social aggression in children and adolescents_ a meta-analytic re.pdf
TRANSCRIPT
-
University of MiamiScholarly Repository
Open Access Dissertations Electronic Theses and Dissertations
2011-08-02
Social Aggression in Children and Adolescents: AMeta-Analytic ReviewCathy LongaUniversity of Miami, [email protected]
Follow this and additional works at: http://scholarlyrepository.miami.edu/oa_dissertations
This Open access is brought to you for free and open access by the Electronic Theses and Dissertations at Scholarly Repository. It has been accepted forinclusion in Open Access Dissertations by an authorized administrator of Scholarly Repository. For more information, please [email protected].
Recommended CitationLonga, Cathy, "Social Aggression in Children and Adolescents: A Meta-Analytic Review" (2011). Open Access Dissertations. Paper 625.
-
UNIVERSITY OF MIAMI
SOCIAL AGGRESSION IN CHILDREN AND ADOLESCENTS: A META-ANALYTIC REVIEW
By
Cathy Longa
A DISSERTATION
Submitted to the Faculty of the University of Miami
in partial fulfillment of the requirements for the degree of Doctor of Philosophy
Coral Gables, Florida
August 2011
-
2011 Cathy Longa
All Rights Reserved
-
UNIVERSITY OF MIAMI
A dissertation submitted in partial fulfillment of the requirements for the degree of
Doctor of Philosophy
SOCIAL AGGRESSION IN CHILDREN AND ADOLESCENTS: A META-ANALYTIC REVIEW
Cathy Longa Approved: Debbiesiu Lee, Ph.D. Assistant Professor of Educational and Psychological Studies
Terri A. Scandura, Ph.D. Dean of the Graduate School
Soyeon Ahn, Ph.D. Assistant Professor of Educational and Psychological Studies
Ora Prilleltensky, Ed.D. Clinical Assistant Professor of Educational and Psychological Studies
Annette La Greca, Ph.D. Professor of Psychology and Pediatrics
-
LONGA, CATHY (Ph.D., Counseling Psychology) Social Aggression in Children and Adolescents: (August 2011) A Meta-Analytic Review Abstract of a dissertation at the University of Miami. Dissertation supervised by Professors Debbiesiu Lee and Soyeon Ahn. No. of pages in text. (112)
Social aggression has been widely studied; however, findings have been inconsistent
leading to confusion within the current literature. Previous research has linked social
aggression to negative outcomes; including poor peer relations, internalizing symptoms,
and low levels of empathy; as well as positive attributes, including prosocial behaviors,
high social status, and social intelligence. This meta-analysis examined the relationship
between social aggression and various correlates, both positive and negative, as well as
how age and gender moderate these relationships. With 896 correlations derived from
108 studies (of a total of 107 published articles), the results using the random-effects
model for computing overall effect sizes indicated that social aggression is related with
maladaptive correlates, such as externalization ( r = 0.46), internalization ( r = 0.16),
negative individual traits ( r = 0.32), as well as negative peer relations ( r = 0.28).
However, findings also suggest that social aggression is associated with popularity ( r =
0.22) and social skillfulness ( r = 0.16). Implications, limitations, and future directions
are discussed.
-
iii
TABLE OF CONTENTS LIST OF FIGURES .............................................................................................................v LIST OF TABLES ............................................................................................................. vi CHAPTER 1. INTRODUCTION ........................................................................................1 CHAPTER 2. LITERATURE REVIEW .............................................................................4
Theoretical Models ......................................................................................................... 6
Early childhood. .......................................................................................................... 8 Middle Childhood. ...................................................................................................... 9 Adolescence. ............................................................................................................. 12
Previous Meta-Analyses on Social Aggression ............................................................ 14 Current Study ................................................................................................................ 17 Meta-Analysis ............................................................................................................... 18
CHAPTER 3. METHODS .................................................................................................20
Search Procedure .......................................................................................................... 20 Coding of Studies .......................................................................................................... 23 Power ............................................................................................................................ 25 Effect Size ..................................................................................................................... 26 Analysis......................................................................................................................... 27
CHAPTER 4. RESULTS ...................................................................................................29
Characteristics of Studies .............................................................................................. 29 Publication Bias ............................................................................................................ 30 Overall Model ............................................................................................................... 30
Gender differences. ................................................................................................... 32
-
iv
Age differences. ........................................................................................................ 32 Relationship Between Individual Characteristics and Social Aggression .................... 36
Gender differences. ................................................................................................... 39 Age differences. ........................................................................................................ 42
Relationship Between Peer Relations and Social Aggression ...................................... 46
Gender differences. ................................................................................................... 46 Age differences. ........................................................................................................ 49
CHAPTER 5. DISCUSSION .............................................................................................53
Future Interventions ...................................................................................................... 65 Future Research ............................................................................................................ 69 Limitations .................................................................................................................... 71 Conclusion .................................................................................................................... 72
REFERENCES ..................................................................................................................74 APPENDIX ........................................................................................................................93
-
v
LIST OF FIGURES
Figure 1. Study selection chart...........................................................................................21 Figure 2. Funnel plot of effect sizes against study sizes ....................................................31 Figure 3. Overall model .....................................................................................................33
-
vi
LIST OF TABLES
Table 1. Overall Model of Relationships with Social Aggression .................................... 31 Table 2. Differences by Gender in Overall Model. .......................................................... 34 Table 3. Differences by Age in Overall Model. ................................................................ 35 Table 4. Relationship between Individual characteristics and Social Aggression. ........... 37 Table 5. Differences by Gender in the Relationship between Social Aggression and Individual Characteristics. ................................................................................................ 40 Table 6. Differences by Age in the Relationship between Social Aggression and Individual Characteristics ................................................................................................ 43 Table 7. Relationship between Peer Relations and Social Aggression. ............................ 47 Table 8. Differences by Gender in the Relationship between Social Aggression and Peer Relations. .......................................................................................................................... 48 Table 9. Differences by Age in the Relationship between Social Aggression and Peer Relations. .......................................................................................................................... 50
-
1
CHAPTER 1
INTRODUCTION
It is no question that children can be extremely cruel to one another, as aggression
is a widely studied construct within the psychological literature. Aggressive behaviors
can take a variety of forms: physical aggression, verbal insults, relationship manipulation,
and nonverbal expressions of disdain. Much of the previous research on aggression has
focused on physical aggression (Underwood, Galen, & Paquette, 2001). Recently, there
has been an explosion of interest in forms of aggression that emphasize nonphysical
strategies: indirect aggression (Bjorkqvist, Lagerspetz, & Kaukiainen, 1992), relational
aggression (Crick & Grotpeter, 1995), and social aggression (Galen & Underwood,
1997).
Much has been made in the media and popular psychology books (e.g., Dellasega
& Nixon, 2003; Simmons, 2002; Wiseman, 2002) concerning the negative consequences
of social aggression, particularly as it relates to girls. It has been suggested that socially
aggressive behaviors are maladaptive, hurtful, and related to negative outcomes in both
the victim and the aggressor (e.g., Bjorkqvist, Osterman, & Lagerspetz, 1994; Crick,
Casas, & Mosher, 1997; Heilbron &Prinstein, 2008). There have even been efforts to
determine if socially aggressive behaviors belong within the Diagnostic and Statistical
Manual [DSM] (Keenan, Coyne, & Lahey, 2008).
Although the introduction of these constructs has generated a very useful body of
research (Underwood, Galen, & Paquette, 2001), within this literature many
disagreements, discrepancies, and questions remain. Included among these incongruities
is whether social aggression represents a form of maladjustment arising from deficiencies
-
2
and leading to negative outcomes, or a form of socially skillful, though perhaps not
desirable, behaviors leading to positive social attributes. One way to explain these
differences in the literature is to consider a developmental perspective, where aggression
is proposed to change with development. As children mature, their social skills, cognitive
abilities, and emotional development also change. Therefore, it is reasonable to expect
that the nature, correlates, and consequences of social aggression may also change with
age. Thus, it is fair to expect differing associations between social aggression and various
outcomes based on developmental stage.
Two previous meta-analyses (Archer, 2004; Card, Stucky, Sawalani, & Little,
2008) have been conducted to explore and better understand social aggression. Archers
(2004) meta-analytic review of gender differences indicated more evidence of social
aggression in girls among children; however the gender differences were not as consistent
as would have been expected from previous individual studies. Archers (2004) study is
limited in that he utilized a more narrow definition of social aggression, searching only
for indirect aggression, therefore including only 41 studies of this construct. In
addition, he used a fixed-effects model for his analysis, which may have affected the
generalization of findings.
Card et al.s (2008) meta-analysis found that girls use more social aggression than
boys, but that this result was trivial in magnitude. In addition, Card et al. found that
social aggression was more strongly associated with internalizing problems (i.e.,
depression and anxiety) than physical aggression, with no gender moderation of these
relationship found. Card et al.s (2008) study is limited by several issues, such as
excluding many studies that solely focused on social aggression, failing to examine age
-
3
as a potential moderator, and not examining social adjustment or potentially positive
outcomes associated with social aggression.
Although both studies provide useful information concerning social aggression,
the studies have limitations and questions remain. Given these discrepancies in the
literatures, it seems that a comprehensive review and analysis of the literature is
warranted. By including more studies, examining age as a moderator, and examining a
wider range of correlates, a better understanding of social aggression can be yielded.
Thus, this study aimed at clarifying and providing a better understanding of the correlates
(both positive and negative) of social aggression, while examining the role age and
gender play in these relationships.
-
4
CHAPTER 2
LITERATURE REVIEW
Researchers have struggled for decades to describe and measure aggressive
behavior (see Coie & Dodge, 1998, for a review). Despite the numerous articles devoted
to this topic, there are significant challenges for understanding aggression. One
significant difficulty is the vast number of subtypes suggested by investigators. For
example, Underwood, Galen, and Paquette (2001) argue that aggression is difficult to
define operationally because neither the intentions of an aggressor nor the perceptions of
harm by a victim can be directly observed. What is considered aggressive behavior relies
greatly on social judgments, of both the aggressor and the perceiver (Underwood, Galen,
& Paquette, 2001).
In general, definitions of aggression involve the intent to inflict harm on others
(Archer & Coyne, 2005). Harre and Lamb (1983) noted that researchers have proposed
over 200 different definitions of aggressive behavior, but that most of these definitions
share two common features: (1) the behavior is intended to harm, and (2) the behavior is
perceived as hurtful by the victim. Direct acts of verbal and physical aggression in
situations of interpersonal conflict readily fit such a definition. However, research has
demonstrated that children engage in a variety of aggressive behavior. Hence, although
the injury necessary for a behavior to be labeled aggression has been most often
interpreted by researchers to mean bodily harm, these terms could also apply to damage
to ones self-esteem or social standing (Galen & Underwood, 1997).
In considering the various types of aggression it is important to note that all forms
of aggression can be considered social strategies that have evolved to represent different
-
5
means to harm others (Archer & Coyne, 2005). In arguing the significance of
nonphysical forms of aggression, Olweus (1996) stated that negative actions can be
carried out by physical contact, by words, or in other ways, such as making faces and
nasty gestures or by intentional exclusion from a group (p. 16). While physical
aggression involves overt actions such as threatening physical injury, nonphysical forms
of aggression are behaviors intended to cause social or interpersonal harm through overt
or covert ways, such as gossiping, excluding from social interactions, and spreading
rumors (Archer & Coyne, 2005; Heilbron & Prinstein, 2008).
Nonphysical forms of aggression have been given three different names in the
literature: indirect aggression (Lagerspetz, Bjorkqvist, & Peltonen, 1988), relational
aggression (Crick & Grotpeter, 1995), and social aggression (Cairns, Cairns,
Neckerman, Feguson, & Gariepy, 1989). Indirect aggression refers to behaviors such as
gossiping or exclusion that harms others without direct confrontation of the victim
(Lagerspetz et al., 1988). Relational aggression involves harming others by manipulating
peer relationships (Crick, 1995). Finally, social aggression refers to behaviors aimed at
damaging the victims self-esteem or social status. Social aggression includes actions
such as negative facial expressions, verbal rejection, rumors, or social exclusion (Galen
& Underwood, 1997).
Researchers using these three labels disagree over which term is the most
practical for describing these nonphysical forms of aggression. Further, these is some
dispute regarding how similar indirect, relational, and social aggression really are, as well
as which term best depicts the behaviors (Bjorkqvist, 2001; Underwood, et al., 2001).
Although there may be subtle differences, the constructs converge around the common
-
6
theme of behaviors that attack a victims actual or perceived social relations with others,
often (though not always) in ways that avoid direct confrontation (Card, et al., 2008). For
the purposes of this paper, the term social aggression will be utilized to refer to the
behaviors in question.
Along with the general constructs of aggression, bullying can be characterized as
a subset of aggressive behaviors. As with aggressive behaviors, bullying intentionally
causes harm to the recipient (Aluede, Adeleke, Omoike, & Afen-Akpaida, 2008). This
harm can be both physical and psychological and has an influence on the victims
physical, emotional, social, and educational well-being (Smith, 2004). Rigby (2002)
points out that it seems like splitting hairs to distinguish between aggression and
bullying (p. 30).
Recent research on bullying has differentiated between the various types of
bullying. Like aggression, bullying can manifest in several ways, including physical
bullying, verbal bullying, and social bullying (e.g., disrupting the social relationships
between victims and peers; Berger, 2007). Given that bullying is a subset of the broader
category of aggression, and that researchers have acknowledged that both bullying and
aggression can be classified as indirect or social, it seems odd that researchers continue to
study these constructs as completely separate phenomena. Therefore, this study included
both social aggression and nonphysical forms of bullying to obtain a better understanding
of the similar behaviors that frequently occur within the child and adolescent population.
Theoretical Models
To varying extents, theoretical models are beginning to be constructed to explain
social aggression. In general, these theories can be divided into two groups: those who
-
7
argue that social aggression is maladaptive and engaged in by children who have a deficit
in their social or cognitive competence (e.g., Crick & Dodge, 1994; Michiels, Gietens,
Onghena, Kuppens, 2008; Prinstein & Cillessen, 2003), and those who argue that social
aggression is used by socially skilled and manipulative individuals (e.g., Hawley, 2007;
Garandeau & Cillessen, 2006). The first group links aggression with negative outcomes,
including poor peer relations, internalizing symptoms, and low levels of empathy. The
latter group links aggression with positive attributes, including prosocial behaviors, high
social status, and social intelligence. Interestingly, there is evidence to support both
theories.
One way to explain why some theorists view social aggression as maladaptive and
other theorists see it as beneficial may be that this behavior has different effects at
different developmental stages. A developmental model of social aggression can assist in
our understanding of developmental processes over time by highlighting the importance
of individual and environmental interactions (Pepler & Craig, 2005; Rutter, 1990). From
a developmental perspective, the nature and form of aggression is assumed to change
with development. These changes emerge as a result of maturation within the child
together with changing social interactions and expectations (Pepler & Craig, 2005). Thus,
it is reasonable to expect differing associations between social aggression and various
outcomes based on developmental stage.
Although the conceptualization of aggression as a form of psychopathology
predominantly dominates the empirical literature, there are notable exceptions to this
emphasis (e.g., Hawley et al., 2007). For some children and adolescents, socially
aggressive behavior is positively related to dominance, social impact, perceived
-
8
popularity, and higher levels of social power (Heilbron & Prinstein, 2008). A review of
the literature on social aggression reveals conflicting findings at different developmental
stages (i.e., early childhood, middle childhood, adolescence).
Early childhood. Although researchers have only recently begun to investigate
the occurrence of social aggression in early childhood, some researchers suggest that
preschool children as young as age 3 engage in these behaviors (e.g., Crick, Casas,
Mosher, 1997; Juliano, Werner, Cassidy, 2006; McNeilly-Choque, Hart, Robinson,
Nelson, Olsen, 1996). Crick and colleagues (1999) suggest that the manifestations of
social aggression at these ages are simple and obvious (e.g., directly telling a peer they
will not play with them unless certain conditions are met) and enacted as a response to an
immediate problem. Thus, young children are beginning to learn social skills and as such
social aggression at this developmental stage may take the form of overt, straightforward
forms (Underwood, 2003).
Although the study of social aggression in early childhood remains a relatively
new area of research, several studies have examined the link between social aggression
and measures of problematic adjustment, positive attributes, social skills, and language
ability with mixed findings. For example, social aggression has been associated with
impulsive, oppositional, anxious, or depressive behaviors (Juliano, et al., 2006), as well
as high levels of deception (Ostrov, 2006) and poor expressive language skills (Estrem,
2005). However, other studies have found social aggression related with a variety of
markers of positive adjustment in young children, including verbal skills (Bonica, et al.,
2003) and receptive language ability (Hawley, 2003), moral maturity (in girls; Hawley,
2003) and the ability to express guilt (Hawley, 2003).
-
9
With regard to social relationships, research has indicated that social aggression is
associated with current and future peer rejection (Crick, et al., 1997; Crick, Ostrov, &
Werner, 2006; Ostrov & Crick, 2007), peer exclusion (Ostrov & Keating, 2004), lower
prosocial behavior (Crick, et al., 1997; Ostrov & Keating, 2004), and lower levels of peer
acceptance (McNeilly-Choque et al., 1996). In contrast, other studies have suggested that
social aggression is linked with average to above average social skills (Carpenter &
Nangle, 2003), social competence (Vaugh, Vollenweider, & Bost, 2003), increased
friendship intimacy (Burr, et al., 2005), higher numbers of stable and mutual friendships
over the course of the school year (Burr, et al., 2005), higher peer acceptance (Crick, et
al., 1997), and higher sociability (Nelson, Robinson, & Hart, 2005).
Middle Childhood. The socially aggressive acts employed by grade-schoolers
likely reflect the social and cognitive advances of the middle-childhood period (Crick et
al., 1999). Although important at any age, needs for acceptance by peers and for mutually
satisfying friendships become increasingly relevant during the grade-school years and are
typically met through relationships with same-gender peers (Bukowski & Hoza, 1989;
Crick et al., 1999; Maccoby, 1990). Bjrkqvist and colleagues (1992) posits that because
physical aggression becomes socially unacceptable as children mature, social aggressive
strategies replace overt aggression. Furthermore, these researchers propose that as
children mature cognitively, they reduce their use of physical forms of aggression and
increase their use of social forms of aggression (Bjrkqvist, et al., 1992). It seems likely
that the increase in the importance of social issues (i.e., socially accepted behaviors), in
addition to the language and cognitive skills acquired during middle childhood (e.g.,
increases in memory and vocabulary, the ability to view ones own thoughts, feelings and
-
10
behaviors from another persons perspective), contributes to childrens ability to use the
peer group as an effective means for hurting others (Crick et al., 1999).
Accordingly, the socially aggressive behaviors expected at this developmental
stage are suggested to involve relatively covert actions, with less confrontation and a
focus on the interactions with other peer group members (Crick et al., 1999). These
covert forms of social aggression require a certain degree of cognitive complexity for
both the aggressor (i.e., to recognize that the action will be an effective strategy) and the
victim (i.e., to recognize that others are being mean to them on behalf of the aggressor;
Crick et al., 1999). The ability to understand anothers perspective and information in the
broader social network makes the nonconfrontational aggressive strategy more feasible at
this period of development (Xie, Cairns, & Cairns, 2005). Thus, as social competence
increases, children learn additional ways of harming others, including ways that do not
place him or her at the direct risk of retaliation (Bjrkqvist, et al., 1992). However, other
researchers argue that it is the lack of social-emotional competencies that contribute to
the development of aggressive behaviors (Benson, 2006).
Social-emotional learning is described as the process of acquiring competencies to
regulate emotions, understand the perspectives of others, establish and maintain positive
relationships with others, and manage challenging interpersonal situations (Durlak et al.,
2011; Elias et al., 2007). It is posited that these abilities provide a basis for positive
social actions, less behavioral problems, and better emotional regulation (Greenberg et
al., 2003). Theorists suggest that by achieving social-emotional competence, children
and adolescents learn to internalize values and beliefs that promote caring and concern
-
11
for others, as well as taking responsibility for ones choices and behaviors (Bear &
Watkins, 2006).
Developmental researchers suggest that failure to acquire social-emotional
competencies is associated with a variety of personal and social difficulties (e.g.,
Eisenberg, 2006; Guerra & Bradshaw, 2008). Research has suggested that many students
do not develop these social-emotional abilities as they mature which, in turn, has a poor
effect on behavior. For example, in one study, social competencies such as empathy and
conflict resolution skills were only noted in about 30% of a national sample of middle-
and high-school students (Benson, 2006). Similar to research among children in early
childhood, studies examining social aggression among children in middle childhood have
also examined the association between social aggression and measures of internalizing
and externalizing difficulties. Numerous studies have linked social aggression with
internalizing and externalizing symptoms (Crick, 1997), specifically higher levels of
anxiety (Marsee, Weems, & Taylor, 2008), depressive symptoms (Crick & Grotpeter,
1995; Henington, et al., 1998; Murray-Close & Crick, 2007; Zimmer-Gembeck, Hunter,
& Pronk, 2007), general aggression (Crick, Ostrov, & Werner, 2006), delinquency (in
boys; Crick, Ostrov, & Werner, 2006), impulsivity (Musher-Eizenman, et al., 2004),
attention-deficit hyperactivity disorder (Zalecki & Hinshaw, 2004), as well as lower
emotional regulation (Bowie, 2010). However, there has been a lack of attention paid to
the measurement of possible indices of positive adaption (Heilbron & Prinstein, 2008).
In terms of social relationships, research has also found discrepant findings within
the middle childhood population. For example, studies have indicated that social
aggression is associated with increased peer rejection (Crick, 1996; Crick & Grotpeter,
-
12
1995; Rys & Bear, 1997; Tomada & Schneider, 1997), lower social acceptance (Crick,
1996; Henington, et al., 1998; Lancelotta & Vaughn, 1989), lower social preference
(Johnson & Foster, 2005; Murray-Close & Crick, 2006; Werner & Crick, 2004; Zimmer-
Gembeck, et al., 2007), and fewer mutual friendships (Johnson & Foster, 2005). In
contrast, other studies have suggested that social aggression is positively associated with
social intelligence (Kaukiainen, et al., 1999), peer-perceived popularity (Andreou, 2006;
Lease, Kennedy, & Axelrod, 2002), and higher rates of friendship exclusivity, self-
disclosure by friends, and social intimacy (Grotpeter & Crick, 1996).
Adolescence. The developmental period of adolescence contains several
challenges including puberty, entering new schools, beginning to separate from parents,
developing an individual identity, and, for many, developing romantic relationships
(Underwood, 2003). Social methods of aggression may be favored by adolescents
because they are considered socially acceptable and less easy to identify by on-lookers
as bullying or aggression. Furthermore, the increased importance of social status and
romantic relationships may have an effect on interpersonal relations of adolescents (Xie,
Cairns, & Cairns, 2005). Competition for friends, a romantic partner, and social visibility
or status may generate interpersonal conflicts that evoke the use of social aggression
(Xie, Cairns, & Cairns, 2005). According to Underwood (2003), younger adolescents
social aggression might be used to protect the boundaries of social groups and for
confirming high peer status. For older adolescents, social aggressive strategies,
particularly gossip, may be used less for self-protection and more in the service of self-
understanding, identity development, and moral negotiation (Underwood, 2003). During
adolescence, social means of aggression are expected to coexist with direct verbal means,
-
13
with the individual using the strategy that is most suitable in the particular situation
(Bjorkqvist, et al., 1992).
Research among adolescent samples have resulted in much the same types of
findings as those for middle childhood. Several studies have documented both
externalizing and internalizing difficulties among socially aggressive adolescents. For
example, research has revealed positive associations with externalizing behavior
(Prinstein, Boergers, & Vernberg, 2001), anger to provocation (Marsee & Frick, 2007),
callous-unemotional traits (Marsee & Frick, 2007), lower empathy (Kaukiainen, et al.,
1999), social anxiety (Loukas, Paulos, & Robinson, 2005), and negative self-
representation (Moretti, Holland, & McKay, 2001). Conversely, several studies have
demonstrated links between social aggression and positive adjustment among adolescents
(Heilbron & Prinstein, 2008). For example, studies have shown that socially aggressive
individuals have higher levels of social intelligence (Kaukiainen, et al., 1999), ability to
influence peers (Xie, et al., 2002), and ratings of Olympian traits (i.e., athletic,
attractive; Xie, et al., 2002).
Investigations into the social relationships of socially aggressive adolescents have
yielded mixed findings. Some studies have suggested that social aggression is positively
associated with higher peer acceptance (Cillessen & Borch, 2006; Salmivalli,
Kaukiainen, & Lagerspetz, 2000), social preference (Prinstein & Cillessen, 2003), peer-
perceived popularity (e.g., Cillessen & Borch, 2006; Cillessen & Mayeux, 2004;
Leadbeater, Boone, Sangster, & Mathieson, 2006; Vaillancourt & Hymel, 2006), social
network centrality (Xie, et al., 2002), and social impact (Zimmer-Gembeck, et al., 2005),
as well as lower levels of peer rejection (Salmivalli, et al., 2000). Conversely, other
-
14
studies have indicated that socially aggressive adolescents have higher levels of peer
rejection (Vaillancourt & Hymel, 2006) and lower levels of social preference (Cillessen
& Mayeux, 2004; Lafontana & Cillessen, 2002; Rys & Bear, 1997; Zimmer-Gembeck, et
al., 2005).
Previous Meta-Analyses on Social Aggression
Social aggression has been explored in two previous meta-analyses (Archer,
2004; Card, et al., 2008). Archers (2004) meta-analytic review of gender differences in
various forms of aggression included 41 published and unpublished studies that examined
social forms of aggression. These researchers found more evidence of social aggression
in the female direction among children. However, the sex differences were not as large or
consistent as would have been expected from previous individual studies of children
(Archer, 2004).
Archer (2004) found that the method of measurement for social aggression (i.e.,
self-report, peer report, teacher report, and observation) was an important moderator to
explain variations in study findings. Specifically, when social aggression was measured
using observations, the largest gender difference on social aggression was found,
indicating that females showed more social aggression (d = -.74); however, this result
was based on only four studies. Data from peer-ratings suggested that females engage in
more social aggression (d = -.19), and gender differences increases with age (d = -.00 for
under age 11; d = -.13 for ages 12-13; d = -.35 for ages 14-17). Similarly, teacher-reports
also revealed social aggression in the female direction (d = -.13). In contrast, although
small, self-reports suggested that males engage in greater levels of social aggression (d =
-
15
.03). This effect also increased with age (d = .03 for ages 6-13; d = .12 for ages 14-17;
Archer, 2004).
While Archers (2004) study provides a basis for understanding gender
differences in aggression, the analysis has several limitations. Archer only used the term
indirect aggression in its search for relevant articles, therefore including only 41 studies
of this construct, Further, inclusion required the sample to contain both males and
females. In addition, Archer used a fixed-effects model for his analysis, which may have
affected the generalization of findings.
In 2008, Card et al. conducted a meta-analysis of 107 published and unpublished
studies examining the correlation between gender and type of aggression (i.e., physical
and social), the relationship between physical and social aggression, and the association
between type of aggression and psychosocial adjustment. In terms of gender differences,
Card et al.s (2008) results found that girls use more social aggression than boys, but that
this result, although statistically significant (r = -.03), was trivial in magnitude and
therefore negligible (p. 1194). Similar to Archer (2004), Card and colleagues found
method of measurement (i.e., self-report, peer report, teacher report, parent report, and
observation) moderation effects when looking at the relationship between gender and
social aggression. However, in contrast to Archer, Card et al.s meta-analysis indicated
that the magnitude of such effects was trivial regardless of the reporter. In particular,
Card et al.s findings found statistically significant gender differences for social
aggression in the female direction based on parent (r = -.08) and teacher (r = -.07), but
not researcher-observation (r = -.05) or peer nominations (r = -.02). Statistically
-
16
significant gender differences in the male direction were found based on self-report (r =
.03).
Card et al.s (2008) meta-analysis also found physical and social aggression to be
strongly correlated (r = .76), yet separable demonstrating the presence of two related, yet
different constructs. In addition, Card et al. explored the relationship between form of
aggression and psychosocial adjustment. Their findings indicate that physical and social
aggression are uniquely related to various aspects of maladjustment. Specifically,
physical aggression was more strongly related to emotional dysregulation and conduct
problems, while social aggression was more strongly associated with internalizing
problems (i.e., depression and anxiety). No gender moderation of these relationships was
found.
Card et al.s (2008) study provides a useful step in the understanding of social
aggression, however is also limited by several issues. First, their criteria for study
inclusion into the meta-analysis (e.g., only studies including both social and physical
aggression) excluded many studies whose focus was solely on social aggression. Second,
these authors did not examine age as a potential moderator for the relationships in
question. Third, the psychosocial adjustment factors studied by Card et al. represent a
limited representation of the possible correlates associated with social aggression. By
focusing solely on maladjustment in the form of internalizing and externalizing
behaviors, Card et al. did not examine social adjustment or the potential positive
outcomes associated with social aggression, thereby limiting the results and subsequent
understanding of this phenomenon.
-
17
Thus, although both meta-analyses (Archer, 2004; Card et al., 2008) have
provided useful information, the studies have limitations and the results raise further
questions. By including more studies using a broader definition of social aggression,
including age as a moderator, and examining all potential correlates (both positive and
negative), a better understanding of social aggression can be yielded. This understanding
can, in turn, affect the generalizability of study findings.
Current Study
From the review above it is clear that there are many inconclusive findings in the
research concerning social aggression. Within the literature, social aggression has been
viewed as either maladaptive or adaptive, each with empirical evidence to support the
claim. Some researchers have suggested that social aggression is related to negative
outcomes such as peer rejection, psychological distress, and behavioral difficulties.
Conversely, other researchers have argued that although socially aggressive behaviors are
not desirable, they are associated with positive outcomes, such as peer acceptance, social
skills, and prosocial behavior. Although a previous meta-analyses explored the
relationship of social aggression and some areas of psychosocial adjustment, it contained
notable limitations.
Furthermore, given the effort to explore and combat social aggression among
girls, it is important to determine the role of gender in the association between social
aggression and different outcomes. Although prior meta-analyses have explored this area,
results are unclear and would benefit from further clarification. In addition, because of
the discrepancy in the results of various studies, it is important to determine if these
inconsistencies are due to another factor, such as age. A resolution of conflicting
-
18
evidence is necessary to further advance the field of social aggression and to apply these
findings to future and current interventions. The more able we are to identify how social
aggression relates to other facts, such as peer relations, social skills, and psychological
distress, the more able we are to create interventions that specifically target these types of
behaviors. Thus, the purpose of this current study is to, (1) examine the relationship
between social aggression and both positive and negative correlations, (2) explore
whether the relation among these variables are moderated by gender, and (3) explore
whether the relation between these variables are moderated by age.
Meta-Analysis
One way to address mixed findings to a research question is to conduct a meta-
analysis (Rosenthal & DiMatteo, 2001). Meta-analysis allows for the combining the
results from multiple, as well as discover moderators and mediators that can examine
mixed findings (Rosenthal & DiMatteo, 2001). This allows for researchers to generate
results that are more generalizable than can be presented in any one study. As Rosenthal
and DiMatteo (2001) note,
Meta-analysis is more than a statistical technique; it is a methodology for
systematically examining a body of research, carefully formulating hypotheses,
conducting an exhaustive search and establishing inclusion/exclusion criteria for
articles, recording and statistically synthesizing and combining data and effect
sizes from these studies, searching for moderator and mediator variables to
explain effects of interest, and reporting results. (p. 62)
-
19
Thus, a meta-analysis was used to better understand the phenomenon of social
aggression and its connection to various correlates by other important moderators, such
as gender and age, which might explain inconsistent findings.
-
20
CHAPTER 3
METHODS
Search Procedure
A comprehensive search for empirical research focusing on social aggression,
broadly defined, was used. Studies to be included in the meta-analysis were selected
using the following search procedures. A broad search was conducted in several relevant
online databases, including: PsycINFO, ERIC (Educational Resources Information
Center), Social Science Citation Index, and Family & Society Studies Worldwide in April
2010. Because of the disagreement among researchers on terms used to describe social
aggression, a wide search with several search terms was carried out to ensure a thorough
collection of relevant articles. Specific keywords included: social aggress* or indirect
aggress* or relational aggress* or covert aggress* or nonphysical aggress* or
bullying. The broad keyword bullying was also used because bullying behaviors were
often not indicated within the keywords of articles. Although related constructs, terms
specific to physical aggression and peer victimization were not included in the search in
order to focus on the specific behaviors of social aggression and the factors associated
with aggressors.
This initial search process resulted in the identification of over 20,000 articles
with potential for inclusion in the meta-analytic database (see Figure 1). These articles
were further narrowed by several criteria within each search database: (1) written in
English, (2) examined a sample of children and/or adolescents, aged 2-17, excluding
college students, (3) be an empirical study, and (4) be within a peer-reviewed journal. In
databases where the age of participants was not provided as a search criteria, additional
-
21
-
22
search terms were added, specifically: child or childhood or children or
adolescent or adolescence or youth. All citations were subsequently placed into
EndNote X3 (a computer citation software program) and analyzed for duplicate
references. This initial search process resulted in the identification of over 4,600 articles
with potential for inclusion in the meta-analytic database.
As shown in Figure 1, abstracts of all collected citations were reviewed to
eliminate articles that clearly did not meet the goals of the present meta-analysis.
Abstracts were excluded if they did not examine social aggression (N = 2488); examined
the general concept of bullying, rather than assessing a particular form of social
bullying (N = 1510)); did not meet the initial criteria (as outlined above, N = 203); did not
provide sufficient information for computing effect size (N = 76); or were primarily
composed of children with mental retardation or pervasive developmental disorders (N =
19) in which social development would presumably be impaired. The remaining 365
studies were included in the initial database. Of those, studies that were relevant to
individual characteristics or peer relations were extracted, yielding a total of 107 studies
included in the present meta-analysis.
A reliability check of included and excluded articles was conducted. A graduate
student in psychology reviewed a random sample of 10% of the initially identified
articles. Studies were randomly chosen by selecting every 10th study in the database (i.e.,
within the database of identified articles, study #1, #10, #20, and so on were selected).
The graduate student independently categorized the chosen studies as included or
excluded. Interrater reliability was assessed by computing the percent of total agreement
between the author and graduate student. The agreement rate for this study was 98%.
-
23
Coding of Studies
Studies which were deemed to be relevant to the present investigation were
collected and coded. Coding involves evaluating each study and collecting the pertinent
information into a coding form. If an article contained multiple samples, each sample was
treated as an independent study sample. Similarly, if an article contained multiple studies,
each study was coded separately. Data from each study was coded on forms created in
Microsoft Access and then automatically transferred to a database created in Microsoft
Excel . This automatic exporting of data minimized opportunity of human error on data
entry.
To ensure reliability of the coding process, another individual was recruited to
code the articles in the database. This individual is a fellow graduate student in
Psychology and was trained on the specific categories that were assessed in the current
research. The authors coding form was compared item-by-item to the research assistants
form for each study. Calculation of the reliability estimate was done via agreement rate
within each study. The inter-rater agreement for this subset of studies was 97%. All
discrepancies found during the reliability check were on items not included in these
analyses (i.e., names of measures, and source of funding).
Each article was coded on five general characteristics:
1) Study characteristics: general article information, such as authors, publication
year, journal, and whether the study was part of a funded project.
2) Sample information: number of participants; location of sample; age; education
level (i.e., preschool, elementary, etc.); gender information; ethnicity.
-
24
3) Social aggression: information regarding what term was used by authors (e.g.,
indirect aggression, relational aggression); information regarding measures;
sources of data (e.g., peer-rating, self-report, teacher-report).
4) Correlates: information regarding the outcomes of each study was identical to that
of measures of social aggression; whether correlate was considered positive (e.g.,
social skills, popularity) or negative (e.g., peer-rejection, psychological distress).
5) Summary statistics: information relating to each of the measures and results,
including correlations and sample size for computing the associated variances.
Each variable was sorted into categories (i.e., individual characteristics, peer
relations) created to address the research questions. Within each category, the correlates
were divided based on their coding as either positive or negative and further separated
into subcategories with the goal of maximizing similarities.
The subcategories for individual characteristics included: externalization,
internalization, positive individual traits, and negative individual traits. Within each
subcategory, several groupings of correlates were created (Please see Appendix for
examples of measures for each subcategory). For externalization, they included: conduct
problems, deception, defiance, delinquency, externalizing behavior, hyperactivity,
impulsivity, inattention, sexual activity, and substance use. Correlates for internalization
were grouped into: anxiety, depression, emotional dysregulation, internalizing symptoms,
and withdrawal. For positive individual traits, groups included: self-efficacy, social
skills, sociability, self-esteem, empathy, cooperative, leadership, language ability, peer-
valued traits (e.g., athleticism, attractiveness, humor, etc.), and positive personality
features (i.e., agreeableness, openness, conscientiousness). Negative individual traits
-
25
were grouped into: anger, dominance, undesirable traits (e.g., bossy, mean, jealousy,
etc.), callous-unemotional traits, narcissism, emotional, and borderline personality
features.
The subcategories for peer relations included positive and negative groupings. For
positive peer relations, groupings of correlates included: peer-acceptance, social support,
social intimacy, peer-liking, popularity, positive social friendships, and prosocial
behavior. Negative peer relations were grouped into: peer-conflict, peer-dislike, peer-
rejection, and peer-victimization. This process of grouping variables was conducted by
the author and research assistant independently. Complete agreement (100%) was
reached for the placement of correlates into categories and subcategories. In addition, the
categories and placement of variables in the respective categories were reviewed by a
professor in Counseling Psychology as an added verification for content validity. Any
discrepancies were discussed until there was complete agreement.
Power
Prospective power analyses are useful to make sure that the power of statistical
tests is sufficient for the effect size to be of practical significance in any given situation
(Hedges & Pigott, 2001). Before a meta-analysis is conducted, a power analysis can
provide the chance of obtaining a statistically significant result provided certain criteria:
expected magnitude of the overall effect size, estimated number of studies included in the
review, and the average sample size within the studies (Hedges & Pigott, 2001).
Given that, at the onset of the analysis process, the factors were undetermined
(i.e., number of studies, typical sample size), the lower acceptable numbers for each
factor were used to determine the range of power. With a small effect size of 0.10, 25
-
26
studies, and a minimum average sample size of 50, the power obtained for finding a
statistically significant overall result is 0.94. Given these factors, analysis of moderators
with as few as five studies would result in a range of power from 0.82 for large between-
group heterogeneity to 0.99 for small between-group heterogeneity. These results indicate
that, even when using the lowest acceptable criteria, the chance of obtaining a statistically
significant result was above 0.80. Thus, the power analysis of the current study adds an
additional confidence yielding statistical inferences with an acceptable power.
Effect Size
The results of each study were standardized into a common metric, known as an
effect size. Effect sizes express the magnitude and direction of the relationship between
two variables (Cooper, Hedges, & Valentine, 2009). The majority of the studies in this
meta-analysis reported correlational coefficients regarding social aggression and the
associated outcomes. Pearsons correlation coefficient (r) is a measure of linear
association, and thus it provides an indication of the degree of the relationship between
two variables. Consequently, Pearsons product moment correlation (r) is the primary
effect size used for the meta-analysis. However, because correlation coefficients are not
normally distributed they were transformed to Fishers z (Borenstein, 2009). Each of the
estimates (e.g., mean, 95% confidence interval) computed based on Fishers z was then
converted back to the r metric using the formula provided by Borenstein (2009). This
conversion allowed the estimates to be interpreted as correlation coefficients. Rosenthal
and DiMatteo (2001) noted several advantages of using the Pearson r in meta-analyses;
including the fact that r is more simply interpreted in terms of practical importance.
-
27
Analysis
As described previously, meta-analyses allows for estimating the overall
relationship using a number of studies examining the same phenomenon (Cooper, 2009)
and further examining the sources of mixed findings, using moderators (Rosenthal &
DiMatteo, 2001). Three methods of meta-analysis are often used in contemporary studies
(Field, 2001): The methods devised by Hedges and Olkin (1985); by Rosenthal and
Rubin (see Rosenthal, 1991), and by Hunter and Schmidt (2004). The current study
employed the methods put forward by Hedges and Olkin (1985) and described by
Cooper, Hedges, and Valentine (2009). According to these authors, first one must
establish which model (i.e., fixed-effect, random-effect, or mixed-effect) is most
appropriate for the data. Essentially, in fixed-effects models (also called the
homogenous case) the effect size is assumed to be from the same underlying
population for all studies included in the meta-analysis (Field, 2001). In contrast,
random-effects models (also called the heterogeneous case) assume that effect sizes
vary from study to study such that the population effect size is likely to be different than
any other study in the meta-analysis (Field, 2001).
To determine which model was more appropriate for the data in the current study,
an overall homogeneity test of effect sizes using the heterogeneity statistics (also known
as Q statistics) was performed using the method described in Konstantopoulos and
Hedges (2009). The heterogeneity Q statistic approximately follows a chi-square
distribution with its associated degree of freedom (k), where k indicates the number of
effect sizes (Konstantopoulos & Hedges, 2009). If the Q statistic was statistically
significant, indicating that the fixed-effects model was not supporting that the effect sizes
-
28
were from the same population, a random-effects model for computing the overall effect
sizes or mixed-effects model with predictors for comparing effect sizes by predictors was
applied. If the Q statistic was not significant, the mean effect size was computed under
the fixed-effect model, in which the effect sizes were weighted by the inverse of their
associated variance.
As discussed in the literature review of this paper, there are several discrepancies
in the relationship between social aggression and various correlates. Moderator analyses
can help clarify these discrepancies. For this meta-analysis, moderators were chosen in
advance based on theory and included: gender and age group (e.g., preschool,
elementary, etc.). Mean correlation by each categorical moderator were compared under
the mixed-effects categorical model.
-
29
CHAPTER 4
RESULTS
Characteristics of Studies
Upon completion of the primary study coding process, it was determined that 108
studies would contribute data for the current meta-analytic database. These 108 studies,
derived from 107 published articles, produced 896 correlation coefficients (r). One article
provided statistical findings from two studies, and thus these findings were treated as two
independent study samples. The majority of studies (s = 76; 70.4%) were from the
United States. International studies were from Asia (s = 6; 5.6%), Australia (s = 2;
1.9%), Europe (s = 11; 10.2%), and North America (s = 12; 11.1%, excluding the United
States). One study used samples from multiple locations (i.e., Russell, Hart, Robinson, &
Olsen, 2003 used samples from the United States and Australia). Articles were published
between 1996 and 2010, with the majority of articles (s = 101; 93.5%) published within
the last ten years. Of the seven articles published in the previous decade, two were
published each year from 1996 to 1998, with one article published in 1999.
The total number of participants across all studies was 52,173. Sample sizes
ranged from 37 to 7,290 (M=487.60, SD = 843.86). The age of participants ranged from
2 to 17 years old, with an average age of 10.2 years (SD = 3.8). The majority of studies (s
= 48; 44.4%) focused on the 10 to 13 year old or middle-school range, with other studies
(s = 60; 55.6%) evenly representing the other age groups (e.g., ages 2-5: s = 22, 20.4%;
ages 6-9: s = 21, 19.4%; ages 14-17: s = 15; 13.9%). Two studies (i.e., Dukes, 2009;
Graves 2007) did not provide participants age information. There were 26,865 (51.5%)
female and 22,285 male participants. Six studies did not provide gender information (i.e.,
-
30
Albrecht, Galambos, & Jansson, 2007; Cillessen, Jiang, West, & Laszkowski, 2005; Ellis
& Zarbatany, 2007; Mayeux & Cillessen, 2008; Ostrov et al., 2009; Yu & Gamble,
2008). Seven studies (i.e., Linder & Gentile, 2009; Loeber et al., 2009; Marsee & Frick,
2007; Mikami, Lee, Hinshaw, & Mullin, 2008; Ohan & Johnston, 2007; Putallaz et al.,
2007; Zalecki & Hinshaw, 2004) used all-female samples, while no studies used all-male
samples.
Publication Bias
One way to assess publication bias is to examine a funnel plot. A funnel plot is a
graphic representation of the relationship between effect sizes and its associated sample
sizes. If there is no publication bias, the plot of effect sizes against sample sizes appears
to be funnel-shaped. In the current study, a scatter plot of 2,078 correlation coefficients
against sample sizes appeared asymmetrical and did not resemble a funnel shape (see
Figure 2), suggesting possible publication bias. In addition to the visual inspection of the
funnel plot, a statistical test commonly used to detect publication bias was conducted.
Eggers regression test for funnel plot asymmetry (Sutton, 2009) was found to be
statistically significant for correlation coefficients (t (2076) = 18.84, p < .01), thus
indicating the presence of potential publication bias. Such results were due to the
inclusion of studies published in peer-reviewed journals, which would limit the
generalizability of findings in the current meta-analysis.
Overall Model
The aim of this meta-analysis was to examine the relationship between social
aggression and various correlates. After coding all of the articles, correlates were grouped
-
31
-
32
into two general categories: individual characteristics and peer relations (as seen in
Figure 3). Under the random-effects model, each category was found to be significantly
correlated with social aggression: individual characteristics (r = 0.25, p < .01), peer relations (r = 0.10, p < .01; see Table 1). Post-hoc comaprisons using Bonferonni adjustments indicated that individual characteristics had a statistically higher mean
correlation with social aggression than peer relations (z = 6.71, p < .01). Several
moderators were examined to determine their role in each relationship and are discussed
below.
Gender differences. Table 2 shows results from the moderator analysis using
gender. The relationship between individual characteristics and social aggression
significantly differed by gender; however, not for peer relations. Under the mixed-effects
model, the relationship between individual characteristics and social aggression, was
significant for female-only, male-only, and mixed-gender groups. Similarly, the
relationship between peer relations and social aggression was significant and positive for
female-only, male-only, and mixed-gender groups. Post-hoc comparisons using
Bonferonni adjustments indicated that the relation of social aggression to individual
characteristics was statistically higher for female-only than for male-only (z = 2.40, p <
.01).
Age differences. Table 3 shows results from the moderator analysis by age.
Mixed-effects models showed that the overall relationship between social aggression and
individual characteristics and peer relations differed by age. The relationship between
social aggression and individual characteristics was significant and positive for
preschool-, elementary-, middle-school-, and high-school-aged samples. Simiarly, the
-
33
-
34
-
35
-
36
relationship between peer relations and social aggression was significant and positive for
preschool-, middle-school-, and high-school-aged groups; however, not significant for
elementary-aged samples. Post-hoc comparisons using Bonferonni adjustments indicated
a statistically higher difference in the relation of social aggression to individual
characteristics for high-school than for middle-school-aged samples (z = 2.83, p < .01).
Post-hoc comparisons also indicated a statistically higher difference in the relation of
social aggression to peer relations for preschoolers as compared to elementary- (z = 3.77,
p < .01) and high-school-aged samples (z = 2.60, p < .01); as well as for middle-school in
comaprison to elementary-school children (z = 2.50, p < .01).
Relationship Between Individual Characteristics and Social Aggression
A more specific goal of the present study was to determine whether positive or
negative correlates were related to social aggression. For each general category,
correlates were separated into those respective groups and analyzed independently.
Again, gender and age moderators were examined to determine their roles in each
relationship.
The general category of individual characteristics was divided into four
subcategories (i.e., externalization, internalization, positive individual traits, negative
individual traits) and then estimated separately for its relationship with social aggression.
Overall, the four types of individual characteristics showed significant and positive
relationships with social aggression (see Table 4). Post-hoc comparisons using
Bonferonni adjustments indicated statistically different relationships between social
aggression and externalization as compared to internalization (z = 3.88, p < .01) and
positive individual traits (z = 5.89, p < .01). Additionally, post-hoc comparisons showed
-
37
-
38
statistically different relationships between social aggression and negative individual
traits as compared to internalization (z = 3.33, p < .01) and positive individual traits (z =
5.42, p < .01). Finally, post-hoc comparisons indicated statistically different relationships
between social aggression and internalization as compared to positive individual traits (z
= 3.05, p < .01).
Correlates categorized into the externalization of individual characteristics were
each estimated separately for their relationship with social aggression. All but one of the
subcategories (i.e., sexual activity) were significantly and positively associated with
social aggression. Post-hoc comparisons using Bonferonni adjustments indicated
statistically lower relationships between social relations and conduct problems as
compared to defiance (z = 2.98, p < .01) and inattention (z = 3.00, p < .01). In addition,
post-hoc comparisons showed statistically lower relationships between social relations
and substance use as compared to defiance (z = 3.64, p < .01), inattention (z = 5.59, p <
.01), delinquency (z = 2.63, p < .01), externalization (z = 1.20, p < .01), and hyperactivity
(z = 2.19, p < .01).
Variables grouped into the internalization of individual characteristics were also
estimated separately for their association with social aggression. The overall correlation
indicated positive and significant relationships between emotional dysregulation and
social aggression, as well as between internalizing symptoms and social aggression. Post-
hoc using Bonferonni adjustments indicated statistically different relationships between
social aggression and internalizing symptoms as compared to emotional dysregulation (z
= 3.05, p < .01).
-
39
Correlates classified into the positive individual traits were estimated separately
for their relationship with social aggression. The overall correlation indicated positive
and significant relationships for social skills, sociability, leadership, language ability, and
peer-valued traits with social aggression. Several correlates indicated negative
relationships with social aggression, such as self-efficacy, empathy, cooperative, and
positive personality traits, however these correlations were not statistically significant.
Variables grouped into the undesirable individual traits were also estimated
separately for their association with social agression. The overall correlation indicated
positive and significant relationships for anger, dominance, callous-unemotional traits,
narcissism, and borderline personality traits. Post-hoc comparisons using Bonferonni
adjustments indicated statistically lower relation between social aggression and
borderline traits as compared to anger (z = 3.40, p < .01), dominance (z = 5.55, p < .01),
callous-unemotional traits (z = 2.69, p < .01), and narcissism (z = 3.80, p < .01).
Gender differences. Table 5 shows results from the moderator analyses by
gender. Mixed-effects models showed that the relationship between social aggression and
externalization, internalization, positive individual traits, and negative individual traits
significantly differed by gender. The relationship between externalization and social
aggression was significant and positive for female-only and mixed-gender groups;
however, not significant for male-only groups. The association between internalization
and social aggression was significant for female-only, male-only, and mixed-gender
groups. Conversely, the correlation between positive individual traits and social
aggression was significant for mixed-gender groups; however, not significant for female-
only or male-only groups. Finally, the relationship between negative individual traits and
-
40
-
41
social aggression was significant for female-only, male-only, and mixed-gender groups.
Post-hoc comparisons using Bonferonni adjustments indicated that the relations of social
aggression to externalization was statistically higher for female-only than for mixed-
gender (z = 2.57, p < .01).
Of the effect sizes related to externalization, three were able to be analyzed for
gender differences: defiance, externalizing behavior, and hyperactivity. Defiance was
positively and significantly related to social aggression for female-only groups and
mixed-gender groups. Female-only groups showed a significant relationship between
externalizing behavior and social aggression, while male-only groups did not. The
association between hyperactivity and social aggression was significant for female-only,
male-only, and mixed-gender groups. Post-hoc comparisons on the relation of defiance to
social aggression indicated statistically higher relationships for female-only than for
mixed-gender groups (z = 2.72, p < .01). Similarly, post-hoc comparisons on the
association between hyperactivity and social aggression indicated statistically different
relationships for female-only than for male-only (z = 5.53, p < .01) and mixed-gender
groups (z = 4.02, p < .01).
All of the effect sizes related to internalization were able to be analyzed for
gender differences. The association between depression and social aggression was
significant for mixed-gender groups; however, not significant for female-only or male-
only groups. Emotional dysregulation was positively and significantly related to social
aggression for female-only groups and mixed-gender groups. Similarly, internalizing
symptoms was significantly associated to social aggression in female-only, male-only,
-
42
and mixed gender groups. Neither anxiety nor withdrawal showed significant
associations for female-only or mixed-gender groups.
Of the effect sizes related to positive individual traits, two were able to be
analyzed for gender differences: language ability and sociability. Language ability was
positively and significantly associated with social aggression for female-only and mixed-
gender groups; however, was negatively and significantly related male-only groups.
Sociability was not significantly related to social aggression for either female-only or
mixed gender groups. Post-hoc comparisons on the association between language ability
and social aggression indicated statistically different relationships for female-only as
compared to male-only groups (z = 3.39, p < .01).
Three effect sizes related to negative individual traits were able to be analyzed for
gender differences: dominance, negative traits, and callous-unemotional traits. Both
dominance and callous-unemotional traits were positively and significantly associated
with social aggression for female-only and mixed-gender groups. Negative traits, such as
being mean or jealous were significantly associated with social aggression for
mixed-gender groups; however, not significant for female-only groups. Post-hoc
comparisons indicated that the relation of social aggression to callous-unemotional traits
was statistically higher for female-only than for mixed-gender groups (z = 5.89, p < .01).
Age differences. Table 6 shows results from the moderator analyses by age.
Mixed-effects models showed that the relationship between social aggression and
externalization, internalization, positive individual traits, and negative individual traits
significantly differed by age of sample. The relationship between externalization and
social aggression was significant for preschool-, elementary-, middle-school-, and high-
-
43
-
44
school-aged groups. The association between internalization and social aggression was
significant for middle-school- and high-school-aged children; however, this relationship
was not significant for preschoolers or elementary-school children. Conversely, the
relationship between positive individual traits and social aggression was significant for
preschool- and elementary-school-aged samples; however, not significant for middle-
school or high-school groups. Finally, the relationship between negative individual traits
and social aggression was significant for preschool-, elementary-, middle-school-, and
high-school-aged groups. Post-hoc comparisons using Bonferonni adjustments indicated
that the relationships of social aggression to externalization was statistically higher for
elementary-school than for high-school-aged children (z = 2.74, p < .01).
Several correlations related to externalization were able to be analyzed for age
differences. Delinquency was significantly related to social aggression for elementary-,
middle-school-, and high-school-aged children. Conduct problems was significantly
associated with social aggression for preschool-, middle-school-, and high-school-aged
groups; however, was not significant for elementary-school children. Both deception and
impulsivity were significantly correlated with social aggression for preschoolers and
middle-school-aged children. Although both externalizing behaviors and inattention were
significantly related to social aggression for middle-school children, only inattention was
also significant for elementary-school children. Finally, substance use was significantly
associated with social aggression for high-school groups, but not for middle-school
groups. Post-hoc comparisons on the association of conduct problems and social
aggression indicated a statistically higher relation for preschoolers as compared to
middle-school-aged children (z = 2.56, p < .01).
-
45
Two correlates related to internalization were able to be analyzed for age
differences. Internalizing symptoms was significantly related with social aggression for
both elementary- and middle-school-aged samples. Depression was significantly
associated with social aggression for high-school gorups; however, this relationship was
not significant for preschoolers or elementary-school aged children. Post-hoc
comparisons on the relation of social aggression to internalizing symptoms indicated
statistically higher relation for middle-school- than for elementary-school children (z =
2.06, p < .01).
Two effect sizes related to positive individual traits were able to be analyzed for
age differences. Language ability was positively and significantly related to social
aggression for preschoolers, but negatively and significantly associated for middle-
school-aged children. Social skills was not significantly related to social aggression for
preschool- or middle-school-aged children.
Four effect sizes related to negative individual traits were able to be analyzed for
age differences. Dominance was significantly associated with social aggression for
preschoolers and high-school children. Both anger and narcissism were significantly
associated with social aggression for middle-school- and high-school-aged children.
Callous-unemotional traits were significantly correlated with social aggression for
elementary-, middle-school-, and high-school-aged groups. Post-hoc comparisons on the
association between social aggression and callous-unemotional traits indicated
statistically lower relation for middle-school children than for elementary- (z = 3.71, p <
.01) and high-school-aged groups (z = 4.02, p < .01).
-
46
Relationship Between Peer Relations and Social Aggression
Peer relations was divided into positive and negative subcategories and estimated
separately for its association with social aggression (see Table 7). While both positive
and negative peer relations were both positively associated with social aggression, only
negative peer relations was significantly related. Correlates categorized into positive peer
relations were each estimated separately for their association with social aggression. The
overall correlation coefficients indicated positive and significant relationships for
intimacy and popularity, as well as negative and significant relationships for peer-liking
and positive social relations. Prosocial behavior, social acceptance, and social support did
not show statistically significant relationships with social aggression. Post-hoc
comparisons using Bonferonni adjustments indicated statistically different associations
between social aggression and popularity as compared to peer-dislike (z = 3.88, p < .01)
and negative friendships (z = 6.01, p < .01). Correlates grouped into negative peer
relations were also estimated separately for their association with social aggression. All
of the subcategories (i.e., social conflict, peer dislike, peer rejection, peer victimization)
were positively and significantly associated with social aggression. Post-hoc comparisons
using Bonferonni adjustments indicated statistically higher relationships between social
aggression and peer-victimization as compared to peer-rejection (z = 2.50, p < .01).
Gender differences. Table 8 shows results from moderator analyses by gender.
Mixed-effects models showed that the relationship between social aggression and both
positive and negative peer relations significantly differed by gender. The relationship
between positive peer relations and social aggression was significant for mixed-gender
-
47
-
48
-
49
groups; however, was not significant for female-only or male-only groups. Conversely,
the association between negative peer relations and social aggression was significant and
positive for female-only, male-only, and mixed-gender groups. Post-hoc comparisons
using Boneronni adjustments indicated that the relationships of social aggression to
negative peer relations was statistically higher for female-only than for mixed-gender
groups (z = 2.65, p < .01).
Of the effect sizes related to positive peer relations, all but one (i.e., intimacy)
were able to be analyzed for gender differences. Popularity was significantly associated
with social aggression for female-only, male-only, and mixed-gender groups. Peer-liking
was significantly related to social aggression in mixed-gender groups, however not for
female-only or male-only groups. Positive friendships was significantly associated with
social aggression for female-only groups; however, was not significantly related for
male-only or mixed-gender groups. Finally, neither social acceptance, social support, or
prosocial behavior was significantly associated with social aggression for female-only,
male-only, or mixed-gender groups.
All of the effect sizes related to negative peer relations were able to be analyzed
for gender differences. Both social conflict and victimization were significantly related to
social aggression for female-only, male-only, and mixed-gender groups. Peer-dislike and
peer-rejection were significantly correlated with social aggression for female-only and
male-only groups, however not for mixed-gender groups.
Age differences. Table 9 shows results from the moderator analyses by age.
Mixed-effects models showed that the relationship between social aggression and both
positive and negative peer relations significantly differed by age of sample. The
-
50
-
51
relationship between positive peer relations and social aggression was significant and
positive for middle-school- and high-school-aged groups, as well as significant and
negative for elementary-school children. However, this association was not significant for
preschoolers. The association between negative peer relations and social aggression was
significant for preschool-, elementary-, and middle-school-aged children, however not
significant for high-school groups. Several effect sizes related to positive peer relations
were able to be analyzed for age differences. Popularity was significantly related with
social aggression for middle-school and high-school groups. Social acceptance and peer-
liking were significantly associated with social aggression for elementary- and high-
school-aged children, however not for preschoolers or middle-school children. Positive
friendships was significantly related to social aggression for preschool and elementary-
school groups, but not for middle-school- or high-school-aged children. Social support
was significantly correlated with social aggression for high-school groups, but not for
elementary-school aged children. Finally, prosocial behavior was not significantly
associated with social aggression for any age group. Post-hoc comparisons using
Bonferonni adjustments on the relation of social aggression and popularity indicated a
statistically higher relation for high-school than for middle-school-aged children (z =
3.61, p < .01).
All of the effect sizes related to negative peer relations were able to be analyzed
for age differences. Peer-dislike was significantly associated with social aggression for
elementary- and high-school-aged children. Both peer-rejection and peer-victimization
were significantly related with social aggression for all age groups. Finally, social
-
52
conflict was significantly correlated with social aggression for elementary and middle-
school groups, but not for high-school-aged children.
-
53
CHAPTER 5
DISCUSSION
This meta-analysis was inspired by the hypothesis that social aggression can be
associated with both positive and negative psychosocial correlates. To explore this
premise, correlates were separated into positive and negative groups and their association
with social aggression was analyzed. In addition, age and gender were considered as
moderators to this relationship. Below, the findings of this meta-analysis are reported;
theoretical and practical implications are discussed; shortcomings are identified; and
directions for future research are suggested.
As described in the literature review of this paper, there is evidence that social
aggression is associated with both positive and negative correlates. In order to examine
this discrepancy, correlates of each category were grouped into positive (i.e., positive
individual traits, positive peer relations) and negative characteristics (i.e., externalization,
internalization, negative individual traits, and negative peer relations). In general, results
suggest that social aggression is more greatly associated with negative correlates.
However, evidence of significant relationships for some positive socially-oriented
variables with social aggression was also identified.
Research linking social aggression to indices of maladjustment has made an
impressive contribution to the scientific literature. Experiences of social aggression in
children and adolescents have been implicated in concurrent social and emotional
difficulties for both boys and girls by previous studies (e.g., Crick et al., 2002a; Crick &
Bigbee, 1998; Prinstein et al., 2001). Similar findings were noted in the current meta-
analysis. Within the category of individual characteristics, positive significant
-
54
relationships with social aggression were found for externalization, internalization, and
negative individual traits. Although significant, the association between positive
individual traits and social aggression was trivial in magnitude. Further, analyses
indicated that positive individual traits had a statistically lower relationship with social
aggression than externalization, internalization, and negative individual traits. That is, for
the correlates grouped into individual characteristics, those that were deemed negative
were more greatly associated with social aggression than those that were considered
positive.
In terms of maladjustment, findings indicate significant relationships between
social aggression and both externalizing and internalizing difficulties. Current results
reveal positive and significant associations with social aggression for several aspects of
externalization, including: deception, defiance, delinquency, externalizing behaviors,
hyperactivity, impulsivity, inattention, conduct problems, and substance use. These
unsurprising results are akin to previous research findings. For example, social
aggression has bee