protecting their social positions in are able to maintain positive affect in their relations
the peer network, whereas rejected groups may with others while deftly preserving their elevated
engage in deviant behavior as a reaction to problems status. Evidence that individual popularity predicts
experienced at home or with peers (Dishion et al., increasing use of relational aggression over time
1991). If central groups use deviancy primarily to (Cillessen & Mayeux, 2004; Rose et al., 2004) supports
mark their maturity and status (Moffitt, 1993), deviant our contention that this behavior is a tool used to main-
behavior of central group members should diminish tain status. Overt aggression, in contrast, is unlikely to
in most cases as youth exit adolescence and adopt be a status-maintaining device in popular groups.
their adult roles. In contrast, deviant behavior of
rejected-deviant groups may continue or escalate
Peer Group Socialization of Prosocial Behavior
if underlying problems are not addressed (Hymel,
Rubin, Rowden, & LeMare, 1990; Kupersmidt, The present study was the first to examine peer
Burchinal, & Patterson, 1995). Unlike central groups, group socialization of prosocial behavior. We found
rejected groups are cut off from other peers and no overall group socialization effect, but central peer
deprived of opportunities for learning important social groups did indeed promote prosocial behavior as we
skills. Low social skills combined with delinquent had anticipated. Although this finding gives some
behavior may contribute to long-term problems such reassurance that central groups are not exclusively
as psychological distress, poor academic functioning, negative in their influence on social behavior, the
and serious criminal behavior (Nelson & Dishion, combination of prosocial and antisocial behavior
1252 Ellis and Zarbatany

inculcated within central groups is reminiscent low participation rates in some groups. In the future, it
of findings on prosocial dominance described by might be useful to conduct a two-stage assessment of
Hawley (1999, 2003). Hawley argued that successful group liking by first identifying the groups and then
access to resources requires more than brute force but obtaining liking ratings for groups rather than indi-
rather a balance of both coercive and prosocial acts. viduals. This type of research would enable a more
Central groups may attract new members by lavish- complete assessment of the socialization dynamics in
ing them with attention and kind acts and then high- and low-status peer groups and determine the
suddenly shift their attention elsewhere (Adler & extent to which group liking fuels group socialization
Adler, 1995). This bistrategic approach is likely essen- effects.
tial for dominance regardless of group gender or age. Our assessment of the behavioral characteristics of
It remains to be determined whether central group groups also is susceptible to criticism because of the
members’ goals to maintain their group desirability less-than-perfect participation rates in most groups.
and exclusivity can be achieved with a strictly uni- Group behavior scores could be calculated only based
strategic (i.e., prosocial) approach. on group members who were in the study. The intra-
class correlations indicated that children perceived
group members as behaviorally similar, however, and
Age Considerations Revisited
we did obtain the expected group status moderation
We did not find the anticipated age differences in effects on behavior. Thus, we are confident that our
group socialization of deviant behavior or relational assessment of group behavioral characteristics was
aggression; both increased in high status groups of all not compromised by low participation in some
ages (and genders). Although deviance and aggres- groups. Still, the two-stage assessment approach that
sion gain greater acceptance from childhood to ado- we described previously for group social preference
lescence (Rose et al., 2004), the age range of our might circumvent group behavioral assessment prob-
sample may have been too limited to detect moder- lems created by low participation rates. Obviously,
ating effects of grade on the group status effect. such an approach would require large samples and
Inclusion of both younger and older samples would many participating groups.
be necessary to determine at what point in develop- The requirements of multilevel analysis demanded
ment status-motivated socialization of deviance and that each child be placed into only one peer group.
relational aggression is most intense. At the very least, Many children were initially identified as multiple
our findings suggest that the central group member- group members but were assigned to a single group
ship motivates acquisition of more ‘‘mature’’ social based on the relative strength of their association, or
behavior as early as age 10. could not be classified into a single group and were
lost from our sample. When children affiliate with
more than one peer group they may experience
Limitations and Future Directions
different and possibly conflicting sources of influ-
Our findings suggest that group centrality (power) ence. Similarly, there may be differences in socializa-
rather than group liking induces the greatest change tion effects of self-identified and peer-identified peer
in group members’ behavior (Lightfoot, 1992). How- groups. Future research comparing socialization
ever, this conclusion must be weighed against possi- across children’s multiple groups would help
ble methodological interpretations of the data. In our elaborate on the effect of group centrality status as
analysis we ruled out group size and group stability well as effects of non-status-related group character-
as potential causes of the differentiated effects pro- istics (e.g., shared time, group organization) on social-
duced by group centrality and group social prefer- ization of status- and non-status-related behavior.
ence. In contrast with previous research (Kupersmidt,
DeRosier, & Patterson, 1995), though, considerable var-
General Conclusions
iability was evident in peer liking within groups
compared with the other variables examined. Thus, The present study adds to our knowledge of peer
the socialization context was not particularly homoge- group socialization and shows that peer groups
neous for liking or conducive to obtaining the expected are powerful contexts that shape both positive and
socialization effect for prosocial behavior. Because our negative aspects of children’s behavior. We have
measure of group liking was based on a compilation of shown that peer group socialization is a more com-
individual members’ liking rather than liking of the plex phenomenon than previously thought, with
group as an aggregate, it is possible that group mere exposure to group behavior insufficient to
heterogeneity on social preference was attributable to instigate behavioral change. Rather, socialization is
Peer Group Status as a Moderator 1253

most likely to occur in highly central peer groups and Cairns, R. B., Cairns, B. D., Neckerman, H. J., Gest, S. D., &
in behavioral domains that contribute to central Gariepy, J. (1988). Social networks and aggressive behav-
group status. The advantageous position enjoyed by ior: Peer support or peer rejection? Developmental
central groups may encourage dominance-related Psychology, 2, 815 – 823.
skills that help maintain status and access to valued Cairns, R. B., Gariepy, J. L., Kindermann, T., & Leung, M.
(1991). Identifying social clusters in natural settings.
resources such as space, equipment, and attention
Unpublished manuscript, University of North Carolina
from peers. Thus, children are active players and at Chapel Hill.
likely use peer group membership to maximize their Cairns, R., Leung, M. C., Buchanan, L., & Cairns, B. D.
social benefits. (1995). Friendships and social networks in childhood
and adolescence: Fluidity, reliability and interrelations.
References Child Development, 66, 1330 – 1345.
Cillessen, A. H. N., & Mayeux, L. (2004). From censure to
Abrams, D., Rutland, A., & Cameron, L. (2003). The reinforcement: Developmental changes in the associa-
development of subjective group dynamics: Children’s tion between aggression and social status. Child Devel-
judgments of normative and deviant in-group and out- opment, 75, 147 – 163.
group individuals. Child Development, 74, 1840 – 1856. Coie, J. D., Dodge, K. A., & Coppotelli, H. (1982). Dimen-
Adler, A. (1969). Individual psychology. Totowa, NJ: Little- sions and types of social status: A cross-age perspective.
field and Adams. Developmental Psychology, 18, 557 – 570.
Adler, P. A., & Adler, P. (1995). Dynamics of inclusion and Crick, N. R., & Grotpeter, J. K. (1995). Relational aggres-
exclusion in pre-adolescent cliques. Social Psychology sion, gender, and social-adjustment. Child Development,
Quarterly, 58, 145 – 162. 66, 710 – 722.
Adler, P. A., Kless S. J., & Adler, P. (1992). Socialization to Crockett, L., Losoff, M., & Petersen, A. C. (1984). Percep-
gender roles: Popularity among elementary school boys tions of the peer group and friendship in early adoles-
and girls. Sociology of Education, 65, 169 – 187. cence. Journal of Early Adolescence, 4, 155 – 181.
Aiken, L. S., & West, S. G. (1991). Multiple regression: Testing Crosnoe, R., & Needham, B. (2004). Holism, contextual
and interpreting interactions. Newbury Park, CA: Sage. variability, and the study of friendships in adolescent
Baumeister, R. F., & Leary, M. R. (1995). The need to belong: development. Child Development, 75, 264 – 279.
Desire for interpersonal attachments as a fundamental Dishion, T. J., Patterson, G. R., Stoolmiller, M., & Skinner, M. L.
human motivation. Psychological Bulletin, 117, 497 – 529. (1991). Family, school, and behavioral antecedents to
Bendixen, M., & Olweus, D. (1999). Measurement of early adolescent involvement with antisocial peers.
antisocial behavior in early adolescence and adoles- Developmental Psychology, 27, 172 – 180.
cence: Psychometric properties and substantive find- Dishion, T. J., Spracklen, K. M., Andrews, D. M., &
ings. Criminal Behavior and Mental Health, 9, 323 – 352. Patterson, G. R. (1996). Deviancy training in male ado-
Brener, N. D., Billy, J. O. G., & Grady, W. R. (2003). lescent friendships. Behavior Therapy, 27, 373 – 390.
Assessment of factors affecting the validity of self- Eder, D. (1985). The cycle of popularity: Interpersonal
reported health-risk behavior among adolescents: Evi- relations among female adolescents. Sociology of Educa-
dence from the scientific literature. Journal of Adolescent tion, 58, 154 – 165.
Health, 33, 436 – 457. Ellemers, N. (1993). The influence of socio-structural
Brown, B. B. (1990). Peer groups and peer cultures. In variables on identity enhancement strategies. European
S. S. Feldman & G. R. Elliott (Eds.), At the threshold: Review of Social Psychology, 4, 427 – 457.
The developing adolescent (pp. 171 – 196). Cambridge, MA: Espelage, D. L., Holt, M. K., & Henkel, R. (2003). Examina-
Harvard University Press. tion of peer-group contextual effects on aggression during
Brown, B. B., Clasen, D. R., & Eicher, S. A. (1986). early adolescence. Child Development, 74, 205 – 220.
Perceptions of peer pressure, peer conformity disposi- Gest, S. D., Farmer, T. W., Cairns, B. D., & Xie, H. (2003).
tions, and self-reported behavior among adolescents. Identifying children’s peer social networks in school
Developmental Psychology, 22, 521 – 530. classrooms: Links between peer reports and observed
Brown, B. B., & Lohr, M. (1987). Peer-group affiliation and interactions. Social Development, 12, 513 – 529.
adolescent self-esteem: An integration of ego-identity Gest, S. D., Graham-Bermann, S. A., & Hartup, W. W.
and symbolic-interaction theories. Journal of Personality (2001). Peer experience: Common and unique features of
and Social Psychology, 52, 48 – 55. number of friendships, social network centrality, and
Bryant, A. L., Schulenberg, J., Bachman, J. G., O’Malley, P. M., sociometirc status. Social Development, 10, 23 – 40.
& Johnston, L. D. (2000). Understanding the links among Hartup, W. W. (1993). Adolescents and their friends. In
school misbehavior, academic achievement, and cigarette B. Laursen (Ed.), Close friendships in adolescence (pp.
use: A national panel study of adolescents. Prevention 3 – 22). San Francisco: Jossey-Bass.
Science, 2, 71 – 87. Hartup, W. W., & Stevens, N. (1997). Friendships and
Bryk, A. S., & Raudenbush, S. W. (1992). Hierarchical linear adaptations in the life course. Psychological Bulletin, 121(30),
models. Newbury Park, CA: Sage. 355 – 370.
1254 Ellis and Zarbatany

Harris, J. R. (1995). Where is the child’s environment? A Masten, A. S., Morison, P., & Pellegrini, D. S. (1985). A
group socialization theory of development. Psychological revised class play method of peer assessment. Develop-
Review, 102, 458 – 489. mental Psychology, 21, 523 – 533.
Hawley, P. H. (1999). The ontogenesis of social dominance: Moffitt, T. E. (1993). Adolescence-limited and life-course
A strategy-based evolutionary perspective. Developmen- persistent antisocial behavior: A developmental taxon-
tal Review, 19, 97 – 132. omy. Psychological Review, 4, 674 – 701.
Hawley, P. H. (2003). Prosocial and coercive configurations Nelson, S., & Dishion, T. (2004). From boys to men:
of resources control in early adolescence: A case for the Predicting adult adaptation from middle childhood
well-adapted Machiavellian. Merrill-Palmer Quarterly, 49, sociometric status. Development and Psychopathology, 16,
279 – 309. 441 – 459.
Horn, S. (2006). Group status, group bias, and adolescents’ Parkhurst, J. T., & Hopmeyer, A. G. (1998). Sociometric
reasoning about the treatment of others in school con- popularity and peer-perceived popularity: Two distinct
texts. International Journal of Behavioral Development, 30, dimensions of peer status. Journal of Early Adolescence, 18,
208 – 218. 125 – 144.
Hymel, S., Rubin, K. H., Rowden, L., & LeMare, L. (1990). Poulin, F., Cillessen, A., Hubbard, J. A., Coie, J. D.,
Children’s peer relationships: Longitudinal prediction of Dodge, K. A., & Schwartz, D. (1997). Children’s friends
internalizing and externalizing problems from middle to and behavioral similarity in two social contexts. Social
late childhood. Child Development, 61, 2004 – 2021. Development, 6, 224 – 236.
Jenkins, P. H. (1997). School delinquency and the school Rodkin, P. C., Farmer, T. W., Pearl, R., & Van Acker, R.
social bond. Journal of Research in Crime and Delinquency, (2000). Heterogeneity of popular boys: Antisocial and
34, 337 – 367. prosocial configurations. Developmental Psychology, 36,
Kiesner, J., Cadinu, M., Poulin, F., & Bucci, M. (2002). 14 – 24.
Group identification in early adolescence: Its relation Rose, A., Swenson, L., & Waller, E. (2004). Overt and
with peer adjustment and its moderator effects on peer relational aggression and perceived popularity: Devel-
influence. Child Development, 73, 196 – 208. opmental differences in concurrent and prospective
Kindermann, T. A. (1993). Natural peer groups as contexts relations. Developmental Psychology, 40, 378 – 387.
for individual development: The case of children’s Ryan, A. M. (2001). The peer group as a context for the
motivation in school. Developmental Psychology, 29, development of young adolescent motivation and
970 – 977. achievement. Child Development, 72, 1135 – 1150.
Kupersmidt, J. B., Burchinal, M., & Patterson, C. J. (1995). Santor, D. A., Messervey, D., & Kusumakar, V. (2000).
Developmental patterns of childhood peer relations as Measuring peer pressure, popularity and conformity in
predictors of externalizing behavior problems. Develop- adolescent boys and girls: Predicting school perfor-
ment and Psychopathology, 7, 825 – 843. mance, sexual attitudes, and substance abuse. Journal
Kupersmidt, J. B., DeRosier, M. E., & Patterson, C. (1995). of Youth and Adolescence, 29, 165 – 182.
Similarity as the basis for children’s friendships: The Stewart-Knox, B., Sittlington, J., Rugkasa, J., Harrisson, S.,
roles of sociometric and withdrawn behavior, academic Treacy, M., & Abaunza, P. (2005). Smoking and peer
achievement and demographic characteristics. Journal of groups: Results form a longitudinal qualitative study of
Social and Personal Relationship, 12, 439 – 452. young people in Northern Ireland. British Journal of
Ladd, G. W. (1983). Social networks of popular, average Social Psychology, 44, 397 – 414.
and rejected children in school settings. Merrill-Palmer Tajfel, H. (1972). Some developments in European social
Quarterly, 29, 283 – 307. psychology. European Journal of Social Psychology, 2,
LaFontana, K. M., & Cillessen, A. H. N. (2002). Children’s 307 – 322.
perceptions of popular and unpopular peers: A multi- Tarrant, M. (2002). Adolescent peer groups and social
method assessment. Developmental Psychology, 38, 635 – 647. identity. Social Development, 11, 110 – 123.
Lease, M., Musgrove, K., & Axelrod, J. (2002). Dimensions Urberg, K. A., Degirmencioglu, S., & Pligrim, C. (1997).
of social status in preadolescent peer groups: Likeability, Close friend and group influence on adolescent cigarette
perceived popularity and social dominance. Social Devel- smoking and alcohol use. Developmental Psychology, 33,
opment, 11, 508 – 533. 834 – 844.
Lightfoot, C. (1992). Constructing self and peer culture: Xie, H., Li, Y., Boucher, S. M., Hutchins, C., & Cairns, B.
A narrative perspective on adolescent risk taking. In (2006). What makes a girl (or boy) popular (or unpop-
L. T. Winegar & J. Valsiner (Eds.), Children’s development ular)? African American children’s perceptions and
within social context: Vol. 2. Research and methodology developmental differences. Developmental Psychology, 42,
(pp. 229 – 245). Mahwah, NJ: Lawrence Erlbaum Associates. 599 – 612.

You might also like