/

University of Southern California Dissertations and Theses

/

Examining the longitudinal relationships between community violence exposure and aggressive behavior among a sample of maltreated and non-maltreated adolescents

(USC Thesis Other)

Examining the longitudinal relationships between community violence exposure and aggressive behavior among a sample of maltreated and non-maltreated adolescents

PDF

Transcript (if available)

Content

EXAMINING THE LONGITUDINAL RELATIONSHIPS BETWEEN COMMUNITY
VIOLENCE EXPOSURE AND AGGRESSIVE BEHAVIOR AMONG A SAMPLE OF
MALTREATED AND NON-MALTREATED ADOLESCENTS

by

Kristopher Ian Stevens

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(PSYCHOLOGY)

August 2011

Copyright 2011 Kristopher Ian Stevens
ii

TABLE OF CONTENTS
List of Tables iii

List of Figures iv

Abstract v

Chapter One: Background and Significance 1
Societal problem 1
Community violence exposure and aggressive behavior 2
Theoretical models 5
Methodological problems 7
Focus of the current study 17
Specific Aims 18

Chapter Two: Research Design and Methods 20
Participants 20
Procedures 22
Measures 28

Chapter Three: Results 33
Preliminary analyses 33
Substantive analyses 55

Chapter Four: Discussion 69

References 85

Appendices
Appendix A: Tables 92
Appendix B: Community Violence Inventory 99
Appendix C: Youth Self-report Questionnaire 100
Appendix D: Child Behavior Checklist 102
Appendix E: Teacher Report Scale 103

iii

LIST OF TABLES

Table 1. Sample demographics 23
Table 2. Attrition across time points 25
Table 3. Means and standard deviations for the variables of interest 36
Table 4. Correlations between the variables of interest 37
Table 5. Percentage of sample exposed to instances of community violence 38

Table 6. Percentage of sample with no exposure at T2 and T3 versus those
exposed at both time points 39

Table 7. Cross-lagged model addressing Specific Aim 1 58
Table 8. Cross-lagged model addressing Specific Aim 2 61
Table 9. Cross-lagged model addressing Specific Aim 3 64
Table 10. Cross-lagged model addressing Specific Aim 4 67

Table 11. Descriptive statistics for the variables of interest at T2 and T3
for the full sample 92

Table 12. Descriptive statistics for the variables of interest at T2 and T3
for females and males 93

Table 13. Descriptive statistics for the variables of interest at T2 and T3
for the maltreated and non-maltreated groups 94

Table 14. Descriptive statistics for the variables of interest at T2 and T3
for the physically and non-physically maltreated groups 95

Table 15. Correlations between the variables of interest at T2 and T3
for females and males 96

Table 16. Correlations between the variables of interest at T2 and T3
for females and males 97

Table 17. Correlations between the variables of interest at T2 and T3
for females and males 98

iv

LIST OF FIGURES

Figure 1. CFA of community violence data at T2 41

Figure 1. CFA of community violence data at T3 41

Figure 3. CFA of aggressive behavior data at T2 48

Figure 4. CFA of aggressive behavior data at T3 48

Figure 5. Second order CFA for aggressive behavior data at T2 49

Figure 6. Second order CFA for aggressive behavior data at T3 49

Figure 7. Latent variable for aggressive behavior at T2 50

Figure 8. Latent variable for aggression at T3 50

Figure 9. Fully cross-lagged model examining Specific Aim 1 57

Figure 10. Fully cross-lagged model examining Specific Aim 2 61

Figure 11. Fully cross-lagged model examining Specific Aim 3 65

Figure 12. Fully cross-lagged model examining Specific Aim 4 68

v

ABSTRACT
The current investigation sought to elucidate the longitudinal relationships
between community violence exposure and aggressive behavior in a sample of 389
maltreated and non-maltreated young adolescent males and females. Of primary interest
was examining the effect of adolescents’ early community violence exposure and
aggressive behavior on their later community violence exposure and aggressive behavior
as well as examining how these relationships differ by gender and maltreatment history.
The present study was conducted in the context of a larger longitudinal study examining
the impacts of child maltreatment on adolescent development. The data used in the
current investigation were drawn primarily from the second (T2) and third (T3) data
collection waves of the parent study.
To examine the longitudinal relationships, latent variable cross-lagged panel
analysis methodology was employed. Results indicated that for the full sample there was
no association between early community violence exposure and later aggressive
behavior. However, early aggressive behavior contributed to later community violence
exposure. Examinations of gender differences revealed that for males, later aggressive
behavior was influenced by early exposure. There was no association for females. For
both males and females, early aggressive behavior contributed to later community
violence exposure. Examination of differences based on maltreatment status indicated
that for non-maltreated adolescents, later aggressive behavior was influenced by early
community violence exposure. There was no association for maltreated adolescents.
Early aggressive behavior placed both maltreated and non-maltreated adolescents at risk
vi

for later exposure to community violence. Examination of specific maltreatment type
indicated that early community violence exposure predicted later aggressive behavior for
the non-maltreated and non-physically maltreated groups. There was no association for
the physically maltreated group. Early aggressive behavior placed both non-maltreated
and non-physically maltreated adolescents at risk for exposure to violence later on; there
was no association for the physically maltreated group.
These results contribute to the field in several significant ways. The results lend
support to the small but growing body of literature demonstrating that early aggression
places children and adolescents at risk for later victimization. Although the relationship is
much more complicated than that, the current study provides evidences that the Exposure
→ Behavior model that has dominated the field is incomplete. Results also shed light on
group differences in the longitudinal associations between community violence exposure
and aggressive behavior. The results of this study have important implications for both
future research and clinical practice.
1

CHAPTER ONE: BACKGROUND AND SIGNIFICANCE
Community Violence: A Societal Problem
Among Western industrialized nations, the United States is one of the most
violent (Scarpa, 2003). In fact, out of 23 developed countries, the United States ranks first
in homicides at a rate that is nearly four times greater than the next country (Fingerhut &
Kleinman, 1990). Most alarming is that the portion of the population that is exposed at
the highest rates also happens to be its most vulnerable: children and adolescents. Several
studies have indicated high prevalence rates. According to the National Crime
Victimization Survey (2009), the 2008 prevalence rate for violent victimization (i.e.,
experiencing homicide, rape, robbery, and/or assault) was 42 per 1,000 for individuals 12
to 19 years of age. Among children living in inner-city Chicago neighborhoods, 65%
reported witnessing a serious assault and 33% reported witnessing a murder (Jenkins &
Bell, 1991). Another study found that 39% of adolescents between the ages of 12 and 17
years old reported witnessing someone being shot, stabbed, sexually assaulted, mugged,
or threatened with a weapon, 13.8% witnessed an assault with a weapon and 20.9%
witnessed an assault without a weapon in the past year (Thompson, Proctor, Weisbart,
Lewis, English, Hussey, & Runyan, 2007). These rates of community violence exposure
have led to comparisons of inner-city America to war zones in distant parts of the world
(Guerra, Husemann, & Spindler, 2003).
2

Since the Centers for Disease Control identified community violence as one of the
major health problems affecting children in the United States in the early 1990s
(American Psychological Society, 1997), there has been a virtual explosion in the amount
of research conducted to determine the extent of the issue and the effects such exposure
could have on development. This boom is illustrated by the fact that before 1993 there
were only nineteen published studies – by 2010, there were five hundred and forty-seven.
This body of research has consistently found links between community violence exposure
and several maladaptive developmental outcomes (Lynch, 2003; Margolin & Gordis,
2000; Overstreet, 2000). One of the strongest links has been between community
violence exposure and increased risk for aggressive behavior problems (Gorman-Smith &
Tolan, 1998; Margolin & Gordis, 2000; Overstreet, 2000).
Community Violence Exposure and Aggressive Behavior
Empirical studies. Several studies have examined the relationship between
community violence exposure and aggressive behavior in adolescents residing in urban
communities in the United States. Durant, Cadenhead, Pendegrast, Slavens, and Linder
(1994) found that among urban black adolescents, exposure to violence and victimization
were related to self-reported aggressive behavior. In a large cross-sectional sample of
sixth, eighth, and tenth graders (Schwab-Stone, Ayers, Kasprow, Voyce, Barone, Shriver,
& Weissberg , 1995), exposure to community violence was significantly and positively
associated with participation in aggressive and antisocial behavior. This relationship held
up even after controlling for demographic variables (i.e., sex, grade, socioeconomic
status, and ethnicity). Farrell and Bruce (1997) examined the influence of exposure to
3

community violence on violent behavior among 436 sixth graders across three time
points in an eight-month period. An examination of the longitudinal data revealed a
significant positive association between exposure and self-reported violent behavior.
Additionally, Gorman-Smith and Tolan (1998) examined the influence of exposure on
aggressive behavior in a community sample of 245 fifth and seventh grade African-
American and Latino boys from the inner city. Exposure was assessed using structured
interviews with the children, while measures of aggressive behavior were obtained from
parent, teacher, and child interviews. Results indicated that exposure was associated with
increases in aggressive behavior during adolescence even after controlling for early
aggressive behavior. Additionally, Mrug and Windle (2009) found that early community
violence exposure is associated with later aggressive behavior in a sample of inner-city
adolescents.
Literature reviews. Within the past decade, several qualitative literature reviews
have been put forth to summarize the research on the link between community violence
exposure and maladaptive developmental outcomes (Calvert, 1999; Lynch, 2003;
Margolin & Gordis, 2000; Overstreet, 2000). All of these reviews have come to the same
conclusions that community violence exposure is positively associated with aggressive
behavior. While all of the reviews established that there is a relationship, they have not
been able to conclude anything about the actual size of the relationship, which is
important information for guiding the conceptualization and implementation of research
and intervention. To address this, the present author conducted a quantitative review of
the extant literature between the years of 1993 and 2006 (Stevens, 2011). The review
4

included 56 independent samples which incorporated a total of 27,302 children and
adolescents. Results of the meta-analysis indicated that the effect size among the 56
samples was actually quite large (r = .40; range r = .0 - .82). This effect size indicates
that community violence exposure accounted for 16% of the variance in aggressive
behavior problems. This provides strong empirical support for conceptualizing
community violence exposure as one of the leading causes of aggression. However, 84%
of the variance went unexplained. Thus, it appears that there are other factors at play in
the relationship between community violence exposure and aggressive behavior
problems. As discussed below in detail, the literature on community violence exposure is
marred by several methodological issues that could impact the validity of these findings.
Group differences. Only a few studies have been designed with the capacity to
examine gender differences in the association between community violence exposure and
aggressive behavior (Boyd, Cooley, Lambert, & Ialongo, 2003; Farrell & Bruce, 1997;
Mrug & Windle, 2009; O’Keefe, 1997). The results of these studies have been
inconsistent. Although most studies have not found differences between genders in the
associations (Boyd et al., 2003), those that have found differences are often contrary to
one another. For instance, Farrell and Bruce (1997) found that exposure to community
violence was related to changes in the frequency of aggressive behavior only for girls.
That is, females displayed more aggressive behaviors as a result of community violence
exposure. O’Keefe (1997), on the other hand, found that exposure to community violence
significantly predicted aggressive behavior only for boys. Mrug and Windle (2009) also
found that boys were more likely than girls to become more aggressive as a result of
5

community violence exposure. For the most part, studies that have demonstrated gender
differences have found that the associations are greater for boys than for girls. This
potential for gender differences in the relationships between community violence
exposure and aggressive behavior, coupled with the scarcity of research considering
gender, highlights an important gap in the extant literature that warrants attention.
Furthermore, there is a dearth of knowledge about how these relationships might differ
across various ethnic groups. This is important information to have because ethnic
minorities tend to disproportionately reside in neighborhoods that have high rates of
crime and violence.
Theoretical Models for Understanding the Relationship
Over the past two decades, psychologists have devoted their attention primarily to
examining the link between community violence exposure and subsequent aggressive
behavior. In doing so, they have been working from an Exposure → Behavior
perspective. In other words, community violence exposure leads directly to aggressive
behavior. One theory that has been used by community violence researchers to explain
how aggressive behavior is influenced by violence exposure is the Social Learning
Theory (Bandura, 1986).
Social Learning Theory posits that people learn behaviors by observing others’
behavior. If people observe desired outcomes in the observed behaviors, they are more
likely to model, imitate, and adopt the behavior themselves. Thus, this framework
suggests that adolescents who are aggressive have acquired behavioral tendencies
through vicarious and observational learning (Bandura, 1986). In a high-risk community,
6

many things come together to facilitate the learning of aggression (Huesmann, 1988;
Proctor, 2006). Adolescents may come to see the use of aggression as acceptable and a
normal aspect of relationships with others (Bandura, 1986). In fact, children and
adolescents who are exposed to community violence when they are with their caregivers
or peers may look to them to see how to respond and may pick up that aggressive and
violent behavior is acceptable and the norm. While this framework provides great utility
in explaining how community violence influences subsequent aggressive behavior,
methodological issues in the field challenge its validity. These issues will be discussed
below in greater detail.
Another framework that has been utilized by researchers has been the Ecological-
Transactional Model. Community violence researchers have utilized the ecological-
transactional model which integrates Bronfenbrenner’s (1977, 1986) ecological
perspective on human development and the developmental psychopathology model. This
perspective views individuals as developing in a set of imbedded contexts. In an
ecological model, risk and protective factors are hypothesized to be present within and
transact across several contexts, such as the individual, family, community, and the
broader socio-cultural environment (Belsky, 1980; Belsky, 1984; Bronfenbrenner, 1977;
Bronfenbrenner, 1989). That is, the community context can influence individual
development on several levels. An important aspect of the ecological framework is that it
recognizes bidirectional effects and person-context interactions (Bronfenbrenner, 1989;
Cicchetti & Lynch, 1993). In other words, not only do contexts influence individuals, but
individuals also influence and even construct the contexts in which they interact (Aber,
7

Gephart, Brooks-Gunn, Connell, & Spencer, 1997). This model provides a useful
heuristic for children living in violent communities. Community violence is an enduring
distal stressor in the exosystem that may impact the children through the family context
(Bronfenbrenner, 1989). For those that are directly victimized, it may be more proximal.
Methodological Issues with the Community Violence Exposure Literature
As stated previously, there has been an explosion in the number of studies
examining the impacts of community violence. While this increased attention is
warranted, it must be cautioned that quantity of research does not often give way to
quality. Several reasons have been given as to why the literature on community violence
has lacked in quality. These reasons highlight serious methodological concerns that draw
into question the reliability and validity of study findings and limit the capacity to draw
comparisons across studies. Understanding how these issues impact the results of studies
and how they can be better accounted for in future research is an important step in
moving the field forward.
Operationalization. One of the most important issues in the study of community
violence is the issue of definition. There has been a lack of consensus across studies as to
how community violence is operationalized. What “community” means and what is
meant by “violence” often differ from study to study (Trickett, Duran, & Horn, 2003).
For the most part, community has been defined as the neighborhood surrounding where
the individual lives. But some researchers have also defined community as including the
home and the school setting. Defining community to include other settings such as the
home becomes problematic because it may lead children to report on acts of violence that
8

would normally be reported as other forms of violence, such as domestic violence or
child maltreatment. While some studies operationalize violence as witnessing or being a
victim of recurrent community violence (i.e., being chased, threatened, mugged, shot at,
killed, etc.), others operationalize violence as either witnessing or being a victim. Still
others might include in the operationalization hearing about or viewing violent acts in the
media. Operationalization of key study variables is an important step, for it determines
how the phenomenon of interest is measured and how it is entered into analyses.
Content. Another issue is the lack of consensus across studies in which aspects of
community violence are observed. Some studies have looked exclusively at violence
exposure (which includes aspects of witnessing and being a victim but does not
distinguish between them), others have looked at either witnessing violent events or
being directly victimized, and still others look at the individual contributions of both
witnessing and victimization. This issue is important for two main reasons. First, if
researchers are not studying the same aspects of community violence exposure, it is
difficult to compare results across studies. The second reason is that it is difficult to
determine the differential effects of the various aspects of community violence
(witnessing and victimization). While many theorists and researchers argue that
witnessing and victimization are distinct aspects of violence exposure, the evidence has
been mixed as to whether they lead to differential outcomes (Margolin & Gordis, 2000;
Scarpa, 2003). This highlights that more work needs to be done to determine whether
these are distinct aspects of community violence exposure.
9

Measurement. Another methodological issue is that there is a lack of consensus in
how community violence exposure is measured across studies (Trickett, Duran, & Horn,
2003). Although several well-known measures are used within the community violence
literature, there is no “standard” tool by which exposure is measured. In fact, most of the
measures used by researchers are either created for their study or adapted from known
measures to meet the needs of the researcher and their study. These researchers typically
select which items to include, discard others, and change the scaling of the response
options. All of this is usually done without examining the psychometric properties of the
assessment, and the effects of removing particular items from the measure go unknown.
Thus, it is difficult to know whether researchers are actually measuring the same thing,
let alone community violence. The extant literature on community violence is lacking in
studies that have employed measurement models to assess the psychometric properties of
the assessments they used to examine exposure. Without examining the properties of the
measures, how do researchers know what they are measuring or how well they are
measuring it? An examination of the literature reveals that few studies report the
reliability of the measure used and even fewer examine validity (Trickett, Horn, & Duran,
2003).
Use of self-report. The community violence literature has relied a great deal on
the use of self-report measures not only to assess exposure, but also to assess the outcome
of interest. Several researchers have warned that this becomes problematic due to the
issue of shared method variance. When children report on both predictor and outcome
variables, the relationships tend to be stronger than when multiple-informants are used.
10

When the same measure is used to assess the independent and the dependent variables,
the resulting correlation between the variables could be explained by the fact that
measurement variance is shared between the two variables (Olweus, 1993b). The
previously discussed meta-analysis indicated that children who reported on both
independent and dependent variables had higher effect sizes than when multiple-
informant reports were used (r = .45 and r = .26, respectively), the difference between
these effect sizes is possibly the result of inflation due to shared method variance. This
potential for shared variance highlights the need in the field to move toward a multiple-
informant and multiple-method strategy of assessing exposure and/or other variables of
interest. An additional problem is that self-report measures of community violence might
tap into cognitive and perceptual biases rather than their actual experiences. In other
words, reporters might be responding to a schema that their neighborhood is a dangerous
place rather than the reality of the experience.
Study design. For the most part, researchers have looked at the effects of exposure
on the outcomes of interest using cross-sectional designs rather than longitudinal designs.
In fact, the previously mentioned meta-analysis indicated, that between 1993 and 2006,
only 20% of studies conducted employed a longitudinal design. Reliance on cross-
sectional designs makes it difficult to examine how exposure to community violence may
have different impacts at various developmental stages and would also make it difficult to
examine immediate versus delayed effects over time (Margolin & Gordis, 2000).
11

Several problems arise when making conclusions about studies that employ such
designs. First, it could be that there is a “third” unmeasured variable that accounts for the
correlation between community violence and aggressive behavior. That is, community
violence exposure and aggressive behavior may be the direct effects of a third variable, or
it could be that aggressive behavior is directly caused by a third variable and only
indirectly by community violence exposure.
Second, while community violence exposure and aggressive behavior are
correlated, the claim that community violence exposure causes aggressive behavior
cannot be firmly established nor can the actual direction of the relationship. That is,
rather than community violence exposure influencing aggressive behavior, it could very
well be that aggressive behavior leads to community violence exposure. Or it could be
that the relationship is bi-directional and both exposure and aggression exert influence on
one another. To examine the direction of influence, the relationships need to be examined
over time. Unfortunately, since the field has been driven by an Exposure → Behavior
model, the designs of longitudinal studies have not often considered the possibility of bi-
directionality or a Behavior → Exposure relationship.
Researchers studying the relationships between victimization and aggressive
behavior have found support for either bi-directionality or a Behavior → Exposure
relationship. For instance, Schwartz, McFadyen-Ketchum, Dodge, Petit, and Bates (1999)
demonstrated that early aggressive behavior places children at risk for being victimized
by peers later on. In regards to community violence exposure, there is a small but
growing body of research that lends support to the claim that early aggressive behavior
12

leads to community violence exposure later on (Boyd et al., 2003; Mrug & Windle, 2009;
Farrell & Bruce, 1997; Farrell & Sullivan, 2004). These studies all employed longitudinal
designs to examine how community violence exposure and aggression relate to each
other over time. Boyd and colleagues (2003) found that children who were aggressive in
the first grade were more likely than non-aggressive first graders to be exposed to
community violence during the middle school years. Similarly, Mrug and Windle (2009)
found that young adolescents who exhibited early externalizing problems that included
aggressive behaviors and conduct problems were more likely to be exposed to
community violence two years later. Furthermore, several of these studies have found
gender differences in the relationships. For instance, Farrell and Bruce (1997) found only
an effect for females. Females who were aggressive early on were more likely than non-
aggressive females to become exposed to community violence. They did not find this
effect for boys. Mrug and Windle (2009), however, did not find any gender differences in
this association. Rather they found that aggressive behavior predicted later exposure to
community violence regardless of gender. While not much attention has been given to
this area, these studies point out the importance of generating more research to better
understand how relationships between exposure and aggressive behavior may be different
for males and females.
Lack of theory. An additional methodological issue is while some researchers
have relied heavily on theoretical frameworks to guide their investigations, many
researchers have not. This poses serious problems for comparing the results across
studies. If researchers are not unified in their approach from study conceptualization to
13

implementation, then the field will not be able to move forward. Theory is essential to
organize studies in terms of the operationalization of community violence, the aspects of
community violence that are measured, the measure that is used, the design of the study
and the manner in which the data is analyzed.
Additionally, as the small but growing body of research above indicates, the
theory used by community violence researchers, in particular Social Learning Theory
(Bandura, 1986), may be incomplete. As discussed earlier, community violence
researchers have been primarily guided by an Exposure → Behavior model. But the
studies discussed above provide evidence for a Behavior → Exposure model. Researchers
in the sociological and criminological realms have been using Behavior → Exposure
models for several years. One model, in particular, that has received a good deal of
attention is the Lifestyle Exposure Theory (Jenna & Brownsfield, 1997). This theory
posits that violent offending and other forms of antisocial behavior are indicators of a
lifestyle that places individuals at risk for violent victimization. That is, aggressive
adolescents may be more likely to select themselves into environments where they are
more likely to be exposed to community violence than non-delinquent children (Gorman-
Smith & Tolan, 1998; Schwartz & Proctor, 2000). Evidence for this has been found in the
literature (Cicchetti & Lynch, 1993).
Co-occurring risk. Another methodological issue is that children and adolescents
exposed to one form of violence are often exposed to multiple forms of victimization.
Coined by Finkelhor (2007), polyvictimization asserts that children who are exposed to
one form of violence are more likely to be, or have been, exposed to other forms of
14

violence. Finklehor (2007) argues that studies often fail to account for co-occurring
victimization and violence exposure. Failure to do so could lead to a serious
overestimation of the impact of the single form of victimization of interest to the
researchers, because all of the trauma may be related to the unmeasured form of
victimization or a combination of the measured and unmeasured, rather than the single
form that is measured. The existing literature may have, to some degree, exaggerated the
strength and consistency of the relationship between community violence exposure and
aggression. Additionally, children and adolescents exposed to violence also face other
risks that tend to go unaccounted for in research, such as a mother’s mental health
problems or substance abuse, poverty, and poor nutrition (Margolin & Gordis, 2000).
Co-occurring risk factors and victimization need to be better accounted for in
research studies. It is important to note that the literature examining how various forms of
victimization and other risk factors relate to one another and lead to maladaptive
outcomes is fairly non-existent. Future research needs to address this gap in the literature.
Understanding whether the experience of multiple forms of victimization creates an
increased risk for maladaptive outcomes in children and adolescents could lead to the
development of effective interventions that ameliorate the effects of violence exposure.
Child Maltreatment, Aggressive Behavior, and Community Violence Exposure
One form of victimization that has been strongly linked to increased risk for
aggressive behaviors in adolescents is child maltreatment; in fact, this is one of the most
consistent findings in both the aggression and the maltreatment literature (Bolger &
Patterson, 2001; Cicchetti & Lynch, 1995; Dodge, Bates, and Pettit 1990; Lynch &
15

Cicchetti, 1998; Smith & Thornberry, 1995; Trickett & McBride-Chang, 1995).
Individuals with a history of child maltreatment are at elevated risk to become exposed to
other forms of victimization (Finklehor, 2007). For example, Schwartz, Toblin, Abou-
ezzedine, Tom, and Stevens (2005) found that children who experience harsh parenting
are more likely to become victims of aggression at the hands of their peers. In terms of
community violence exposure, research has indicated that neighborhoods with high rates
of child maltreatment are also marked by high incidence rates of violent crimes
committed against persons (Coulton, Crampton, Irwin, Spilsbury, & Korbin, 2005). Thus,
it is likely that children and adolescents who have been maltreated are also exposed to
increased levels of violence in their communities. Yet, research does not exist to examine
how these two forms of victimization might relate to one another and also relate to
aggressive behaviors in adolescents.
Such research is necessary given that children and adolescents with a history of
child maltreatment are at risk for additional forms of victimization it is important to
understand how these forms of violence exposure relate to one another and aggressive
behavior. The relationship between community violence exposure and aggressive
behavior may differ for adolescents with a history of child maltreatment and for those
with no such history. According to the “social push” hypothesis put forth by Raines and
Venables (1984), the adverse experience of early child maltreatment influences
adolescents’ aggressive behavior above and beyond any other experience of violence.
That is, for maltreated children, the experience of community violence will not have
much of an effect on their levels of aggressive behavior. For individuals without the
16

adversity of early child maltreatment, other factors come into play to influence aggressive
behavior. According to the hypothesis, community violence exposure would influence
aggressive behavior more so for adolescents without the adverse experience of child
maltreatment.
Issues in the maltreatment literature. Methodological issues that make it difficult
to compare the findings across studies have also plagued the literature on child
maltreatment. One problem faced by researchers is that most studies have predominantly
employed either a cross-sectional or retrospective design. The issue surrounding cross-
sectional designs is that, by nature, they tend to measure outcomes attributed to
maltreatment around the same time that the maltreatment occurred. Thus, neither the
delayed effects of maltreatment, nor the effects of maltreatment occurring at different
developmental periods can be observed (Trickett & McBride-Chang, 1995). The
drawback of retrospective designs is the reliance on the participant’s memory. Over time,
memories are distorted. Additionally, the classification of the participants as maltreated
also relies on personal recollections and perceptions of their experiences, which differs
from studies that use samples referred from official agencies. The use of longitudinal
designs would greatly increase the understanding of both the proximal and distal effects
of child maltreatment and also the impacts of child maltreatment that occurs at different
developmental periods.
Another issue concerning the design of studies examining the effects of
maltreatment is that most studies examining child maltreatment have had no meaningful
comparison groups. Thus, it becomes difficult to make firm conclusions about outcomes
17

associated with child maltreatment. The outcomes associated with child maltreatment are
also associated with poverty and other factors. If comparison groups are not included, it is
difficult to parse out how much of an observed outcome is due to maltreatment and how
much is due to poverty and other factors (Trickett & McBride-Chang, 1995).
Furthermore, the manner in which researchers have defined child maltreatment
has been inconsistent and vague to allow for the interpretation of what was actually
experienced. Most studies simply classify adolescents as either “maltreated,” “physically
abused,” or “neglected” without any further information given about the groups. These
vague descriptions of maltreatment experiences make it difficult to examine the
differential outcomes associated with the various forms of maltreatment. It is also
difficult to compare the findings of one study with another because the samples are not
necessarily the same. Research needs to work toward a more accurate way of classifying
their samples and use this information to examine the differential outcomes related to the
various forms of child maltreatment (Trickett & McBride-Chang, 1995). Currently, there
is no body of research examining the relationships between community violence
exposure and aggressive behavior and whether those relationships differ depending on
specific maltreatment experiences.
Focus of the Current Study
The current investigation seeks to elucidate the longitudinal relationships between
community violence exposure and aggressive behavior in a sample of 389 maltreated and
non-maltreated young adolescent males and females. Of primary interest will be to
examine the effect of adolescents’ early community violence exposure and aggressive
18

behavior on their later community violence exposure and aggressive behavior as well as
examining how these relationships differ by gender and maltreatment history. The present
study will be conducted in the context of a larger longitudinal study examining the
impacts of child maltreatment on adolescent development. The data used in the current
investigation will be drawn primarily from the second (T2) and third (T3) data collection
waves of the parent study.
This study will contribute to the extant literature in several significant ways. First,
the design of the study will seek to address several of the methodological issues that have
plagued both the community violence and maltreatment literature. Second, the current
study will examine whether community violence exposure contributes to aggressive
behavior, whether aggressive behavior contributes to community violence exposure, or
whether community violence exposure and aggression relate to each other in a reciprocal
manner. Third, the current study will shed light onto how the longitudinal relationships
differ by group membership. Results of this study could have important implications for
the conceptualization and implementation of research and clinical intervention
The following aims are addressed by the current study:
Specific Aim 1: In a sample of maltreated and non-maltreated female and male
adolescents, determine the longitudinal relationships between community violence
exposure and aggressive behavior.
Specific Aim 2: Determine if the longitudinal relationships between community violence
exposure and aggressive behavior are different for male and female adolescents.
19

Specific Aim 3: Determine if the longitudinal relationships between community violence
exposure and aggressive behavior are different for maltreated and non-maltreated
adolescents.
Specific Aim 4: Determine if the longitudinal relationships between community violence
exposure and aggressive behavior are different among physically maltreated, non-
physically maltreated, and non-maltreated adolescents.
20

CHAPTER TWO: RESEARCH DESIGN AND METHODS
Participants
The participants for this study are 389 adolescents who are enrolled in the Young
Adolescent Project (YAP). The YAP is an ongoing National Institute of Child Health and
Development (NICHD) funded longitudinal study that examines the effects of child
maltreatment on adolescent development. The YAP has so far completed three waves of
data collection and is currently engaged in a fourth. The present study draws mostly upon
data from the second (T2) and third waves (T3) of the YAP. T2 and T3 took place
approximately 18 months apart from one another. Slight attrition occurred between T1
(N=454) and T3 (N=309) and will be discussed in more detail in this section. To
maximize the data available for analyses, the participants who completed T1, T2 and T3
(N=306), only T1 and T2 (N=80), or only T1 and T3 (N=3) will be included in this study.
The statistical methods used to include all 389 participants will be discussed in more
detail in the following section.
Recruitment of the Sample
Maltreated young adolescents were recruited from active substantiated cases of
neglect and abuse brought to the attention of the Los Angeles County Department of
Child and Family Services (LACDCFS). To control for the diversity inherent in the large
area serviced by the LACDCFS, participants were drawn only from 10 zip codes in Los
21

Angeles County. Based on LACDCFS data and census tract information, these zip codes
were chosen because of their manageable travel distance to USC (to facilitate family
participation) and also large numbers of the most prevalent ethnicities in the urban Los
Angeles area (African American, Hispanic, and Caucasian).
A total of 303 male and female adolescents (and their primary caretakers) were
recruited into the study. Inclusion criteria for the neglect sample at the first time-point
were: (1) a new substantiated report of any type of maltreatment to the LACDCFS in the
preceding month; (2) children between the ages of 9 and 12 years of age; (3) child
identified as Black, Latino or White (non-Hispanic); and (4) at time of referral, child
resided in one of 10 zip codes described above. After receiving approval from
LACDCFS, the courts, and USC University Park Institutional Review Board, families
meeting the above criteria were contacted via mail, and asked to return a postcard
indicating their willingness to participate. Phone calls were then made to families
indicating their willingness to participate. Recruitment of the maltreatment sample began
in spring 2002 and ended in Fall 2004.
A comparison sample of non-maltreated adolescents (N=151) was recruited from
the same 10 zip codes as the maltreated sample via school lists of families with 9 to 12
year-olds obtained from a marketing firm. Caretakers of potential comparison adolescents
were contacted via mail and asked to return a postcard to indicate their willingness to
participate. Phone calls were then made to families indicating their willingness to
participate. Recruitment of the comparison participants began after that of the maltreated
participants, and was completed in Fall 2005.
22

Sample Demographics
As displayed in Table 1, 41% of the maltreated sample was Black, 35% Latino,
11% White, and 13% biracial. Approximately 50% of the maltreated group was female.
The average age of the maltreatment group at T1 was 10.84 years of age. As for the non-
maltreated group, 32% was Black, 47% Latino, 10% White, and 11% biracial.
Approximately 40% of the non-maltreated group was comprised of females. The average
age of the non-maltreated group at T1 was 11.11 years of age.
Upon entry into the study, the maltreated and non-maltreated samples were fairly
similar in age, gender, and ethnic distribution, but differed in terms of living
arrangements. A greater proportion of the non-maltreated group lived with a biological
parent than the maltreated group (93% and 51%, respectively), the latter being more
likely to reside in foster care or with extended biological family.
The neighborhood characteristics of the maltreated and non-maltreated groups
were examined for significant differences using the 2000 US Census data. In particular,
comparisons were made of social, educational, economic, and demographic variables. Of
the 72 characteristics examined, nine significant differences emerged, but none that were
likely to produce an effect because of a relationship with another variable. These results
indicate that the neighborhoods are similar environments for both groups.
Procedures
Approximately 12 months after their T1 interview, families were contacted and
scheduled for the T2 data collection. Upon arrival to the research offices, informed
consent was obtained from both the adolescent and caretaker. Trained research assistants
23

interviewed the adolescents and caretakers in separate rooms. Adolescents completed
several standardized measures that assessed cognitive abilities, behavior problems, health
and physical development, psychiatric symptoms, family functioning, and trauma history.
Caretakers completed several standardized caretaker-report-on-adolescent questionnaires
designed to measure their adolescent’s behavior problems, health and physical
development, psychiatric symptoms, and trauma history. Caretakers also completed
several standardized self-report questionnaires designed to obtain information on their
psychiatric symptoms, family functioning, trauma history, attitudes and beliefs about
parenting, and parenting behaviors. Assessments were completed in one, same-day
session that lasted between four and five hours. At the close of the interview, participants
were debriefed and compensated for their participation following the guidelines specified
by the National Institute of Health’s Volunteer Program. This procedure was repeated 18
months later for the T3 data collection.

Table 1. Demographics of the sample over three waves of data collection.
24

In addition to collecting adolescent self-report and caregiver report-on-adolescent,
teacher report-on-adolescent data was also collected at T2. Permission to contact the
teacher was obtained from the caregiver during the informed consent process. Caregivers
were asked to identify a current teacher that knows their adolescent well. A battery of
standardized questionnaires was mailed to the teachers. Teachers were informed that the
adolescent and their family were participants in a study that was being conducted by the
University of Southern California. The teachers were not informed of the nature of the
study or why the adolescent was selected to participate. The questionnaire packet was
designed to obtain information on the adolescent’s behavior problems and
psychopathological symptoms in the school setting. As with the adolescents and their
caregivers, teachers were compensated for their time and effort. The response rate at T2
was modest. Despite the efforts of the research staff to increase teacher participation,
only 60% (N=236) of the questionnaire packets were returned. This procedure was
repeated 18 months later for T3. The response rate improved dramatically; approximately
71% (N=226) of the packets were obtained from teachers.

25

Table 2. Attrition across T1, T2, and T3.

Table 2 illustrates the sample retention from T1 to T3. Eighty-six percent of the
participants (N=391) returned at T2 (82% of the maltreated group; 91% of the non-
maltreated group). Seventy-one percent returned at T3 (64% of the maltreated group;
85% of the non-maltreated group). Most of the participants who attrited from the study
became unreachable via phone, mail, and family and friends whom they listed as contacts
at T1 and T2. Only 15 families formally withdrew from the study. These data will be
more fully explored in the following section.
26

Operationalizing Child Maltreatment Experience in the YAP Sample.
Official classification of child maltreatment often simplifies the experience the
child has, and in the process a good deal of information that could be important to
understanding how a child’s particular experience may contribute to their negative
outcomes is lost. The Maltreatment Case Record Abstraction Instrument (MCRAI:
Trickett, Mennen, Kim, & Sang, 2009) was developed for the YAP to better characterize
adolescent’s maltreatment experiences. The MCRAI coded for the type of maltreatment
and for specific information regarding the incident, including: the perpetrator,
relationship of perpetrator to child, age of child at onset of maltreatment, frequency,
duration, and specifics of the maltreatment (i.e., hospitalization, marks apparent, etc.).
Additionally, information on parental functioning in regards to substance abuse, domestic
abuse, and psychiatric and physical health problems were also included.
Comparisons of the LACDCFS classifications of the adolescent’s experiences and
the MCRAI classifications reveal enormous disparities. For instance, while LACDCFS
reports indicate that 56 out of 303 (18% of the sample) adolescents in the sample were
physically maltreated; the MCRAI found that 148 (49% of the sample) of the adolescents
met criteria for having experienced physical child maltreatment. In addition, the MCRAI
identified more children who met criterion for experiencing multiple forms of
maltreatment than was indicated by the official report from LACDCFS.
Results of the MCRAI (Trickett et al., 2009) indicate that the maltreated group
has the following characteristics: the mean number of referrals to LACDCFS for
suspicion of maltreatment was 4.9 (standard deviation = 3.3) with a range from 1 to 17
27

reports; seventy-two percent of the maltreated group experienced general to severe
neglect (N=218), 49% experienced physical maltreatment (N=149), 20% sexual abuse
(N=61), 48% emotional abuse (N=145), 52% experienced caretaker incapacity (N=158),
and 53% were at substantial risk (N=161). Substantial risk occurs when there is no
evidence for maltreatment but a case is substantiated for a sibling. Of those that are at
substantial risk, 91% experienced an additional form of maltreatment (N=147).
Approximately 52% (N=158) of the maltreated group have multiple classifications.
Maltreatment experience will be used as a grouping variable in the structural
models to examine whether maltreatment history or the type of maltreatment plays a role
in the association between community violence exposure and aggressive behavior. Two
methods will be used to create grouping variables. The first method will be to treat child
maltreatment as a dichotomous variable. That is, the variable will contain two categories,
maltreated and non-maltreated. The second method is to create a “maltreatment” variable
with three categories: physically maltreated, non-physically maltreated, and non-
maltreated. As discussed previously, one of the methodological issues in the
maltreatment literature is that researchers usually tend to use simple dichotomies (e.g.,
maltreated versus non-maltreated) to examine their samples and as a result limit the
external validity of their results. This is because the maltreated group is comprised of
individuals with various forms of experiences that could range from neglect to sexual
abuse (Trickett & McBride-Chang, 1995). The differential outcomes of these various
experiences go undetected. The MCRAI offers the opportunity to examine whether there
are differential outcomes based on specific forms of maltreatment. The physically
28

maltreated group will contain all adolescents that experienced at least one instance of
physical maltreatment and/or sexual abuse (sexual abuse will be included in this category
because of the physical nature of the experience). It should be noted that adolescents who
comprise the physically maltreated are likely to have experienced other forms of
maltreatment (i.e., neglect, emotional abuse, etc.), thus this category is not a clean one.
The non-physically maltreated group will include participants that have experienced other
forms of maltreatment but have not experienced any incidents of physical maltreatment
(as indicated by the official records supplied by LACDCFS).
Child Self Report Measures
Community violence exposure. The Community Violence Index (CVI, see
Appendix B: based on Richters & Saltzman, 1990) was used to assess children’s
exposure to community violence at T2 and T3. This measure includes 19 items that cover
a wide variety of violent events that occur within the community. Two subscales are
thought to comprise this measure, witnessing (11 items; seen someone threaten to beat
someone up, seen someone chased, seen someone choked or strangled, etc.) and
victimization (8 items; been beaten up, been shot at, been robbed, etc.). As will be
discussed later, the witnessing and victimization subscales were not examined separately
in the current study. Instead, they were combined to assess adolescents’ overall
community violence exposure.
Adolescent self-reports of exposure were collected in a one-on-one interview
format. Adolescents were instructed to only report incidents that have occurred within the
contexts of their neighborhood and their school, and to disregard incidents that have
29

occurred within their home, observed in the media, or been learned through hearsay. For
each item, adolescents were asked to indicate whether they experienced the item (yes or
no) and how many times they experienced that form of community violence during their
lifetime and over the past year. Responses were initially recorded as count variables but
were rescaled to an ordinal scale (0 incidents = 0; 1 incident = 1; 2-3 incidents = 2; 4-5
incidents = 3; 6 or more incidents = 4) to better account for the relatively low base rates
of the various experiences and, accordingly, the highly skewed distributions. In the
current study, exposure over the past year at T2 and T3 was examined.
Aggressive behavior and delinquency. The Youth Self-Report (YSR, see Appendix
C; Achenbach, 1991) is an extensively researched tool for the assessment of adolescent’s
behavioral (i.e., aggression, delinquency, hyperactivity, etc.) and psychological problems
(i.e., depression, anxiety, etc.). It represents the self-report version of a multiple
informant strategy that also includes from the caregiver report on adolescent (Child
Behavior Checklist: Achenbach, 1991) and teacher report on adolescent (Teacher Report
Form: Achenbach, 1991). The YSR is comprised of 112 items. Respondents use a 3-point
scale (0 = not at all true, 1 = somewhat true, and 2 = very true) to rate how well a series
of problems describe them. The measure is arguably the most extensively normed
measure available for assessing adolescent problem behavior (Lambert, Schmitt, Samms-
Vaughan, An, Fairclough & Nutter, 2003). The measure features high test-retest
reliability and high internal consistency. Validity is supported by studies which have
found that the YSR is related to important functional outcomes (Achenbach, 1991). The
current study makes use of 20 items that comprise the aggressive behavior subscale (i.e.,
30

destroys things belonging to others, fights, temper tantrums, physically attacks people
etc.). Additionally, 11 items that comprise the delinquency subscale were used in the
current study (i.e., lying or cheating, running away from home, stealing at home or
outside, using alcohol or drugs, etc.). The data from T2 were used. A composite index
was constructed by summing the scale items. This composite was used as a covariate in
the substantive models.
Caregiver Report Measures
Adolescents’ aggressive behavior. The Child Behavior Checklist (CBCL, see
Appendix D) was utilized to assess adolescent’s aggressive behavior. The CBCL
represents the caregiver report on adolescent version of the multiple informant strategy to
assess adolescent problem behavior. As with the YSR, the CBCL is comprised of 112
items. Items are rated on a 3-point scale assessing the frequency of occurrence (0 = not at
all true, 1 = somewhat true, and 2 = very true). As with the YSR, the CBCL is an
extensively researched instrument and yields strong psychometric properties. The
measure features high test-retest reliability and high internal consistency. Validity is
supported by studies which have found that the CBCL is related to important functional
outcomes (Achenbach, 1991). The current study makes use of 20 items (that correspond
to the items from the YSR) that comprise the aggressive behavior subscale (i.e., argues,
fights, louder than other children, destroys things, etc.).
Caregiver Psychopathology. The caretaker’s depressive and anxiety symptoms
were assessed using the Brief Symptom Inventory (BSI; Derogatis & Melisaratos, 1983).
The BSI is a well-validated measure of adult psychological symptoms and has been used
31

profitably in studies of child maltreatment (Cox, Kotch, & Everson, 2003). The measure
includes six items tapping the depression construct that are rated on a 5-point scale from
0 (not at all) to 4 (extremely). Respondents rate how frequently they have experienced the
item description in the past seven days. These scales were used to examine attrition.
Simple composite scales were developed by summing the scales. Data were used from T1
to examine attrition at T2 and data from T2 to examine T3 attrition.
Socio-demographic variables. Several demographic variables will be used to
examine attrition across time points. These variables include: adolescent’s age, ethnicity,
living situation, family composition, caregiver’s level of education, maltreatment status,
and family SES. Age and family SES were used as covariates in the substantive structural
models. Both have been found to be associated with both community violence and
aggressive behavior in the literature (Mrug & Windle, 2009; Schwab-Stone et al., 1995).
Other demographic variables that account for attrition were used as covariates in the
substantive structural models.
Teacher Report Measures
Adolescents’ aggressive behavior. The Teacher Report Scale (TRS, see Appendix
E: adapted from the Aggressive Behavior – Teacher Checklist; Dodge & Coie, 1987) was
used to assess adolescents’ aggressive behavior in the school setting at T2 and T3. The
TRS is an 11-item measure that assesses the use of proactive aggression (6 items: i.e.,
student teases and name calls; student uses force to dominate others; etc.) and reactive
aggression (5 items: i.e., student strikes back when teased; student blames others in
fights; etc.) in the classroom. Items are rated on a 5-point scale assessing the frequency of
32

occurrence (1 = almost never happens; 3 = in between; and 5 = happens very often). In
the current study, responses were recoded onto a three-point scale (i.e., 1-2 = 0; 3 = 1;
and 4-5 = 2) so that responses could be compared directly to the responses on the CBCL
and YSR. The psychometric properties of this measure are not known, but similar scales
drawn from other studies have demonstrated reliability and validity.
33

CHAPTER THREE: RESULTS
Preliminary Analyses
Item-level missingness. Longitudinal designs must contend with the fact that there
will be some amount of missing data. Data collection in the parent study was carefully
conducted and therefore item-level missingness was kept at a minimum. Most items have
rates of missing less than one percent and no variables featured missingness greater than
five percent. For missingness at such low rates, a single imputation is sufficient to yield
accurate estimates of parameter values. The percentage of missing data in the current
study was very low, about one percent. Thus, multiple imputation using NORM software
program (Schafer, 1999) was used to impute data missing at the item level.
Sample attrition. Substantial attrition was witnessed across the measurement
periods, with 86% of the sample being retained at T2 and 71% of the sample at T3. This
attrition rate is acceptable, given the highly distressed nature of the sample. Attrition data,
sorted on the basis of demographic variables, are displayed in Table 2.
Logistic regression analyses were conducted in order to help determine the
patterns of missingness and examine the degree to which observed variables could
explain attrition. Analyses first examined the loss of data from T1 to T2 and next,
examined attrition from T2 to T3. Psychosocial variables, including caregiver and
adolescent report data, along with demographic variables were entered in the model
examining the data loss from T1 to T2.

34

Results of the logistic regressions examining attrition between T1 and T2
indicated that poverty status predicted attrition with an odds ratio (OR) of 1.92 (95% CI
1.04-3.52) as did maltreatment history (OR=2.70 95% CI 1.21-6.03). The Cox & Snell R-
Square, however, was .035. For the models examining attrition between T2 and T3,
psychosocial variables from T2 were added to the model. Results of the analyses
examining data loss from T2 to T3 indicated that several variables predicted whether
participants dropped out. As with the Time 1 to T2 analyses, poverty status (OR1.66;
95% CI: 0.92 -2.98), and maltreatment history (OR 2.80; 95% CI: 1.48-5.32) predicted
attrition as well as Latino (OR 3.91; 95% CI: 2.06-6.68) and Caucasian (OR 3.73; 95%
CI: 1.64-8.48) ethnicity. Cox and Snell R-Square statistic was .097.
Models were also explored to determine the patterns of missingness and examine
the degree to which observed variables could explain the non-response of teachers. At
T2, only 60% of the teachers responded and at T3, approximately 71% returned data.
Similar models to the ones run to examine the attrition of study participants were
conducted. Results for teacher non-response at T2 indicated that maltreatment history
(OR 1.92; 95% CI: 1.23-3.00) and living arrangement (removed from home) predicted
non-response (OR 1.72; 95% CI: 1.12-2.68). The Cox & Snell R-Square, however, was a
mere .023. Results for T3 data indicated that gender (male) predicted non-response (OR
1.63; 95% CI: 1.08-2.46), as well as maltreatment history (OR 1.86; 95% CI: 1.20-2.87).
The Cox & Snell R-Square, however, was .038.

35

These findings suggest that models including the variables that were associated
with attrition would contribute to the functioning of full information maximum likelihood
(FIML) procedures in providing unbiased parameter estimates. These variables will be
included in the structural models, described later, as covariates.
Descriptive information. The raw score means, standard deviations, ranges,
variance, skewness, and kurtosis for the variables of interest in this study are presented in
Table 3. Basic group differences were found between the mean level of community
violence exposure for females and males at both T2 and T3. Simple t-tests revealed that
males had a significantly higher mean than girls at both time points (t=-4.486 (387),
p<.000 and t=-2.647 (306), p=.04). There were no significant differences found for the
non-maltreated and maltreated groups on mean levels of community violence exposure.
Basic group differences were found between the mean level of aggressive behavior for
non-maltreated and maltreated adolescents at both T2 and T3 (t=-4.404 (387), p<.000 and
t=-3.703 (306), p<.000). Simple t-tests demonstrated that the maltreated group had a
significantly higher mean than the non-maltreated group at both T2 and T3. There were
no significant differences found among the mean levels of aggressive behavior for
females and males at either T2 or T3. Consonant with prior research (Trickett &
McBride-Chang, 1995), the physically maltreated group had the highest scores on
aggression at T2 and T3 (7.82 and 7.33, respectively, compared to 5.42 and 5.12 for the
36

non-maltreated group and 7.09 and 6.41 for the non-physically maltreated) as revealed by
one-way ANOVAS (F=10.564 (2), p<.000 and F=8.102 (2), p<.000). Appendix A
includes additional tables (Tables 11 – 14) with descriptive information about the
variables by gender, maltreatment status, and specific maltreatment type.

Table 3. Means and standard deviations for the variables of interest at T2 and T3.

Correlations were computed between the variables of interest (Table 4).
Significant correlations were found among all of the aggressive behavior variables for T2
(r = .244 to .360, p<.05) and T3 data (r = .192 to .307, p<.05). The community violence
exposure variables were also found to have significant correlations for both T2 (r = .567
to .637, p<.05) and T3 (r = .494 to .569, p<.05). The aggressive behavior variables had
significant correlations with the community violence variables at T2 (r = .185 to .294,
p<.01) and for most of the aggressive behavior variables at T3 (r = .160 to .352, p<.01).
The correlation between community violence exposure at T3 and caregiver report of
adolescent’s aggressive behavior at T3 was only marginally significant (r = .105, p<.10).
37

Appendix A includes additional tables (Tables 15 – 17) that examine correlations among
the variables by gender, maltreatment status, and specific maltreatment type.

Table 4. Correlations among the variables of interest at T2 and T3.

Table 5 displays the nineteen community violence exposure items and the
percentage of adolescents who reported at least one experience of one of the types of
violence at T2 and T3. The percentages are quite similar across the two time points. As
expected, the less severe items (i.e., Seen some threatened to be beaten up, seen someone
beaten up, etc.) appear much more frequently than the less severe items (i.e., Seen a
murder, been stabbed, etc.).

39
38

Table 5. Percentages of the sample exposed to instances of community violence at T2 and
T3.

Further analyses revealed that for the full sample, 56% of adolescents who
reported exposure to at least one incident of community violence exposure at T2 also
reported exposure at T3. Basic group differences in the stability of exposure were
observed. Among gender, 49% of females who reported exposure at T2 also reported
exposure at T3; for the males it was 63%. Similarly, among maltreatment status, 46% of
the non-maltreated group adolescents who reported violence exposure at T2 also reported
it at T3 while for the maltreated group it was 63%. These results are summarized in Table
6.

39

Table 6. Percentage of sample with no exposure at T2 and T3 versus those exposed at
both time points.

Determining Factor Structure for Child-Rated Community Violence Exposure
The psychometric characteristics of the CVI were examined before conducting the
substantive analyses. Specifically, analyses investigated whether the CVI reliably
measured adolescents' exposure to community violence among the full sample and also
whether the measure functioned similarly across subgroups within the sample (male vs.
female; non-maltreated vs. maltreated; non-maltreated vs. physically maltreatedvs. other
maltreated). Additionally, longitudinal invariance was assessed for the full sample and
for each of the subgroups because cross-lagged analyses assume constructs are being
measure consistently across time points.
Structural equation modeling (SEM) conventions assert that there should be at
least 5 – 10 participants per parameter to yield accurate model estimates (Loehlin, 2004).
This consideration is important not only in examining model with the full sample but
especially when examining multiple group models. To lessen the number of parameters
in the model and increase the accuracy of model estimates, a parceling strategy will be
employed. Parceling has been defended as item level modeling can compromise
40

reliability, features a larger ratio of unique-to-common factor variance and is more likely
to evidence distributional violations (Bandalos & Finney, 2001; Yang, Nay, & Hoyle,
2010; Little, Cunningham, Shahar & Widaman, 2002). Little and colleagues (2002)
suggest that when the researcher is interested in relations between latent constructs rather
than explorations of the psychometric properties of the measure, as in the current
investigation, the obscuration of “nuisance” factors at the item level do not compromise
the interpretability of the parceled results. The nineteen items of the CVI were randomly
assigned to one of three parcels. Due to the odd number of items, two of the parcels
contained six items and one parcel contained seven items. The parcels were identically
configured for T2 and T3. These parcels were then modeled as manifest indicators of a
latent construct representing adolescents' community violence exposure.
A restricted factor analyses was conducted to determine if the three parcels
provided measurable and reliable scales for the CVI for both the T2 and T3 data. The
factor structure was tested using Structural Equation Modeling (SEM) with Mplus
(Version 5.21: Muthen & Muthen, 2009). Several fit indices such as the χ2 (chi-square)
goodness-of-fit statistic, comparative fit indices (CFI), and the root mean square error of
approximation (RMSEA) were used to evaluate the fit of the model to the data. Overall, a
small χ2 is optimal, a CFI close to 1.00, and an RMSEA lower than .080 when the model
specified to load on one and only one factor. The first item on each latent factor was set
to 1.0 to establish the metric and allow the models to be identified.

41

The three indicator model demonstrated perfect fit to the data for both T2 (χ2=0,
df=0, CFI=1.00, RMSEA=.000, 90% CI=.000-.000) and T3 data (χ2=0, df=0, CFI=1,
RMSEA=.0, 90% CI =.000 - .000). Perfect fit was obtained because the models were
fully saturated (i.e., no degrees of freedom). All loadings were significant and their
standardized values ranged from .716 to .823 at T2 and from .707 to .896 at T3.

Figure 1. CFA of community violence exposure data at T2.

Figure 2. CFA of community violence exposure data at T3.

42

Invariance of community violence exposure across groups. Evidence for
invariance between genders was obtained using multiple group structural equation
modeling (MGSEM). The MGSEM approach allows the researcher to test the
equivalency of measurement across groups. Invariance is tested by comparing a model in
which the parameters are allowed to be freely estimated to a model which the factor
loadings and intercepts are held equal across groups. The χ
2
of the two models are
compared. If there is not a significant difference between the models then it is concluded
that the constructs were measured the same across groups. If the χ
2
is significant, then the
construct is not the same across groups and the models are examined to determine how
they are different. If necessary, the models are revised. Guided by Byrne and Stewart
(2006), the current study sought at least weak metric invariance between groups. Weak
metric invariance is obtained when the pattern and strength of factor loadings are
equivalent across groups.
First an unrestricted model was tested in which factor loadings and intercepts
were freely estimated for each gender. The unconstrained models fit adequately for both
T2 (χ2 =5.408, df=2, CFI=.990, RMSEA=.094, 90% CI=.000 - .193) and T3 (χ2 =1.32,
df=2, CFI=1.00, RMSEA=.000, 90% CI=.000 - .142) data. Standardized loadings for
females ranged from .716 to .823 at T2 and from .707 to .896 at T3. Standardized
loadings for males ranged from .775 to .819 at T2 and from .775 to .819 at T3. Next,
factor loadings and intercepts were restricted to be equal across the groups. A
43

nonsignificant χ2 difference between the unrestricted and the restricted models indicates
invariance of measurement. The χ2 difference was not significant for either the T2
(Δχ2=0.927, Δdf=1, p>.05) or the T3 (Δχ2=3.026, Δdf=2, p>.05) data. These results
support measurement invariance for the community violence exposure variable across
gender at both T2 and T3.
Evidence for invariance across the non-maltreated and maltreated groups was
obtained using the same procedure. The unconstrained models fit adequately for both T2
(χ2 =7.195, df=2, CFI=.988, RMSEA=.116, 90% CI=.033 - .212) and T3 (χ2 =3.09, df=2,
CFI=.996, RMSEA=.060, 90% CI=.000 - .183) data. Standardized loadings for non-
maltreated group ranged from .679 to .835 at T2 and from .750 to .887 at T3.
Standardized loadings for the maltreated group ranged from .740 to .838 at T2 and from
.519 to .800 at T3. The χ2 difference the unrestricted and the restricted models was not
significant for either the T2 (Δχ2=4.841, Δdf=2, p>.05) or the T3 data (Δχ2=6.411,
Δdf=2, p>.05). These results provided evidence for measurement invariance for the
community violence exposure variable across the non-maltreated and maltreated groups
at T2 and T3.
Invariance across the non-maltreated, physically maltreated, and neglect groups
was examined. The unconstrained models fit adequately for both T2 (χ2 =7.195, df=2,
CFI=.988, RMSEA=.116, 90% CI=.033 - .212) and T3 (χ2 =3.09, df=2, CFI=.996,
RMSEA=.060, 90% CI=.000 - .183) data. Standardized loadings for the non-maltreated
group ranged from .630 to .831 at T2 and from .756 to .883 at T3. Standardized loadings
for the physically maltreated group ranged from .637 to .762 at T2 and from .441 to .888
44

at T3. Standardized loadings for the non-physically maltreated group ranged from .800 to
.888 at T2 and from .516 to .705 at T3. The χ2 difference between the unrestricted and
the restricted models was not significant for either the T2 data (Δχ2=4.841, Δdf=2, p>.05)
or the T3 data (Δχ2=6.411, Δdf=2, p>.05). These results support measurement invariance
for the CVI across the non-maltreated, physically maltreated, and non-physically
maltreated groups at T2 and T3. These results support measurement invariance for the
community violence exposure variable across non-maltreated, physically maltreated, and
non-physically maltreated groups at both T2 and T3.
Invariance of community violence exposure across time. It is also important to test
whether the construct is invariant across time, given the focus of this study on the
longitudinal relationships. For the full sample as well as each subgroup, it s important to
know whether the construct measured at T2 is the same as the one measured at T3.
Longitudinal invariance was assessed across the two time periods by comparing a model
in which the parameters were allowed to be freely estimated to a model in which the
factor loadings and intercepts were held equal across time. For the full sample, model fit
for the unrestricted model was good (χ2=11.27, df=4, CFI=.990, RMSEA=.068, 90%
CI=.023 - .117). Standardized loadings ranged from .717 to .813 for T2 and from .679 to
.854 for T3. The χ2 difference between the unrestricted and the restricted models was not
significant for (Δχ2=3.33, Δdf=3, p>.05). These results provided evidence that for the full
sample community violence exposure was measured consistently between T2 and T3.

45

For males and females, model fit for the unrestricted model was good (χ2=23.21,
df=14, CFI=.986, RMSEA=.058, 90% CI=.000 - .099). Standardized loadings for the
females ranged from .467 to .794 at T2 and from .704 to .841 for T3. Standardized
loadings for the males ranged from .779 to .819 at T2 and from .629 to .841 for T3. The
χ2 difference between the unrestricted and the restricted models was not significant for
(Δχ2=12.23, Δdf=7, p>.05). These results provide evidence that the community violence
exposure variable was measured consistently at T2 and T3 for both males and females.
That is, there was no difference in factorial invariance between males and females on the
community violence exposure variable as it was measured at T2 and T3.
For the non-maltreated and maltreated groups, model fit for the unrestricted
model was good (χ2=27.74, df=14, CFI=.982, RMSEA=.071, 90% CI=.030 - .110).
Standardized loadings for the non-maltreated group ranged from .681 to .825 at T2 and
from .743 to .888 for T3. Standardized loadings for the maltreated group ranged from
.733 to .843 at T2 and from .491 to .847 for T3. The χ2 difference between the
unrestricted and the restricted models was not significant for (Δχ2=1.069, Δdf=2, p>.05).
These results provide evidence that the community violence exposure variable was
measured consistently at T2 and T3 for both non-maltreated and maltreated adolescents.
That is, there was no difference between non-maltreated and maltreated adolescents on
the community violence exposure variable as it was measured at T2 and T3.
For the non-maltreated, physically maltreated, and non-physically maltreated
groups, model fit for the unrestricted model was good (χ2=21.64, df=21, CFI=.999,
RMSEA=.015, 90% CI=.000 - .077). Standardized loadings for the non-maltreated group
46

ranged from .647 to .825 at T2 and from .743 to .883 for T3. Standardized loadings for
the physically maltreated group ranged from .660 to .756 at T2 and from .622 to .782 for
T3. Standardized loadings for the non-physically maltreated group ranged from .791 to
.885 at T2 and from .552 to .698 for T3 The χ2 difference between the unrestricted and
the restricted models was not significant for (Δχ2=5.25, Δdf=6, p>.05). These results
provide evidence that the community violence exposure variable was measured
consistently at T2 and T3 for non-maltreated, physically maltreated, and non-physically
maltreated adolescents. That is, there was no difference between the three groups on the
community violence exposure variable as it was measured at T2 and T3.
Determining Factor Structure for Adolescent Self-Report, Caregiver Report, and Teacher
Report of Adolescent’s Aggressive Behavior: Integrating Data from Multiple Informants
As with community violence exposure, the psychometric properties of
adolescents' aggressive behavior were also examined. However, there were several
differences in the procedure utilized. Unlike community violence exposure, adolescents'
aggressive behavior was assessed using multiple informants (adolescent, caregiver, and
teacher), thus efforts were made to determine the best way to integrate the data from the
respondents. A three step procedure was used. The steps and the results obtained at each
are outlined below.
Similarly to the procedure used to create parcels for the CVI items, the twenty
items composing the aggressive behavior subscale of the YSR were randomly assigned to
one of four parcels, thus each parcel contained five items. The same procedure was
conducted for the 20 items comprising the aggressive behavior subscales on the CBCL.
47

As for the 11 items of the TRS, they were randomly assigned to one of three parcels. The
parcels were identically configured at T2 and T3. These parcels were modeled as
manifest indicators of latent constructs representing adolescent-, caregiver-, and teacher-
reports of adolescent’s aggressive behavior.
A CFA was conducted to determine how well the parceled variables indicate their
respective latent variable. A three factor model was fit. The first factor, representing
adolescent self-report of aggressive behavior was indicated by the four parcels
constructed from the YSR data. The second latent variable, representing caregiver report
of adolescents' aggressive behavior, was indicated by the four parcels constructed from
the CBCL items. And the third latent variable, representing teacher report on adolescents'
aggressive behavior, was indicated by the three parcels constructed from the TRS. These
latent variables were set as oblique to one another. To specify the models the first
indicator of each latent variable was constrained to 1. Results of the model at T2
indicated good fit to the data (χ2=62.69, df=40, CFI=.991, RMSEA=.038, 90% CI=.018 -
.056). Loadings for the four parcels indicating adolescents' self-report ranged from .660
to .834. For the caregivers, the four parcels had loadings between .816 and .896. And
finally, teachers ranged from .932 to .964. Results for T3 were similar to the results at T2.
The model fit the data well (χ2=59.00, df=39, CFI=.990, RMSEA=.041, 90% CI=.016 -
.061). Loadings for the four parcels indicating adolescents' self-report ranged from .657
to .789. For the caregivers, the four parcels had loadings between .828 and .856. And
finally, teachers ranged from .904 to .958. Correlations between the three factors were
moderate at T2 (r
range
=.21 - .35) and T3 (r
range
=.23 - .31).
48

Figure 3. Oblique model examining multiple-informant report of adolescents’ aggressive
behavior at T2.

Figure 4. Oblique model examining multiple-informant report of adolescents’ aggressive
behavior at T3.

Results of the second order CFA examining whether the latent variables
representing adolescent, caregiver, and teacher report of adolescents’ aggressive behavior
indicate a higher-order latent variable of aggressive behavior indicated good fit to the
data (χ2=62.69, df=40, CFI=.991, RMSEA=.038, 90% CI=.018 - .056). Loadings for the
three latent variables representing adolescent-, caregiver-, and teacher-report on the
second-order latent variable, aggressive behavior, were .469, .657, and .613, respectively.
Similar results were found for the data at T3. Results of the second order CFA indicated
49

good fit to the data (χ2=59.00, df=39, CFI=.990, RMSEA=.041, 90% CI=.016 - .061).
Loadings for the three latent variables representing adolescent-, caregiver-, and teacher-
report on the second-order latent variable, aggressive behavior, were .507, .775, and .534,
respectively.

Figure 5. Second order factor model for aggression at T2.

Figure 6. Second order factor model for aggression at T3.

50

A three factor model incorporating adolescent-, caregiver-, and teacher-report as
the three manifest indicators of a latent variable, aggressive behavior, was conducted for
the T2 and T3 data. These models demonstrated perfect fit to the data for both T2 (χ2=0,
df=0, CFI=1.00, RMSEA=.000, 90% CI=.000 - .000) and T3 data (χ2=0, df=0, CFI=1.00,
RMSEA=.000, 90% CI =.000 - .000). Perfect fit was obtained because the models were
fully saturated (i.e., no degrees of freedom). Standardized loadings ranged from .502 to
.740 at T2 and from .505 to .814 at T3.

Figure 7. Latent variable for aggression at T2 indicated by manifest variables
representing adolescent, caregiver, and teacher report.

Figure 8. Latent variable for aggression at T3 indicated by manifest variables
representing adolescent, caregiver, and teacher report.

51

Invariance of aggression across groups. Evidence for invariance between genders
was obtained using multiple group structural equation modeling (MGSEM). First an
unconstrained model was tested in which factor loadings and intercepts were freely
estimated for each gender. The unconstrained models fit adequately for both T2 (χ2
=4.93, df=2, CFI=.985, RMSEA=.087, 90% CI=.000 - .188) and T3 (χ2 =.584, df=2,
CFI=1.00, RMSEA=.000, 90% CI=.000 - .111) data. Standardized loadings for females
ranged from .459 to .761 at T2 and from .584 to .777 at T3. Standardized loadings for
males ranged from .533 to .751 at T2 and from .411 to .894 at T3. Next, factor loadings
and intercepts were restricted to be equal across the groups. A nonsignificant χ2
difference between the unrestricted and the restricted models indicates invariance of
measurement. The χ2 difference was not significant for either the T2 (Δχ2=0.672, Δdf=2,
p>.05) or the T3 (Δχ2=3.802, Δdf=2, p>.05) data. These results support measurement
invariance for the aggressive behavior variable across gender at T2 and T3.
Evidence for invariance across the non-maltreated and maltreated groups was
obtained using the same procedure. The unconstrained models fit adequately for both T2
(χ2 =2.683, df=1, CFI=.996, RMSEA=.042, 90% CI=.000 - .155) and T3 (χ2 =1.736,
df=2, CFI=1.00, RMSEA=.000, 90% CI=.000 - .154) data. Standardized loadings for
non-maltreated group ranged from .421 to .888 at T2 and from .363 to .911 at T3.
Standardized loadings for the maltreated group ranged from .477 to .796 at T2 and from
52

.519 to .788 at T3. The χ2 difference the unrestricted and the restricted models was not
significant for either the T2 (Δχ2=0.979, Δdf=2, p>.05) or the T3 data (Δχ2=0.55, Δdf=2,
p>.05). These results support measurement invariance for the aggressive behavior
variable across the non-maltreated and maltreated groups at T2 and T3.
Invariance across the non-maltreated, physically maltreated, and non-physically
maltreated groups was examined. The unconstrained models fit well for both T2 (χ2
=2.739, df=4, CFI=1.00, RMSEA=.000, 90% CI=.000 - .112) and T3 (χ2 =2.876, df=5,
CFI=1.00, RMSEA=.000, 90% CI=.000 - .101) data. Standardized loadings for the non-
maltreated group ranged from .300 to .685 at T2 and from .412 to .845 at T3.
Standardized loadings for the physically maltreated group ranged from .515 to .816 at T2
and from .486 to .800 at T3. Standardized loadings for the non-physically maltreated
group ranged from .370 to .793 at T2 and from .519 to .775 at T3. The χ2 difference
between the unrestricted and the restricted models was not significant for either the T2
data (Δχ2=2.266, Δdf=3, p>.05) or the T3 data (Δχ2=1.064, Δdf=3, p>.05). These results
support measurement invariance for the aggression variable across the non-maltreated,
physically maltreated, and non-physically maltreated groups at T2 and T3.
Invariance of aggressive behavior across time. Longitudinal invariance was
assessed across the two time periods by comparing a model in which the parameters were
allowed to be freely estimated to a model in which the the factor loadings and intercepts
were held equal across time. For the full sample, model fit for the unrestricted model was
good (χ2=11.27, df=4, CFI=.990, RMSEA=.068, 90% CI=.023 - .117). Standardized
loadings ranged from .717 to .813 for T2 and from .679 to .854 for T3. The χ2 difference
53

between the unrestricted and the restricted models was not significant for (Δχ2=3.33,
Δdf=3, p>.05). These results provide evidence that among the full sample the aggressive
behavior variable was measured consistently at T2 and T3. That is, aggressive behavior
was measured the same at both time points.
For males and females, model fit for the unrestricted model was good (χ2=23.21,
df=14, CFI=.986, RMSEA=.058, 90% CI=.000 - .099). Standardized loadings for the
females ranged from .467 to .794 at T2 and from .704 to .841 for T3. Standardized
loadings for the males ranged from .779 to .819 at T2 and from .629 to .841 for T3. The
χ2 difference between the unrestricted and the restricted models was not significant for
(Δχ2=12.23, Δdf=7, p>.05). These results provide evidence that the aggressive behavior
variable was measured consistently at T2 and T3 for both males and females. That is
there was no difference between males and females on the aggression variable as it was
measured at T2 and T3.
For the non-maltreated and maltreated groups, model fit for the unrestricted
model was good (χ2=27.74, df=14, CFI=.982, RMSEA=.071, 90% CI=.030 - .110).
Standardized loadings for the non-maltreated group ranged from .681 to .825 at T2 and
from .743 to .888 for T3. Standardized loadings for the maltreated group ranged from
.733 to .843 at T2 and from .491 to .847 for T3. The χ2 difference between the
unrestricted and the restricted models was not significant for (Δχ2=1.069, Δdf=2, p>.05).
54

These results provide evidence that the aggressive behavior variable was measured
consistently at T2 and T3 for both the non-maltreated group and maltreated group. That
is, there was no difference between the groups on the aggression variable as it was
measured at T2 and T3.
For the non-maltreated, physically maltreated, and non-physically maltreated
groups, model fit for the unrestricted model was good (χ2=21.64, df=21, CFI=.999,
RMSEA=.015, 90% CI=.000 - .077). Standardized loadings for the non-maltreated group
ranged from .647 to .825 at T2 and from .743 to .883 for T3. Standardized loadings for
the physically maltreated group ranged from .660 to .756 at T2 and from .622 to .782 for
T3. Standardized loadings for the non-physically maltreated group ranged from .791 to
.885 at T2 and from .552 to .698 for T3 The χ2 difference between the unrestricted and
the restricted models was not significant for (Δχ2=5.25, Δdf=6, p>.05). These results
provide evidence that the aggressive behavior variable was measured consistently at T2
and T3 for both the non-maltreated group, physically maltreated, and non-physically
maltreated groups. That is, there was no difference between the groups on the aggression
variable as it was measured at T2 and T3.

55

Substantive Analyses
As with the preliminary analyses, the substantive analyses were conducted using
structural SEM in Mplus (version 5.21: Muthen & Muthen, 2009). The same fit indices
described in the previous sections were used to evaluate the fit of the models to the data.
To evaluate the fit of nested models, χ2 difference tests were used. To account for
missing data due to attrition between T2 and T3 and to maximize the data to be included
in the structural models, a FIML estimator was used.
Specific Aim 1: Using cross-lagged regression models determine the longitudinal
relationships between community violence exposure and aggressive behavior in a sample
of maltreated and non-maltreated female and male adolescents. In cross-lagged
regression models, distinct constructs are expected to influence themselves over time, and
to exert an influence on another construct over time. The autoregressive parameters
indicate the degree of longitudinal stability and coefficients from the crossed path
indicate that one common factor may exert an impact on the other. A significant crossed
path suggests that one factor can explain variance above and beyond the variance
explained by the autoregressive effect. Community violence exposure and aggressive
behavior were modeled following the guidelines of Martens and Haase (2006), with
correlated error terms for manifest indicators and disturbance terms for the endogenous
variables. Longitudinal and group invariance constraints are also applied, as determined
by psychometric model testing. Following traditional cross-lagged analysis methodology,
56

a fully cross-lagged model is compared to a model that included only autoregressive
paths. The decrement in model fit (as assesssed by change in
2
) is examined. Significant
decrement in fit from the fully cross-lagged model to the autoregressive only model
indicates that the fully cross-lagged model accounts for the data better than the
autoregressive model.
Utilizing the measurement models described above, latent variable cross-lagged
analyses will be carried out to examine the longitudinal relationships of community
violence exposure and aggressive behavior in the full sample of adolescents. Longitudinal
invariance constraints were imposed on the models. Common risk factors, including
adolescent’s delinquency, age, and family SES were used as covariates. Additionally, the
variables that predicted attrition at T2 or T3 were used as covariates (i.e., ethnicity,
poverty status, maltreatment status, and living arrangement). These variables were
modeled with non-zero covariances. Community violence exposure and aggressive
behavior at each of the time points were regressed on the covariates.
The fully cross-lagged model fit the data well (χ2=129.38, df=79, CFI=.982,
RMSEA=.040, 90% CI=.027 - .053) and was a significant improvement over the
autoregressive paths only model (Δχ2=13.16, Δdf=2, p<.01), indicating that the data is
better accounted for by the fully-cross-lagged model The auto-regressive paths of the
fully cross-lagged model were significant and strong (community violence exposure:
=.857, p<.01; aggressive behavior: =.791, p<.01). The cross-lagged path regressing T3
57

aggressive behavior on T2 community violence exposure was non-significant ( =.093,
p=.168). However, the cross-lagged path regressing T3 community violence exposure on
T2 aggressive behavior was significant ( =.114, p<.01). Results are illustrated in Figure
9 and Table 7.

Figure 9. Fully cross-lagged model addressing Specific Aim 1.

58

Table 7. Results of the structural model addressing Specific Aim 1.

Specific Aim 2: Using multiple group cross-lagged regression models determine if
the longitudinal relationships between community violence exposure and aggressive
behavior are different for male and female adolescents. To examine gender differences in
these relationships MGSEM was used. Longitudinal invariance constraints will be
imposed on the models. As with the analysis used to address Specific Aim 1, a fully
cross-lagged model was compared to an autoregressive paths only model to determine
whether the fully cross-lagged model best accounts for the data. The model was
conducted simultaneously for males and females. First, an unconstrained model was run
for both groups, this model provided a basis from which to compare to more restricted
models, with measurement and structural constraints. Second, the factor loadings and
intercepts were constrained to be equal across groups as specified by the measurement
59

model analyses. The change in the χ2 from the unconstrained model to the measurement
constrained model provides evidence as to whether measurement parameters across
groups were significantly different from each other. A non-significant difference in the χ2
indicates that the variables are measured similarly across groups and time. Third, the
measurement constrained model was compared to a model that imposes structural
constraints. And finally, if χ2 difference tests indicated a significant worsening of fit
between the measurement constrained model and the structural constrained model,
parameters were explored in turn to determine whether group affiliation moderated the
relationships between the variables.
The unrestricted fully cross-lagged model fit the data well (χ2=225.71, df=167,
CFI=.978, RMSEA=.043, 90% CI=.027 - .056), and was a significant improvement over
the autoregressive paths only model (Δχ2=17.47, Δdf=4, p>.05), indicating that the data
is better accounted for by the fully-cross-lagged model. The autoregressive paths of the
fully cross-lagged model were significant and strong for both males (community violence
exposure: =.870, p<.01; aggressive behavior: =.768, p<.01) and females (community
violence exposure: =.802, p<.01; aggressive behavior: =.747, p<.01). The cross-lagged
path regressing T3 aggressive behavior on T2 community violence exposure was non-
significant for the females ( =.003, p=.972) but significant for males ( =.192, p<.01).
Additionally, the cross-lagged path regressing T3 community violence exposure on T2
aggressive behavior was only marginally significant for females ( =.104, p<.10) but
significant for the males ( =.136, p<.01). Results are illustrated in Table 8 and Figure 10.
60

Constraining the factor loadings and intercepts to be equal across groups did not
result in a significant decrement in overall fit (Δχ2=9.67, Δdf=6, p<.05), but there was a
significant decrement when structural constraints were imposed (Δχ2=6.57, Δdf=3,
p>.05). These results indicate that gender moderated at least one of the parameters. Each
of the structural parameters were examined to determine which were moderated by
maltreatment experience. A significant decrement in the model constraining a specific
parameter to be equal across groups indicated an interaction effect. The χ2 difference test
showed that the cross-lagged parameter between early aggressive behavior and later
exposure to community violence was not significantly different for females and the males
(Δχ2=4.61, Δdf=1, p>.05). Thus, this parameter was not moderated by gender. The
parameter between early community violence exposure and later aggressive behavior was
significantly different between the females and males (Δχ2=.205, Δdf=1, p<.05). Results
of the χ2 difference tests indicate that the cross-lagged path between early community
violence exposure and later aggressive behavior is moderated by gender.
As with the full sample, the best predictor of later aggressive behavior for males
and females is early aggressive behavior. The same is true of exposure to community
violence. If adolescents are exposed to community violence early on, it places them at
risk for more exposure later on regardless of their gender. Early exposure to community
violence did not predict aggressive behavior later on for females, but it did predict it for
males. Additionally, early aggressive behavior places both males and females at risk for
later exposure to community violence and aggressive behavior.

61

Table 8. Results of the structural model addressing Specific Aim 2.

Figure 10. Fully cross-lagged MGSEM examining Specific Aim 2.

62

Specific Aim 3: Using multiple group cross-lagged regression models of
community violence exposure and aggressive behavior determine if the longitudinal
relationships between community violence exposure and aggressive behavior are
different for maltreated and non-maltreated adolescents. MGSEM procedures similar to
those described in the previous aim were used to address this aim. The unconstrained
fully cross-lagged model fit the data well (χ2=210.58, df=150, CFI=.978, RMSEA=.046,
90% CI=.030 - .059) and was a significant improvement over the autoregressive paths
only model (Δχ2=22.79, Δdf=4, p<.05), indicating that the data is better accounted for by
the fully-cross-lagged model. The autoregressive paths of the fully cross-lagged model
were significant and strong for both the non-maltreated group (community violence
exposure: =.842, p<.01; aggressive behavior: =.492, p<.01) and the maltreated group
(community violence exposure: =.913, p<.01; aggressive behavior: =.789, p<.01). The
cross-lagged path regressing T3 aggressive behavior on T2 community violence exposure
was non-significant for the maltreated group ( =.069, p>.05) but significant for the non-
maltreated group ( =.223, p<.05). Additionally, the cross-lagged path regressing T3
community violence exposure on T2 aggressive behavior was only marginally significant
for the maltreated group ( =.076, p<.10) but significant for the non-maltreated group
( =.217, p<.01). Results are illustrated in Table 9 and Figure 11.
Constraining the factor loadings and intercepts to be equal across groups did not
result in a significant decrement in overall fit (Δχ2=9.63, Δdf=8, p>.05) but imposing
structural constraints did significantly worsen in fit from the model with measurement
constraints (Δχ2=41.65, Δdf=4, p<.01). These results indicate that maltreatment history
63

moderated at least one of the parameters. Each of the structural parameters were
examined to determine which were moderated by maltreatment experience. A significant
decrement in the model constraining a specific parameter to be equal across groups
indicated an interaction effect. The χ2 difference test showed that the cross-lagged
parameter between early aggressive behavior and later exposure to community violence
was significantly different for the non-maltreated and the maltreated group (Δχ2=8.68,
Δdf=1, p<.05. The parameter between early community violence exposure and later
aggressive behavior was also significantly different between the non-maltreated and
maltreated groups (Δχ2=9.33, Δdf=1, p<.01). The results of these χ2 difference tests
indicate that maltreatment status moderated both of the cross-lagged relationships.
As with the analyses presented in the previous aims, early exposure to community
violence was the best predictor for future exposure. The same was true for early and later
aggressive behavior. Early exposure to community violence was not a predictor of later
aggressive behavior for adolescents with a history of maltreatment. However, early
exposure was a predictor of later aggressive behavior for adolescents without such a
history. Additionally, early aggressive behavior was a predictor for later exposure to
community violence for both the maltreated and non-maltreated groups, although the
relationship was weaker for the maltreated group.

64

Table 9. Results of the structural model addressing Specific Aim 3.

65

Figure 11. Fully cross-lagged MGSEM examining Specific Aim 3.

Specific Aim 4: Using multiple group cross-lagged regression models of
community violence exposure and aggressive behavior determine if the relationship
between community violence exposure and aggressive behavior is moderated by specific
maltreatment type (non-maltreated vs. physically maltreated vs. non-physically
maltreated). MGSEM procedures similar to those described in the previous two aims
were used to address this aim. The unconstrained fully cross-lagged model fit the data
well (χ2=283.60, df=228, CFI=.979, RMSEA=.043, 90% CI=.024 - .059), and was a
significant improvement over the autoregressive paths only model (Δχ2=36.83, Δdf=6,
p<.01), indicating that the data is better accounted for by the fully-cross-lagged model.
The autoregressive paths were significant and strong for non-maltreated (community
violence exposure: =.801, p<.01; aggressive behavior: =.826, p<.01), physically
maltreated (community violence exposure: =.855, p<.01; aggressive behavior: =.825,
p<.01), and non-physically maltreated (community violence exposure: =.934, p<.01;
69
66

aggressive behavior: =.723, p<.01) groups. The cross-lagged path regressing T3
aggressive behavior on T2 community violence exposure was non-significant for the
physically maltreated group ( =.174, p=.122) but was marginally significant for the non-
matreated group ( =.237, p<.10) and significant for the non-physically maltreated group
( =.300, p<.05). Additionally, the cross-lagged path regressing T3 community violence
exposure on T2 aggressive behavior was not significant for the physically maltreated
group ( =.054, p=.394), but was significant for the non-maltreated group ( =.262, p<.01)
and marginally significant for the non-physically maltreated group ( =.112, p<.10).
Results are illustrated in Table 10 and Figure 12.
Constraining the factor loadings and intercepts to be equal across groups did not
result in a significant decrement in overall fit (Δχ2=9.14, Δdf=6, p>.05) but imposing
structural constraints did significantly worsen in fit from the model with measurement
constraints (Δχ2=62.24, Δdf=8, p>.05). These results indicate that maltreatment
experience moderated at least one of the structural parameters. The χ2 difference test
showed that the cross-lagged parameter between early aggressive behavior and later
exposure to community violence was significantly different for the non-maltreated,
physically maltreated, and non-physically maltreated adolescents (Δχ2=12.28, Δdf=2,
p<.05). The parameter between early community violence exposure and later aggressive
behavior was also significantly different between the non-maltreated, physically
maltreated, and non-physically maltreated groups (Δχ2=6.08, Δdf=2, p<.05). Results of
the χ2 difference tests indicated that maltreatment type moderated both cross-lagged
paths.
67

These results indicated that for all three groups, early exposure to community
violence is the strongest predictor for later exposure. The same holds true for the
relationship between early and later aggressive behavior. Early exposure to community
violence predicted later aggressive behavior for both adolescents in the non-maltreated
group and the non-physically maltreated group. Early exposure, however, was not a
predictor for later aggressive behavior for adolescents with a history of physical
maltreatment. Additionally, early aggressive behavior was predictive of later exposure to
community violence for both the non-maltreated and non-physically maltreated groups,
although the association was weaker for the non-physically maltreated group. Early
aggression was not a predictor for later exposure for the physically maltreated group.

Table 10. Results of the structural model addressing Specific Aim 4.

68

Figure 12. Fully cross-lagged MGSEM examining Specific Aim 4.
69

CHAPTER FOUR: DISCUSSION
The Current Study
The current investigation sought to elucidate the longitudinal relationships
between community violence exposure and aggressive behavior in a sample of maltreated
and non-maltreated young adolescents. Of primary interest was examining the effect of
adolescents’ early community violence exposure and aggressive behavior on their later
community violence exposure and aggressive behavior as well as examining how these
relationships differ by gender and maltreatment experiences. The present study was
conducted in the context of a larger longitudinal study examining the impacts of child
maltreatment on adolescent development. The data used in the current investigation were
drawn primarily from the T2 and T3 data collection waves of the parent study. These two
time points were separated by 18 months. To examine the relationships between the
variables at T2 and T3, latent variable cross-lagged panel analysis methodology was
employed.
Longitudinal Relationship of Community Violence and Aggressive Behavior
The first aim of this study was to examine the longitudinal relationships between
community violence and aggressive behavior among the full sample of adolescents. The
results of the analysis revealed that community violence exposure and aggressive
behaviors at T3 were strongly predicted by community violence exposure and aggressive
behaviors at T2, respectively. The magnitude of the relationship between aggressive
behavior at T2 and T3 was to be expected; the extant literature is replete with studies
demonstrating that aggressive behavior is stable throughout childhood and adolescence
70

(Huesmann, 1984; Kokko & Pulkkinen, 2005; Olweus, 1979). The strength of the
relationship between community violence exposure at T2 and T3 also provides evidence
for the stability of exposure across time, but this finding is contrary to the literature
examining community violence exposure (Margolin, Vickerman, Ramos, Duman-
Serrano, Gordis, Iturralde, Oliver, & Spies, 2009). There exists very little evidence to
make firm conclusions as to the stability of community violence exposure over time. This
scarcity of evidence is due in large part to the lack of longitudinal designs also the fact
that studies that have employed such designs have had inconsistent findings. While some
studies have found exposure to be stable over time, others have not (Margolin et al.,
2009). In examining exposure across time points in the current sample, 56% of the
adolescents who reported exposure at T2 also reported exposure at T3. This is much
higher than the 10% of participants in the Margolin et al. (2009) study which examined
exposure over three time waves.
In examining the relationship between community violence exposure at T2 to
aggressive behavior at T3, there was no evidence demonstrating that exposure to
community violence predicted aggressive behavior. This finding is contrary to what was
expected. The majority of studies examining the association between the two variables
have found a significant relationship (Durant et al., 1994; Farrell & Bruce, 1997;
Gorman-Smith & Tolan, 1998; Schwab-Stone et al., 1995). Thus it was to be expected
that the Exposure → Behavior model would be supported. In an empirical review of the
literature, published between 1993 and 2006, a large effect size was detected between the
community violence exposure and aggressive behavior (Stevens, 2011).
71

While the bulk of the literature supports a strong and significant relationship
between early community violence exposure and later aggressive behavior, it should be
noted that many of the studies comprising this canon are fraught with methodological
problems that might lead to inflated strengths in the relationships. The foremost of these
issues is that the majority of the studies have been cross-sectional in design. If aggressive
behavior is measured around the same time that the community violence occurred, it
becomes difficult to disentangle how these variables relate to one another and how they
exert influence on the other (Margolin & Gordis, 2001). In other words, it is not possible
to determine which of the constructs directly leads to the other or whether there is bi-
directionality in this association. In fact, most of the research and theory focusing on
community violence exposure have viewed it as influencing aggressive behaviors, very
little attention has been given to whether aggressive behavior and its correlates place
individuals at risk for being exposed to violence. To truly determine how these variables
relate to one another they must be examined over time. However, only a few studies have
employed longitudinal designs. In utilizing cross-lagged panel analysis methodology, the
current study was able to examine how community violence exposure and aggressive
behavior relate to one another over time and provides fairly robust evidence for how
these variables relate to one another in the current sample of adolescents.
The finding that aggressive behavior at T2 predicts community violence exposure
at T3 supports the literature that has demonstrated early aggressive behavior as a risk
factor for later exposure to community violence. These results lend support to the
Behavior → Exposure model. As discussed previously, only a few studies have reported
72

similar results (Boyd et al., 2003; Mrug & Windle, 2009; Farrell & Bruce, 1997; Farrell
& Sullivan, 2004). For instance, Boyd and colleagues (2003) found that children who
were aggressive in the first grade were more likely than non-aggressive first graders to be
exposed to community violence during the middle school years. Similarly, Mrug and
Windle (2009) found that young adolescents who exhibited early externalizing problems
that included aggressive behaviors and conduct problems were more likely to be exposed
to community violence two years later. While there are not very many studies that have
examined the influence of aggressive behavior on community violence exposure,
researchers examining other forms of violence exposure have found that early aggressive
behavior serves as a risk factor for being victimized later on (e.g., Schwartz et al, 2005).
Gender as a Moderator
The second aim of this paper was to examine whether the longitudinal
relationships between community violence exposure and aggressive behavior differed
between males and females. A multiple group cross-lagged panel analysis was used to
examine how the relationships between the variables differed. As with the full sample,
the best predictor of aggressive behavior at T3, for both males and females, was
aggressive behavior at T2. The same was true of exposure to community violence;
exposure at T2 strongly predicted exposure at T3. The strength of these associations did
not differ significantly between the male and the female groups. The temporal stability of
aggressive behavior for both male and female adolescents has been well documented in
the literature (Kokko & Pulkkinen, 2005). As stated previously, there is not much support
in the literature for community violence exposure being stable over time, especially
73

whether stability differs by gender. In the current sample of adolescents it appears that
exposure is stable. As discussed previously, 63% of males exposed at T2 were also
exposed at T3, 49% for females.
In examining the relationship between community violence exposure at T2 and
aggressive behavior at T3, exposure to community violence did not predict aggressive
behavior for the female group, but did predict aggressive behavior for the male group.
That is, males who reported higher levels of exposure to community violence at T2 were
reported as demonstrating more aggressive behaviors at T3. Surprisingly, there is a dearth
of research examining gender as a moderator in the relationship between community
violence exposure and aggressive behavior. The findings of the studies that have
examined gender differences have been inconsistent and have made it difficult to draw
firm conclusions regarding gender as a moderator. While some studies have found that
males become more aggressive as a result of exposure (Mrug & Windle, 2009; O’Keefe,
1997), others have found that females become more aggressive (Farrell & Bruce, 1997;
Farrell & Sullivan, 2004). For instance, Farrell and Bruce (1997) found that there was no
significant relationship between early community violence exposure and later aggressive
behavior for males, but did find one for females. That is, females in their study exhibited
aggressive and violent behaviors as a result of being exposed to community violence.
However, O’Keefe (1997) found that community violence exposure influences aggressive
behavior for boys but not for girls. More recently, Mrug and Windle (2009) found that
males are more likely to experience externalizing problems, such as aggression and
conduct behaviors, if they are exposed to community violence. No such relationship was
74

found for females. A substantive difference between these studies is that Farrell and
Bruce (1997) evaluated gender differences by conducting separate analyses for males and
females, rather than testing for differences between genders using statistical methods.
The analytical approach used in the current study and that used by Mrug and Windle
(2009) utilized multiple group SEM which is more appropriate and statistically precise
for directly evaluating the presence of subgroup differences, and as such it provides
robust evidence that the obtained relationships apply equally well to young adolescent
males and females. In the current sample, the results provide support for the Exposure →
Behavior model for males, but not for the females.
Examination of the other association between T2 aggressive behavior and T3
community violence exposure indicated that aggressive behavior at T2 strongly predicts
exposure to community violence at T3 for both the male and the female groups. That is,
early aggressive behavior places adolescents, regardless of their gender, at risk for being
exposed to community violence in the future. Only a few studies have reported similar
results (Boyd et al., 2003; Farrell & Bruce, 1997; Mrug & Windle, 2009). Farrell and
Bruce (1997) found only an effect for females. That is, in their sample females who were
aggressive early on were more likely than non-aggressive females to become exposed to
community violence. Like the current study, Mrug and Windle (2009) did not find any
gender differences in this association. Rather they found that aggressive behavior
predicted later exposure to community violence regardless of gender. The current study
and Mrug and Windle employed more robust statistical techniques to examine gender
differences thus there is support for the obtained relationships applying equally well for
75

males and females in the current study. As with the results of the full model, these results
provide support for the Behavior → Exposure model. However, the relationship is more
complicated for males. It seems that there is bi-directionality among community violence
exposure and aggressive behavior. That is, they influence each other.
Maltreatment Experience as a Moderator
The third aim of this study was to examine whether the longitudinal relationship
between community violence exposure and aggressive behavior differed depending on
maltreatment experience. The results of the analyses indicated that community violence
exposure and aggressive behavior at T3 were strongly predicted by community violence
exposure and aggressive behavior at T2, respectively, for adolescents with a history of
child maltreatment and for those without such a history. As with the previous analyses,
this model provides evidence that aggressive behavior and community violence exposure
are stable across time for adolescents with a history of child maltreatment and for those
without a history. In the current study, 63% of the maltreated adolescents exposed at T2
were also exposed at T3, compared to 46% with the non-maltreated group.
In examining the relationship between T2 community violence exposure to T3
aggressive behavior, community violence exposure did not predict aggressive behavior
for adolescents in the maltreated group, but did predict aggressive behavior for the non-
maltreated adolescents. Support is provided for the Exposure → Behavior model for the
non-maltreated group, but not for the maltreated group. As stated previously, this is the
first study to examine maltreatment history as a moderator in the association between
community violence exposure and aggressive behavior, thus there is not much in the
76

literature to compare these findings. Maltreated adolescents have other reasons to be
aggressive that are not associated with community violence exposure. It could very well
be the characteristics of the maltreatment that make them more aggressive. This is
supported by the findings in the literature and the current study that maltreated
adolescents are more aggressive than non-maltreated adolescents (Trickett & McBride-
Chang, 1995). These results could be interpreted in light of the social push hypothesis
(Raine & Venables, 1984). Adolescents who experienced early child maltreatment were
“pushed” by that experience towards aggressive behavior. Aggression for the maltreated
adolescents is tied up in their experience of maltreatment. Thus, the addition of another
form of violence exposure has no or little influence on their aggressive behavior. For
adolescents without such a history, it is much more likely that their aggressive behavior
will be pushed by community violence exposure.
Aggressive behavior at T2 was a predictor for exposure to community violence at
T3 regardless of maltreatment experience. Again, early aggressive behavior is a risk
factor for later community violence exposure regardless of whether adolescents had a
history of child maltreatment or not. The results provide support for the Behavior →
Exposure model for adolescents, regardless of their maltreatment history. However, for
the non-maltreated group, it is more complicated. Like males, it appears that there is bi-
directionality in the relationship. Thus, for the non-maltreated group, community
violence exposure and aggressive behavior exert influence on one another. As stated
previously, this is the first study to examine maltreatment experience as a moderator in
the relationship; much more work needs to be done to truly examine these relationships.
77

Specific Maltreatment Type as a Moderator
The final aim of the study was to examine whether the longitudinal relationship
between community violence exposure and aggressive behavior was moderated by
specific maltreatment type. In other words, are the relationships different for adolescents
who have been physically maltreated, non-physically maltreated, and non-maltreated? As
Trickett and McBride-Chang (1995) point out, there is a need in the field to determine
whether there are differential outcomes for specific forms of maltreatment. The majority
of the extant literature have simply categorized children and adolescents as either
“maltreated” or “non-maltreated” and then drawn conclusions as to the impacts of having
that experience. This results in the loss of crucial information about how the various
types of child maltreatment might differentially impact development. The MCRAI
system employed by the YAP provided the opportunity to examine the relationships
among these groups and examine whether there are differential outcomes. The results
indicated that for all three groups, community violence exposure at T2 is the strongest
predictor for exposure at T3. The same holds true for the relationship between aggressive
behavior at T2 and T3. These results also provide evidence for the stability of aggressive
behavior over time for physical maltreated and non-physically maltreatedadolescents.
Evidence is also provided for the stability of community violence over time for the
physically maltreatedand non-physically maltreated groups.
Again, for groups without a history of physical victimization, it seems that early
exposure to community violence predicts aggressive behavior. However, for those who
were physically maltreatedare not at elevated risk for aggressive behavior based on early
78

exposure to community violence. This provides support for the Exposure → Behavior
model for the non-physically maltreated and non-maltreated groups. It should be noted
that the adolescents in the physically maltreated group had higher levels of aggressive
behavior at both T2 and T3 than did the non-physically maltreated and non-maltreated
groups. So while there was no relationship for the physically maltreated group, it seems
that there is something else that accounts for their high levels of aggressive behavior. As
discussed above, in regards to the maltreated group as a whole, it could be that there is no
relationship between early community violence exposure and later aggressive behavior
for the physically maltreated group because there is something else driving their
aggression that is tied up in their experiences with physical maltreatment. In other words,
physical maltreatment “pushed” the individual to develop aggressive behaviors and the
addition of another form of violence does not have an impact on them. Thus, physical
maltreatment predicts aggressive behavior above and beyond community violence
exposure. This result is interesting in that there seems to be differential impacts of
community violence exposure on aggressive behavior depending on the specific type of
maltreatment that the adolescents experienced.
Aggressive behavior at T2 was predictive of exposure to community violence at
T3 for both the non-maltreated and non-physically maltreated groups. Aggressive
behavior at T2 was not a predictor for exposure at T3 for the physically maltreated group.
Again, early aggressive behavior serves as a risk factor for later violence exposure, but
only for adolescents with a history of non-physical maltreatment and for adolescents in
the non-maltreated group. Early aggressive behavior does not seem to be a risk factor for
79

the adolescents with a history of physical maltreatment. These results provide support for
the Behavior → Exposure model for the non-physically maltreated and non-maltreated
groups. However, early community violence exposure influenced later aggression for
both of these groups. So, the relationships among the community violence exposure and
aggressive behavior are bi-directional for the non-maltreated and non-physically
maltreated adolescents.
Theoretical Considerations
The traditional approach over the last twenty years has been to examine the
degree to which community violence exposure influences aggressive behavior. However,
the results of the current study point less at an Exposure → Behavior model and more at a
Behavior → Exposure model. While a couple of the groups had increased risk for
aggressive behavior based on their exposure to community violence, overwhelmingly the
results suggest that participants, regardless of their group affiliation, were at greater risk
for later community violence exposure based on their earlier aggressive behavior.
Although there is a growing body of research demonstrating this, much more research
effort needs to be made.
Most of the psychological research and theory has focused on the causal
implications of community violence exposure and has not given much thought to how
else community violence exposure and aggressive behavior might relate to one another.
In particular, researchers have relied on social learning theory (Bandura, 1986) to guide
their investigations. Sociological researchers have long thought that community violence
exposure and aggressive behavior are closely related to one another, and that it is often
80

aggressive and delinquent behavior that place children and adolescents at risk for being
victimized. The lifestyle exposure theory (Jenna & Brownsfield, 1997) posits that violent
offending and other forms of antisocial behavior are indicators of a lifestyle that places
individuals at risk for violent victimization. While this theory seems plausible for helping
to explain the results of the current study, it is not that clear cut. It seems that community
violence exposure and aggressive behavior relate to each other in a much more
complicated manner. A great deal of bi-directionality was observed indicating that
community violence exposure and aggressive behavior influence one another. It seems
that aggressive adolescents may be exposed to community violence in ways that cannot
be distinguished from their own aggressive behavior. While both social learning theory
and lifestyle exposure theory are mostly right at explaining the associations between
community violence exposure and aggressive behavior, they are also partly wrong.
Perhaps, the violence involvement model put forth by Halliday-Boykins and Graham
(2001) is more appropriate to help understand the relationships. This model holds that
violence exposure and aggressive behavior are associated because they are both
manifestations of general involvement in violence (Halliday-Boykins & Graham, 2001).
In a study, Halliday-Boykins and Graham (2001) compared four alternative models:
community violence exposure influences violent behavior; violent behavior influences
community violence exposure; both are consequences of common antecedents; and both
are manifestations of the same higher order construct, violence involvement. The findings
81

of their study provided support for the violence involvement model. This highlights the
need for more attention to these issues. It also highlights the importance of theory in
guiding research and unifying the field.
Strengths and Limitations of the Current Study
Strengths. This study contributes to the existing literature in several significant
ways. First, the methodology employed in the current study specifically addressed several
of the methodological issues that mar both the community violence exposure and child
maltreatment literature. Second, this study used cross-lagged analysis methodology with
latent variables to examine the longitudinal relationships between community violence
exposure and aggressive behavior. This analytical design permitted for the examination
of whether community violence exposure contributes to aggressive behavior, aggressive
behavior contributes to community violence exposure, or whether community violence
exposure and aggressive behavior are reciprocally associated with one another. Third, the
demographic composition of the sample allowed for the examination of whether there
were gender differences in the longitudinal relationships. There is a scarcity of literature
examining gender differences in this area, thus the results contribute significantly. And
finally, the composition of the sample and the MCRAI methodology allowed for the
examination of whether there were differences in the longitudinal relationships based on
maltreatment history, and also whether these relationships differed depending on the
specific type of maltreatment. This is the first study to examine maltreatment status and
type as a moderator in the longitudinal relationship between community violence
exposure and aggressive behavior.
82

Limitations. The results of this study must be considered in light of several
limitations. First, this study used only adolescent self-report to assess community
violence exposure at both time points. This is problematic in that the measure may have
tapped into cognitive and perceptual biases rather than actual experience. Adolescents
may have been responding to schemas about their neighborhood being a dangerous place
rather than what they have actual been exposed to. Report of adolescent violence
exposure was also collected from caregivers, but was not used in the current study
because of findings in the literature that caregivers’ reports correlate very weakly to
adolescents’ reports. One way to deal with this issue would be to include data about the
adolescent’s neighborhood from other sources such as official crime records. Problem
with this is that crime data is aggregated at a level that does not often match up to the
area that the adolescent is reporting on. Also not all crime is reported and access to the
statistics is not often permitted.
Secondly, the current study was limited in its examination of contextual factors.
While the study examined gender, maltreatment history, and specific maltreatment type
as moderators, other factors need to be examined in order to determine how they might
impact the relationships. One factor that that was not examined in the current study and
which has received little attention in the literature is whether ethnicity moderates the
relationships. Very few studies have examined this (Mrug & Windle, 2009), but
increasing the knowledge base in important because ethnic minorities live at
disproportional rates in violent and crime-ridden urban areas. In addition, attention was
not given to the underlying mechanisms in the relationships between community violence
83

exposure and aggressive behavior. In particular, delinquent peer affiliation, aggressive
cognitive styles, and family attributes.
And finally, even though the present study sought to build upon the previous
research and examine the possible differential outcomes based on the specific type of
maltreatment that adolescents experienced, it must be noted that these categories were not
clearly delineated. That is, there was a great deal of shared experiences across the
categories. For instance, adolescents in the physically maltreated group were not only
physically maltreated, they also experienced other forms of maltreatment, such as sexual
abuse, neglect, and emotional abuse. Additionally, the adolescents in the non-physically
maltreated group also varied in their experiences; they included all maltreatment types
that were not physical in nature. While this is a step forward from how maltreatment is
examined in the literature, caution should be used in interpreting the results because the
categories were not necessarily pure.
Future Directions and Clinical Implications
Building off the strengths and limitations of the current study, future research
efforts should focus further understanding the longitudinal relationship between
community violence exposure and aggressive behavior. While this study employed a
longitudinal design, the actual time period was relatively short. Future research should
examine the relationship over a more extended period of time and with more
measurement points. Additionally, these longitudinal studies should take into account the
possibility of aggressive behavior contributing to violence exposure and/or the bi-
directionality of influence. Research should focus on better understanding the differences
84

among groups in the relationships between community violence exposure and aggressive
behavior. As detailed in the current study, there is a small amount of research that has
examined group differences. While the current study adds to this knowledge-base, much
more work needs to be done to examine the differential relationships based on group
membership. Additionally, more effort should be devoted to exploring the underlying
mechanisms through which the relationships operate. For example, better understanding
how deviant peer affiliation, aggressive coping styles, and family attributes play a role in
mediating the association between community violence exposure and aggressive
behavior.
The results of this study have important implications for the development and
implementation of clinical interventions for children and adolescents. For the most part,
psychologists have relied primarily on an Exposure → Behavior framework to guide how
they view the relationship of exposure and aggressive behavior. The results of this study
point to the need to also focus attention on the development of interventions that address
the Behavior → Exposure perspective. Additionally, more work needs to be done to
examine the differential impacts of early community violence exposure and aggressive
behavior on later exposure and aggression for various subgroups, so that interventions
could be tailored to best help those impacted by violence.
85

REFERENCES
Achenbach, T. M. (1991a). Manual for the Child Behavior Checklist and 1990 Profile.
Burlington: University of Vermont, Department of Psychiatry.

Achenbach, T. M. (1991b). Manual for the Youth Self-Report and 1991 Profile.
Burlington: University of Vermont, Department of Psychiatry.

Achenbach, T.M., McConaughy, S.H. & Howell, C.T. (1987). Child/Adolescent
Behavioral and Emotional Problems: Implications of Cross-Informant Correlations
for Situational Specificity. Psychological Bulletin, 191, 213-232.

American Psychological Society (1997).

Bandura, A. (1986). Social foundations of thought and action: A social-cognitive theory.
Engle-wood Cliffs, NJ: Prentice-Hall.

Bell, C. C. & Jenkins, E. J. (1993). Community violence and children on Chicago’s
Southside. Psychiatry: Interpersonal and Biological Processes. Special Issue.
Children and Violence, 56, 46-54.

Bolger, K. E. & Patterson, C. J. (2001). Developmental pathways from child
maltreatment to peer rejection. Child Development, 72, 549 – 568.

Boyd, B. C., Cooley, M. R., Lambert, S. F., & Ialongo, N. S. (2003). First-grade child
risk behaviors for community violence exposure in middle school. Journal of
Community Psychology, 31, 297-314.

Bronfenbrenner, U. (1977). Toward an experimental ecology of human development.
American Psychologist, 32, 513-531.

Bronfenbrenner, U. (1989). Ecological systems theory. In R. Vasta (Ed.), Annals of child
development – Six theories of child development: Revised formulations and current
issues (pp. 187-250). Greenwich, CT: JAI Press.

Bronfenbrenner, U. (1986). Ecology of the family as a context for human development:
Research perspectives. Developmental Psychology, 22, 723-742.

Brooks-Gunn, J., Duncan, G.J., & Aber, J.L. (Eds.). (1997a). Neighborhood poverty: Vol.
1. Contexts and consequences for children. New York: Russell Sage Foundation.

Brooks-Gunn, J., Duncan, G.J., & Aber, J.L. (Eds.). (1997b). Neighborhood poverty: Vol.
2. Policy implications in studying neighborhoods. New York: Russell Sage
Foundation.
86

Brooks-Gunn, J., Duncan, G.J., Leventhal, T., & Aber, J.L. (1997). Lessons learned and
future directions for research on the neighborhoods in which children live. In J.
Brooks-Gunn, G. Duncan, & J. L. Aber (Eds.), Neighborhood poverty: Vol. 1.
Contexts and consequences for children (pp. 279-298). New York: Russell Sage
Foundation.

Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K. A.
Bollen & J. S. Long (Eds.), Testing structural equation models (pp. 136-162).
Newbury Park, CA: Sage.

Byrne, B.M. & Stewart, S.M. (2006). The MACS Approach to Testing for Multigroup
Invariance of a Second-Order Structure: A Walk through the Process. Structural
Equation Modeling, 13, 287-321.

Calvert, W. J. (1999). Integrated literature on effects of exposure to violence upon
adolescents. The ABNF Journal, 10, 84-96.

Cicchetti, D., & Lynch, M. (1993). Toward an ecological/transactional model of
community violence and child maltreatment: Consequences for children’s
development. Psychiatry, 56, 96-1993.

Coie, J. D., Lochman, J. E., Terry, R., & Hyman, C. (1992). Predicting early adolescent
disorder from childhood aggression and peer rejection. Journal of Consulting and
Clinical Psychology, 60, 783 – 792.

Collins, L. M., Schafer, J. L., & Kam, C. M. (2001). A comparison of inclusive and
restrictive strategies in modern missing data procedures. Psychological Methods,
6, 330–351.

Cooley-Quille, M., Turner, S. M., & Beidel, D. C. (1995). Emotional impact of children's
exposure to community violence: A preliminary study. Journal of the American
Academy of Child and Adolescent Psychiatry, 34, 1362-1368.

Coulton, C.J.; Crampton, D.S.; Irwin, M.; Spilsbury, J.C.; Korbin, J.E. (2008). How
neighborhoods influence child maltreatment: a review of the literature and
alternative pathways. Child: Care, Health and Development, 34, 403 – 403.

Dempster, A.P., Laird, N.M., & Rubin, D.B. (1977). Maximum likelihood estimation
from incomplete data via the EM algorithm (with discussion). Journal of the
Royal Statistical Society, B39, 1-38.

Dodge, K. A., Bates, J. E., & Pettit, G. S. (1990). Mechanisms in the cycle of violence.
Science, 250, 1678 – 1683.

87

Dodge, K.A., & Coie, J.D. (1987). Social information processing factors in reactive and
proactive aggression in children's peer groups. Journal of Personality and Social
Psychology, 53, 1146-1158.

Dodge, K. A. & Pettit, G. S. (2003). A biopsychosocial model of the development of
chronic conduct problems in adolescence. Developmental Psychology, 39, 349 –
371.

Durant, R. H., Cadenhead, C., Pendergrast, R. A., Slavens, G., & Linder, C. W. (1994).
Factors associated with the use of violence among urban black adolescents.
American Journal of Public Health, 84, 612-617.

Eitle, D. & Turner, R. J. (2002). Exposure to community violence and young adult crime:
The effects of witnessing violence, traumatic victimization, and other stressful life
events. Journal of Research in Crime and Delinquency, 39, 214-237.

Farrell, A. D. & Bruce, S. E. (1997). Impact of exposure to community violence on
violent behavior and emotional distress among urban adolescents. Journal of
Clinical Child Psychology, 26, 2-14.

Farrell, A. D., & Sullivan, T. N. (2004). Impact of witnessing violence on growth curves
for problem behaviors among early adolescents in urban and rural settings.
Journal of Community Psychology, 32, 505-525.

Finkelhor, D., Ormrod, R. K., & Turner, H. A. (2007). Polyvictimization and trauma in
a national longitudinal cohort. Development and Psychopathology, 19, 149-166.

Gorman-Smith, D. & Tolan, P. (1998). The role of exposure to community violence and
developmental problems among inner-city youth. Developmental
Psychopathology, 10, 101-116.

Guerra, N. G., Huesmann, L. R., & Spindler, A. (2003). Community violence exposure,
social cognitions, and aggression among urban elementary school children. Child
Development, 74, 1561-1576.

Halliday-Boykins, C. A. & Graham, S. (2001). At both ends of the gun: Testing the
relationship between community violence exposure and youth violent behavior.
Journal of Abnormal Child Psychology, 29, 383-402.

Horn, J.L. & McArdle, J.J. (1992). A practical and theoretical guide to measurement
invariance in aging research. Experimental Aging Research, 18, 117-144.decade.
Journal of Abnormal Child Psychology, 31, 293-300.

88

Jensen, G. F., & Brownfield, D. (1986). Gender, lifestyles and victimization: Beyond
routine activity. Violence & Victims, 1, 85-99.

Huesmann, L. R. (1988). An information processing model for the development of
aggression. Aggressive Behavior, 14, 13-24.

Kraemer, H.C., Measelle, J.R., Ablow, J.C., Essex, M.J., Boyce, W.T., & Kupfer, D.J.
(2003). A New Approach to Integrating Data From Multiple Informants in
Psychiatric Assessment and Research: Mixing and Maxing Contexts and
Perspectives. American Journal of Psychiatry, 160, 1566-1577.

Lambert, M.C., Schmitt, N., Samms-Vaughan, M.E. An, J.S., Fairclough, M. & Nutter,
C.A. (2003). Is it Prudent to Administer All Items for Each Child Behavior
Checklist Cross-Informant Syndrome? Evaluating the Psychometric Properties of
the Youth Self-Report Dimensions with Confirmatory Factor Analysis and Item
Response Theory. Psychological Assessment, 15, 550-568.

Little, T.D., Cunningham, W.A., Shahar, G. & Widaman, K.F. (2002). To Parcel or Not
to Parcel: Exploring the Question, Weighing the Merits. Structural Equation
Modeling, 9, 151-173.

Loehlin, J. C. (2004). Latent variable models: An introduction to factor, path, and
structural equation analysis. Fourth Edition. Psychology Press: New York.

Lynch, M. & Cicchetti, D. (1998). An ecological-transactional analysis of children and
contexts: The longitudinal interplay among child maltreatment, community
violence, and children's sypmtomatology. Development and Psychopathology, 10,
235-257.

Lynch, M. (2003). Consequences of children's exposure to community violence. Clinical
Child and Family Psychology Review, 6, 265-274.

MacCallum, R.C., Browne, M.W., & Sugawara, H.M. (1996). Analysis and
determination of sample size for covariance structure modeling. Psychological
Methods, 1, 130-149.

Margolin, G. & Gordis, E. B. (2000). The effects of family and community violence on
children. Annual Review of Psychology, 51, 445-479.

Margolin, G. & Gordis, E. B. (2004). Children's exposure to violence in the family and
community. Current Directions in Psychological Science, 13, 152-155.

89

Margolin, G., Vickerman, K. A., Ramos, M. C., Serrano, S. D., Gordis, E. B., Iturralde,
E., Oliver, P. H., & Spies, L. A. (2009). Youth exposed to violence: Stability, co-
occurrence, and context. Clinical Child and Family Psychology Review, 12, 39-
54.

Martens, M.P., & Haase, R.F. (2006). Advanced applications of structural equation
modeling is counseling psychology research. The Counseling Psychologist, 34,
878–911.

Mrug, S.,& Windle, M. (2009). Bidirectional influences of violence exposure and
adjustment in early adolescence: Externalizing behaviors and school
connectedness. Journal of Abnormal Child Psychology, 37, 611-623.

Muthen, B. O. & Muthen, L. (2009). Mplus 5.21 statistical software.

National Crime Victimization Survey (2005). United States Bureau of Justice Statistics.

O'Keefe, M. (1997). Adolescent's exposure to community and school violence:
Prevalence and behavioral correlates. Journal of Adolescent Health, 20, 368-376.

Olweus, D. (1993). Bullying at school: What we know and what we can do. Malden:
Blackwell Publishing.
Overstreet, S. (2000). Exposure to community violence: Defining the problem and
understanding the consequences. Journal of Child and Family Studies, 9, 7-25.

Proctor, L. J. (2006). Children growing up in a violent community: The role of family.
Aggression and Violent Behavior, 11, 558-576.

Richters, J. E., & Saltzman, W. (1990). Survey of children's exposure to community
violence: Parent report. Rockville, MD: National Institute of Mental Health.

Scarpa, A. (2003). Community violence exposure in young adults. Trauma, Violence, &
Abuse, 4, 210-227.

Schafer, J.L. (1999). NORM: Multiple imputation of incomplete multivariate data under
a normal model (Version 2.0), Computer statistical software available from
http://www.stat.psu.edu/~jls/misoftwa.html.

Schafer, J. L., & Olsen, M. K. (1998). Multiple imputation for multivariate missing-data
problems: A data analyst ‟s perspective. Multivariate Behavioral Research, 33,
545-571.

90

Schwab-Stone, M. E., Ayers, T. S., Kasprow, W., Voyce, C., Barone, C., Shriver, T., &
Weissberg, R. P. (1995). No safe haven: A study of violence exposure in an urban
environment. Journal of American Academy of Child and Adolescent Psychiatry,
34, 1343-1352.

Schwartz, D., McFadyen-Ketchum, S. A., Dodge, K. A., Pettit, G. S., & Bates, J. E.
(1999). Early behavior problems as a predictor of later peer group victimization:
moderators and mediators in the pathways of social risk. Journal of Abnormal
Child Psychology, 27, 191-201.

Schwartz, D. & Proctor, L. J. (2000). Community violence exposure and children's social
adjustment in the social peer group: The mediating roles of emotion regulation
and social cognition. Journal of Consulting and Clinical Psychology, 68, 670-683.

Schwartz, D., Toblin, R. L., Abou-ezzeddine, T., Tom, S. R., & Stevens, K. I. (2005).
Difficult home environments and development of aggressive victims of bullying.
In K. A. Kendall-Tackett and S. M. Giacomoni (Eds). Child Victimization. Civic
Research Institute.

Shapiro, D.L., & Levendosky, A.A. (1999). Adolescent survivors of childhood sexual
abuse: the mediating role of attachment style and doping in psychological and
interpersonal functioning. Child Abuse and Neglect, 23(11), 1175-1191.

Smith, C., & Thornberry, T.P. (1995). The relationship between childhood maltreatment
and adolescent involvement in delinquency. Criminology, 33, 451-477.

Stevens, K. I. (2011). Does community violence exposure lead to externalizing
behavior problems? A meta-analytical review of the literature. In preparation.

Tabachnick, B.G., & Fidell, L.S. (2001). Using Multivariate Statistics (4
th
Ed.). Needham
Heights, MA: Allyn & Bacon.

Teisl, M. & Cicchetti, D. (2008). Physical abuse, cognitive and emotional processes, and
aggressive/disruptive behavior problems. Social Development, 17, 1 – 23.

Thompson, R., Proctor, L. J., Weisbart, C., Lewis, T. L., English, D. J., Hussey, J. M., &
Runyan, D. K. (2007). Children's self-reportsabout violence exposure: An
examiniation of the Things I Have Seen and Heard Scale. American Journal of
Orthopsychiatry, 77, 454-466.

Thornberry, T. P., Ireland, T. O., & Smith, C. A. (2001). The importance of timing: The
varying impact of childhood and adolescent maltreatment on multiple problem
outcomes. Development and Paychopathology, 13, 957 – 979.

91

Trickett, P. K., Duran, L., & Horn, J. L. (2003). Community violence as it affects child
development: Issues of definition. Clinical Child and Family Psychology
Review, 6, 223-236.

Trickett, P. K. & McBride-Chang, C. (1995). The developmental impact of different
forms of child abuse and neglect. Developmental Review, 15, 311 – 337.

Trickett, P. K., Mennen, F. E., Kim, K., Sang, J. (2009). Emotional abuse in a sample of
multiply maltreated, urban young adolescents: Issues of definition and
identification. Child Abuse and Neglect, 33, 27-35.

92

APPENDIX A: Tables
Table 11. Descriptive statistics for the variables of interest at T2 and T3 for the full
sample.

103
93

APPENDIX A continued

Table 12. Descriptive statistics for the variables of interest at T2 and T3 for females and
males.

104
94

APPENDIX A continued

Table 13. Descriptive statistics for the variables of interest at T2 and T3 for the
maltreated and non-maltreated groups.
105
95

APPENDIX A continued

Table 14. Descriptive statistics for the variables of interest at T2 and T3 for the physically
maltreated and non-physically maltreated groups.

106
107

96

APPENDIX A continued

Table 15. Correlations between the variables of interest at T2 and T3 for females (below
diagonal) and males (above diagonal).
97

APPENDIX A continued

Table 16. Correlations between the variables of interest at T2 and T3 for non-maltreated
(below diagonal) and maltreated (above diagonal).

108
98

APPENDIX A continued

Table 17. Correlations between the variables of interest at T2 and T3 for non-physically
maltreated (below diagonal) and physically maltreated (above diagonal).

109
99

100

APPENDIX C: Youth Self-report Questionnaire

AGGRESSIVE BEHAVIOR Not true
of me
Somewhat
true of me
Very true of
me
1. I argue
2. I brag
3. I am mean to others
4. I try to get a lot of attention
5. I destroy my own things
6. I destroy things belonging to others
7. I disobey my parents
8. I disobey at school
9. I don’t get along with other kids
10. I am jealous of others
11. I get in fights
12. I physically attack people
13. I scream a lot
14. I show off or clown
15. I swear or use dirty language
16. I talk too much
17. I tease others
18. I have a hot temper
19. I threaten to hurt people
20. I am louder than other kids

24
101

APPENDIX C continued

DELINQUENT BEHAVIOR Not true
of me
Somewhat
true of me
Very true
of me
21. I am pretty honest
22. I feel guilty after doing something I shouldn’t
23. I hang around with kids who get in trouble
24. I lie or cheat
25. My school work is poor
26. I run away from home
27. I set fires
28. I steal at home
29. I steal from places other than home
30. I cut classes or skip school
31. I use alcohol
102

APPENDIX D: Child Behavior Checklist

AGGRESSIVE BEHAVIOR Not true of
me
Somewhat
true of me
Very true
of me
1. Argues
2. Brags
3. Cruelty, bullying
4. Demands a lot of attention
5. Destroys own things
6. Destroys things belonging to others
7. Disobedient towards parents
8. Disobedient at school
9. Doesn’t get along with other kids
10. Jealous of others
11. Gets in fights
12. Physically attacks people
13. Screams a lot
14. Shows off or clowns
15. Swears or uses dirty language
16. Talks too much
17. Teases others
18. Temper tantrums
19. Threatens to hurt people
20. Louder than other kids
103

APPENDIX E: Teacher Report Scale

Abstract (if available)

Abstract The current investigation sought to elucidate the longitudinal relationships between community violence exposure and aggressive behavior in a sample of 389 maltreated and non-maltreated young adolescent males and females. Of primary interest was examining the effect of adolescents’ early community violence exposure and aggressive behavior on their later community violence exposure and aggressive behavior as well as examining how these relationships differ by gender and maltreatment history. The present study was conducted in the context of a larger longitudinal study examining the impacts of child maltreatment on adolescent development. The data used in the current investigation were drawn primarily from the second (T2) and third (T3) data collection waves of the parent study. ❧ To examine the longitudinal relationships, latent variable cross-lagged panel analysis methodology was employed. Results indicated that for the full sample there was no association between early community violence exposure and later aggressive behavior. However, early aggressive behavior contributed to later community violence exposure. Examinations of gender differences revealed that for males, later aggressive behavior was influenced by early exposure. There was no association for females. For both males and females, early aggressive behavior contributed to later community violence exposure. Examination of differences based on maltreatment status indicated that for non-maltreated adolescents, later aggressive behavior was influenced by early community violence exposure. There was no association for maltreated adolescents. Early aggressive behavior placed both maltreated and non-maltreated adolescents at risk for later exposure to community violence. Examination of specific maltreatment type indicated that early community violence exposure predicted later aggressive behavior for the non-maltreated and non-physically maltreated groups. There was no association for the physically maltreated group. Early aggressive behavior placed both non-maltreated and non-physically maltreated adolescents at risk for exposure to violence later on

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

The relationship between pubertal timing and delinquent behavior in maltreated male and female adolescents

PDF

Emotion regulation as a mechanism linking parents’ marital aggression to adolescent behavioral problems: a longitudinal analysis

PDF

Untangling the developmental relations between depression and externalizing behavior among maltreated adolescents

PDF

Family aggression exposure and community violence exposure associated with brain volume in late adolescence: a comparison of automated versus manual segmentation

PDF

Parents' attunement: relation to adolescent problem behavior and the moderating role of marital conflict

PDF

Maltreated adolescents and their families: a longitudinal examination of family functioning, parenting attitudes, & youth mental health

PDF

Trajectories of internalizing problems among maltreated girls and boys: differences by maltreatment type and developmental timing

PDF

Community violence exposure and children's subsequent rejection within the school peer group: the mediating roles of emotion dysregulation and depressive symptoms

PDF

Spouse aggression, depression, and physical health: a multivariate longitudinal study of midlife couples

PDF

The impact of child maltreatment on the mental health of adolescents: a longitudinal study of social anxiety and self-perception

PDF

Social status, perceived social reputations, and perceived dyadic relationships in early adolescence

PDF

How parental psychological control is related to adolescent anger and aggression

PDF

Family aggression, prosocial friends, and the risk of dating and friend victimization in late adolescence

PDF

Using observed peer discussions to understand adolescent depressive symptoms and interpersonal interactions

PDF

The development of electrodermal fear conditioning from ages 3 to 8 years and relationships with antisocial behavior at age 8 years

PDF

Parenting practices among non-offending mothers of sexually abused girls and its impact on the abused girls' behavioral adjustment: perspectives from a multigenerational, longitudinal study

PDF

Offline social functioning and online communication: how social competence translates to an online context

PDF

The relationship between dating status and academic and social functioning among Latinx and Asian-American middle adolescents

PDF

Exposure to alcohol advertising on television and alcohol use among young adolescents

PDF

Actual and perceived social reinforcements of weight-related cognitions and behaviors in adolescent peer groups