You are on page 1of 13

JOURNAL OF RESEARCH ON ADOLESCENCE, 24(2), 337–349

Family Violence, Bullying, Fighting, and Substance Use Among Adolescents:


A Longitudinal Mediational Model
Dorothy L. Espelage Sabina Low
University of Illinois, Urbana-Champaign Arizona State University

Mrinalini A. Rao and Jun S. Hong Todd D. Little


University of Illinois, Urbana-Champaign University of Kansas

Social interaction learning theory provides a framework for understanding the potential overlap among adolescent
problem behaviors such as family violence, aggression, and substance use. The current study assessed the longitudinal,
reciprocal relations among family violence, bullying perpetration, fighting perpetration, and adolescent substance use
using a sample of 1,232 students from four Midwestern middle schools. Students completed measures on family
dynamics, bullying and fighting behaviors, and alcohol and drug use three times over 18 months. Structural equation
modeling results indicated that bullying and fighting perpetration mediated the link between family violence and sub-
stance use only for males, but not for females. For females, family violence at Wave 1 was related to greater substance
use at Wave 3.

Substance use onset typically occurs during adoles- increases the risk of alcohol and drug addiction
cence, with percentages as high as 39% of 8th later in life (deBry & Tiffany, 2008; King & Chassin,
grade students having tried alcohol (Johnston, 2007; Zucker et al., 2006).
O’Malley, Bachman, & Schulenberg, 2012). Adoles- In the past several decades, theories on aggres-
cent substance use is considered a part of a cluster sion and other risky behavior have attributed con-
of problem behaviors (e.g., school failure, aggres- duct problems to the interaction between innate
sion) that co-occurs and supports an antisocial characteristics and individuals who learn the
developmental trajectory to later delinquency (e.g., behavior from their environment (Burke, Loeber, &
Dishion, Nelson, & Bullock, 2004). Indeed, there is Birmaher, 2002). From a social learning perspective,
growing evidence of co-occurrence between bully- Bandura (1986) argued that the external environ-
ing and externalizing behavior more generally ment contributes, in large part, to the acquisition
(Cook, Williams, Guerra, Kim, & Sadek, 2010; and maintenance of aggression and other risk
Sourander, Helstela, Helenius, & Piha, 2000) and behaviors. Since then, research has supported
substance use more specifically (Berthold & Hoo- social interaction learning theory (SIL), insofar as
ver, 2000; Kim, Catalano, Haggerty, & Abbott, maladaptive family dynamics spill over into other
2011). According to the Centers for Disease Control relationships (e.g., peers), which can exacerbate
and Prevention (2009), middle school students adolescent substance use (e.g., Conger & Rueter,
identified as bully–victims and bullies were 3–5 1996; Vakalahi, 2001, for a review). Thus, this study
times more likely to use alcohol and 5–6 times utilized SIL theory as a framework for examining
more likely to use drugs (respectively) than non- the associations among family violence, bullying
bullies or victims. Such findings are worrisome perpetration, fighting, and adolescent substance
given that early involvement with substance use use. To capture a broader range of family dynam-
ics, we assessed both family violence and sibling
aggression in this study.
Research for the current study was supported by the Centers The most studied family dynamic as it relates to
for Disease Control and Prevention (#1U01/CE001677) to Doro- children’s social adjustment is the coercion process
thy Espelage (PI) at the University of Illinois at Urbana-Cham-
paign. Opinions expressed herein do not necessarily reflect
(see Snyder, Reid, & Patterson, 2003). According to
those of the Centers for Disease Control and Prevention or the SIL model, impaired social interactions and
related offices. child dysfunction lead to more reciprocal, aversive
Requests for reprints should be sent to Dorothy L. Espelage,
Department of Educational Psychology, University of Illinois at © 2013 The Authors
Urbana-Champaign, Champaign, IL 61820-6925. E-mail: espelage@ Journal of Research on Adolescence © 2013 Society for Research on Adolescence
illinois.edu DOI: 10.1111/jora.12060
338 ESPELAGE, LOW, RAO, HONG, AND LITTLE

conflict, and violence that becomes a training Olweus, 1993, 2002), bullying can be interpreted as a
ground for problem behaviors and relationships proactive form of aggression with the intent of harm
within and outside the family (Reid, Patterson, & and involve many behaviors such as name-calling,
Snyder, 2002; Slomkowski, Rende, Conger, Simons, rumor-spreading, exclusion, and physical threats. In
& Conger, 2001). Because coercive exchanges contrast, fighting tends to be more reactive and is
within the family preclude the development of associated with hostile attributions and dysregulat-
positive, adaptive social tactics, these patterns gen- ed emotions (Crick, 1996). However, questions
eralize to peer interactions (see also Criss & Shaw, remain as to whether the mediating effects for bully-
2005; Haynie & McHugh, 2006; Snyder, Bank, & ing and fighting on the association between family
Burraston, 2005), which heightens opportunity for violence and substance use differ. Indeed, bullying
misbehavior (e.g., Dishion & Andrews, 1995). perpetration has been found to be supported and
driven by peers more so than fighting (Espelage,
Holt, & Henkel, 2003). If this is the case, youth who
FAMILY VIOLENCE, VIOLENCE, AND PEER
are regularly in fights might then be isolated from
AGGRESSION
peers and therefore might be more likely to use sub-
Coercive exchanges between parents and children stances to cope with peer rejection. Thus, it was
in the home often co-occur with abusive and con- hypothesized that bullying and fighting would
flictual family dynamics, which have been linked emerge as distinct constructs, thus, allowing them to
to bullying. As demonstrated by several cross- be tested as mediators in separate models.
sectional (see Espelage, Bosworth, & Simon, 2000;
Ortega and Espelage, under review) and a handful
FAMILY VIOLENCE AND SUBSTANCE USE
of longitudinal findings (see Voisin & Hong, 2012),
direct and indirect exposure to family violence is Family violence is also correlated with youth
linked to bullying behavior. More specifically, substance use. More specifically, children who are
cross-sectional investigations in the United States maltreated or those exposed to interparental
and other countries have found significant associa- violence are likely to engage in substance use
tions between exposure to physical violence in the during adolescence (Downs & Harrison, 1998) and
home and bullying (Baldry, 2003; Espelage, Low, & adulthood (Dube et al., 2006; Gilbert et al., 2009;
De La Rue, 2012; Holt, Kantor, & Finkelhor, 2009; Marcenko, Kemp, & Larson, 2000). Dube et al.’s
Moretti, Obsuth, Odgers, & Reebye, 2006; Musta- (2006) retrospective study found that those who
noja et al., 2011). For instance, Cluver, Bowes, and were physically or emotional abused and neglected
Gardner’s (2010) research, which included a sample were significantly more likely to initiate and con-
of 1,050 children in South Africa, documented that sume alcohol by age 14. A review by Downs and
children who witnessed interparental violence were Harrison (1998) also reported that childhood abuse
at an elevated risk of engaging in bullying. Further, is a major risk factor for substance use. However,
several longitudinal studies have also established findings on the link between family violence and
this association (Bauer et al., 2006; Bolger & Patter- substance use have been inconsistent (see Gilbert
son, 2001; Bowes et al., 2009). To illustrate, Baldry et al., 2009).
(2003) found that Italian elementary and middle
school children who were exposed to violence at
BULLYING, FIGHTING, AND SUBSTANCE USE
home were involved in bullying, particularly girls.
Girls who witnessed a father’s violence against the There is a well-documented correspondence between
mother or a mother’s violence against the father substance use and school violence (e.g., weapon
were among the most likely to bully others, com- carrying, fighting; DuRant, Smith, Kreiter, & Krow-
pared with girls who did not witness any forms of chuk, 1999; Furlong, Casas, Corral, Chung, & Bates,
interparental violence. In sum, there is a growing 1997; Jimerson & Furlong, 2006), providing validation
consensus that family violence is a training ground for problem-behavior theory (Donovan, Jessor, &
for peer aggression (and in turn, associated risk Costa, 1991). Despite evidence for clustering of risk
behaviors). behaviors, surprisingly fewer studies have specifi-
Despite this growing consensus, it is not clear that cally examined bullying behavior and substance use.
family violence is equally associated with both Studies that have examined bullying behavior have
bullying and fighting perpetration. Based on the cur- consistently found a link between bullying and sub-
rent defining characteristics of bullying (i.e., imbal- stance abuse (Ensminger, Juon, & Fothergill, 2002;
ance of power, intent to cause harm, persistence; Houbre, Tarquinio, Thuillier, & Hergott, 2006; Luk,
FAMILY VIOLENCE, BULLYING, AND SUBSTANCE USE 339

Wang, & Simons-Morton, 2010; Taiwo & Goldstein, relationships of these youth (Duncan, 1999).
2006), although these have all utilized cross-sectional Moreover, although sibling aggression is the most
designs. Several more recent longitudinal studies common type of family violence (Gelles & Straus,
support the link between bullying perpetration and 1988; Roscoe, Goodwin, & Kennedy, 1987), societal
substance use among early adolescents. Hemphill awareness of sibling aggression has lagged behind
et al. (2011) found that greater bullying perpetration other types of family violence. Compared with par-
among Australian youth in Year 7 of school was asso- ent–child abuse and interparental violence, research
ciated with a twofold increase in binge drinking and on intersibling abuse or violence among children in
marijuana use when these students were in Year 10 the same home is generally scant (see Caffaro &
of school. From the U.S.-based Raising Healthy Chil- Conn-Caffaro, 2005), and only a handful of studies
dren project, childhood bullying in Grade 5 was asso- have considered sibling aggression when examining
ciated with heavy drinking and marijuana use at age family context of bullying (e.g., Duncan, 1999). Sec-
21 (Kim et al., 2011). Other studies have shown longi- ond, in addition to placing bullying perpetration
tudinal associations between bullying among older and substance use in a unified theoretical frame-
adolescents and associations with heavy drinking work, this study augments previous research by
and marijuana use into adulthood, but these studies utilizing longitudinal data. Third, because previous
often enter family variables as covariates and do not literature on the relations between bullying and sub-
build mediation models. stance use have isolated specific forms of bullying
Although research findings suggest a correspon- (e.g., Sullivan et al., 2006; Tharp-Taylor, Haviland,
dence among family violence, bullying, and sub- & D’Amico, 2009), we were specific in the contrast-
stance use, much is still unknown about this ing forms of aggression (i.e., bullying, fighting).
association due to a dearth of longitudinal research Finally, given prior literature suggesting that gender
with early adolescent samples and a lack of speci- may modify the link between both family violence
ficity among various forms of aggression. The aim and peer aggression, as well as between aggression
of this study is to fill this research gap by testing and substance use, we examined pathways condi-
and contrasting bullying and fighting as mediators tional upon gender. We hypothesized a positive,
of the relation between family violence and sub- longitudinal association between family violence
stance use in a sample of early adolescents using and substance use, and that both forms of aggres-
longitudinal data. Indeed, aggression may be one sion (i.e., bullying and fighting) would mediate the
of several conduits of the association between fam- relation between family violence and substance use.
ily violence and AOD use (e.g., depression, impul-
sivity); however, the literature clearly points to
aggression as a potential mediator in understand- METHOD
ing how these are mutually entrained over time. In
Design and Sample
addition, because some studies suggest a stronger
relation between peer aggression and substance Participants were 1,232 students from four Midwest-
use for males (e.g., Skara et al., 2008; Sullivan, ern middle schools (Grades 5–7); 49.8% female; 51%
Farrell, & Kliewer, 2006), gender differences in the African American, 34% White, 3% Hispanic, 3%
mediation models were investigated. Asian, and 9% identified as “other.” Students ran-
ged in age from 10 to 15 years of age (M = 13.9;
SD = 1.05). Students completed a survey designed
SUMMARY OF CURRENT STUDY
to collect information about the level of violence in
The aim of this study is to use a SIL framework to their home with parents and siblings or other chil-
investigate the relation among family violence (i.e., dren, frequency of bullying and fighting perpetra-
family violence and sibling aggression), bullying, tion, and year alcohol and drug use (AOD). Students
fighting, and substance use, utilizing a longitudinal completed surveys once in Spring 2008 (Wave 1),
transactional model. This study augments prior Fall 2008 (Wave 2), and Spring 2009 (Wave 3).
research in three important regards. First, we draw
upon a richer, more comprehensive assessment of
Procedure
family violence by incorporating sibling aggression
or aggression among children in the home. Because Parental consent. A waiver of active parental
research suggests a significant difference in parent- consent was approved by the institutional review
ing among bullies and noninvolved youth, it is board and school district administration. Parents of
plausible that those differences extend to sibling all students enrolled in the schools were sent letters
340 ESPELAGE, LOW, RAO, HONG, AND LITTLE

informing them about the purpose of the study. Scale had a low correlation with the Victimization
Several meetings were held to inform parents of Scale (r = .21) and was only moderately correlated
the study in each community. In the early Spring with the Bullying Scale (r = .58), providing evidence
of 2008, investigators attended parent–teacher con- of discriminant validity (Espelage & Holt, 2001).
ference meetings and staff meetings, and the study Cronbach’s alpha coefficients were .70 for Wave 1
was announced in school newsletters and emails and .75 for Waves 2 and 3 in this study.
from the principals. Furthermore, parents were
asked to sign the form and return it only if they Bullying perpetration. The nine-item Illinois
were unwilling to have their child participate in Bully Scale (Espelage & Holt, 2001) assesses the fre-
the investigation. At the beginning of each survey quency of bullying at school. Students are asked
administration, teachers removed students from how often in the past 30 days they did the follow-
the room if they were not allowed to participate, ing to other students at school: teased other stu-
and researchers also reminded students that they dents, upset other students for the fun of it,
should not complete the survey if their parents had excluded others from their group of friends, helped
returned the form. A 95% participation rate was harass other students, and threatened to hit or hurt
achieved. Students were asked to consent to partic- another student. Response options include never, 1
ipate in the study through an assent procedure or 2 times, 3 or 4 times, 5 or 6 times, and 7 or more
included on the coversheet of the survey. times. The construct validity of this scale has been
supported via exploratory and confirmatory factor
Survey administration. Six trained research analysis (Espelage & Holt, 2001). Factor loadings in
assistants, the primary researcher, and a faculty the development sample for these items ranged
member collected data. At least two of these individ- from .52 to .75, and this factor accounted for 31% of
uals administered surveys to classes ranging in size the variance in the factor analysis (Espelage & Holt,
from 10 to 25 students. Students were first informed 2001). Higher scores indicated more self-reported
about the general nature of the investigation. Next, bullying behaviors. The scale correlated moderately
researchers made certain that students were sitting with the Youth Self-Report Aggression Scale
far enough from one another to ensure confidential- (r = .65; Achenbach, 1991), suggesting that it was
ity. Students were then given survey packets, and somewhat distinct from general aggression. Con-
the survey was read aloud to them. It took students current validity of this scale was established with
approximately 40 min on average to complete the significant correlations with peer nominations of
survey. bullying (Espelage et al., 2003). More specifically,
students who reported the highest level of bully
perpetration on the scale received significantly
Measures
more bullying nominations (M = 3.50, SD = 6.50)
Each participant completed a questionnaire about from their peers than students who did not self-
involvement in peer aggression, exposure to family report high levels of bullying perpetration (M = .98;
violence and substance use, as well as a demo- SD = 1.10; Espelage et al., 2003). This scale was not
graphic questionnaire that included questions about significantly correlated with the Illinois Victimiza-
his or her gender age, grade, and race or ethnicity. tion Scale (r = .12) and thus provided evidence of
For race, participants were given six options: Afri- discriminant validity (Espelage et al., 2003). Cron-
can American (not Hispanic), Asian, White (not His- bach’s alpha coefficients were .86 for each of the
panic), Hispanic, Native American, and other (with three waves of data collection in this study.
a space to write in the most appropriate racial
descriptor).
Family Violence, Sibling Verbal and
Physical Aggression
Multiple Forms of Aggression Perpetration
Family violence. The Family Conflict and Hostility
Fighting perpetration. The four-item University Scale (Thornberry, Krohn, Lizotte, Smith, & Tobin,
of Illinois Fighting Scale (Espelage & Holt, 2001) 2003) was used to measure the level of perceived
assesses physical fighting behavior (e.g., I got in a conflict and hostility in the family environment. The
physical fight; I fought students I could easily beat) scale contains three items from a larger survey,
the respondent engaged in over the past 30 days. which was designed for the Rochester Youth Devel-
Response options include never, 1 or 2 times, 3 or 4 opment Study. The three items were as follows:
times, 5 or 6 times, and 7 or more times. The Fighting How often is there yelling, quarreling, or arguing in
FAMILY VIOLENCE, BULLYING, AND SUBSTANCE USE 341

your household? How often do family members lose RESULTS


their temper or blow up for no good reason? How
Preliminary Analyses
often are there physical fights in the household, like
people hitting, shoving, or throwing things? First, preliminary analyses were conducted to deter-
Response options include never, seldom, sometimes, mine the nature and extent of missing data. To
often, and always. Cronbach’s alpha coefficients were address the issue of missing data for the current
.79 for Wave 1 and .81 for Waves 2 and 3 in this sample, a multiple imputation procedure was
study. executed using the PROC MI function in SAS 9.2
(Graham, Cumsille, & Elek-Fisk, 2003; SAS Institute
Sibling verbal and physical aggression. A sib- Inc., 2008). Overall, missingness per item ranged
ling aggression scale was created for this study and from 0% to 4.7%, with a mean percentage of missing
included five items that assessed aggression data across all measured variables of 1.7%. Although
between siblings (Espelage & Stein, 2006). Items Luengo, Garcıa, and Herrera (2010) suggest that
were selected from the University of Illinois Bully- missing data between 1% and 5% are generally man-
ing Scale to parallel that scale. Students were ageable, a multiple imputation procedure was
presented the following instructions “Now, we employed to preserve the integrity of each group of
would like for you to think of your siblings or other respondents and create one parsimonious data set.
children in your family. For each of the following Second, an item-to-construct balance method
questions, choose how many times you did this was used to develop parcels for the four latent
activity or how many times these things happened constructs of family violence, fighting and bully
to you in the LAST 30 DAYS among your siblings or perpetration, and alcohol and drug use (Little,
other children in your home.” Five items emerged Cunningham, Shahar, & Widaman, 2002). To have
as a scale in factor analysis, which includes the fol- a just identified model, items from the scales were
lowing: I upset my brother or sister for the fun of it; parceled into three indicators for each latent con-
I got into a physical fight with my brother or sister; I struct. To do this, a single-item exploratory factor
started arguments with my brother or sister; I hit analysis was run in SPSS using maximum likeli-
back when a sibling hit me first; and I teased my sib- hood estimation with one fixed factor. Items were
lings for the fun of it. Response options include then divided into three parcels based on the factor
never, 1 or 2 times, 3 or 4 times, 5 or 6 times, and 7 or loading of each item onto the single factor, and
more times. A Cronbach’s alpha coefficient of .82 was these items were averaged to create the parcel
found for each wave of data collection in this study. value. Items were averaged, as opposed to
summed, to reflect the actual scale that was used
to record the item-level information as well as
Risk Behavior
maintain comparable metrics between parcels. The
Substance use. Eight items from the Problem advantages of parcels over single items include
Behavior Frequency Scale (Farrell, Kung, White, & greater reliability, more communality, a higher
Valois, 2000) were used, which asked students to ratio of common-to-unique factor variance, reduc-
report how many times in the past year (or since tions in distributional violations, and decreased
the last time they were surveyed for Waves 2 and chance for correlated residuals or dual loadings
3) they used alcohol and or drugs. The scale con- (Little et al., 2002). Internal reliabilities (i.e., Cron-
sists of items, such as “smoked cigarettes,” “drunk bach’s a) for each scale were calculated using par-
liquor,” and “used inhalants.” Response options celed indicators (Roche, Ghazarian, Little, &
include never, 1 or 2 times, 3 or 5 times, 6 or 9 times, Leventhal, 2011). The correlations between the
and 10 or more times. A Cronbach’s alpha coefficient latent constructs are presented for girls and boys in
of .87 was found with a sample of urban adoles- Table 1. Nested chi-squared difference tests were
cents and .88 with a sample of rural adolescents conducted to examine whether there were any dif-
(Farrell et al., 2000). The authors also report posi- ferences in the correlations across boys and girls.
tive correlations with risk behaviors such as self-
reported delinquency and negative correlations
Measurement and Structural Model
with positive behaviors and school attendance (Far-
rell et al., 2000). In the current study, a Cronbach’s Measurement model. To ensure that the con-
alpha coefficient of .90 was found for each wave of structs investigated are measured equivalently
data collection in this study. across the three time points, a confirmatory factor
342 ESPELAGE, LOW, RAO, HONG, AND LITTLE

TABLE 1
Correlations Between Family Conflict (FC), Bullying Perpetration (Bully), Fighting Perpetration (Fight), and Substance Use (AOD)
Latent Constructs for Boys and Girls

FC1 Bully1 Fight1 AOD1 FC2 Bully2 Fight2 AOD2 FC3 Bully3 Fight3 AOD3

FC1 — .62 .40 .14 .64 .51 .32 .32 .44 .46 .33 .20
Bully1 .69 — .69 .22 .40 .51 .39 .33 .40 .54 .43 .42
Fight1 .49 .73 — .24 .30 .35 .44 .26 .29 .37 .48 .38
AOD1 .22 .35 .23 — .11 .18 .27 .79 .21 .23 .29 .29
FC2 .69 .59 .38 .22 — .69 .55 .34 .63 .51 .48 .30
Bully2 .50 .58 .42 .28 .77 — .71 .35 .53 .63 .50 .37
Fight2 .25 .42 .40 .20 .45 .63 — .41 .46 .44 .60 .35
AOD2 .34 .40 .28 .81 .49 .54 .43 — .33 .41 .36 .41
FC3 .53 .49 .31 .16 .78 .60 .42 .37 — .65 .54 .39
Bully3 .54 .61 .35 .25 .74 .75 .46 .45 .81 — .68 .56
Fight3 .37 .49 .38 .28 .56 .56 .68 .45 .62 .69 — .50
AOD3 .35 .41 .28 .52 .45 .41 .37 .62 .46 .43 .45 —

Note. Numbers above the diagonal are correlations for the boys, and the numbers below the diagonal are correlations for the girls.
Bolded numbers indicate that these values are significantly different across gender.

analysis procedure using a structural equation Analyses were also conducted to examine
model (SEM) framework was conducted across the whether fighting perpetration and bullying
variables in each analysis. Three levels of invari- perpetration were unique constructs, or whether
ance were tested: Configural invariance (which they formed one latent variable of aggression in
establishes that the pattern of fixed and free param- the analyses. To test this, all four constructs (i.e.,
eters is the same), weak factorial invariance (which family conflict, fighting perpetration, bullying per-
establishes that the relative factor loadings are pro- petration and AOD) were included in confirmatory
portionally equal across groups), and strong facto- factor analyses that examined the invariance of the
rial invariance (which establishes that the relative constructs across time points. Strong invariance
indicator means are proportionally equal across was achieved. Next, the correlations between the
groups). The effects-coding method of scaling for latent constructs of fighting perpetration and bully-
the latent constructs was used in these analyses. ing perpetration were fixed at 1.0 within each time
This method maintains the original metric of the point, and their correlations with other constructs
observed variables in the latent construct form, and both within time and across time were equated,
the latent construct values, therefore, have mean- essentially collapsing the two constructs to function
ingful interpretations (Little, Slegers, & Card, 2006). as one. The fit of this statistically collapsed model
The loadings, intercept values, residuals, and R2 was compared with the strong invariance model
values for each indicator from the strong invariance with all four constructs. The chi-squared difference
model are presented separately for girls and boys test indicated that the collapsed model was signifi-
in Tables S1 and S2 in the supporting information. cantly different from the four construct model [v2
For this study, the Tucker-Lewis Index (TLI), Com- (30, 651.198)]. This demonstrates that fight perpe-
parative Fit Index (CFI), and Root Mean Square tration and bullying perpetration are unique latent
Error of Approximation (RMSEA) were used. constructs, and consequently, these were tested
Where TLI and CFI scores >.95 are considered an separately in the structural analyses.
acceptable fit (Schermelleh-Engel, Moosbrugger, & Additionally, multigroup comparisons were con-
Muller, 2003), and RMSEA scores of above .10 are ducted to examine invariance across gender. The
considered a poor fit, between .08 and .10 a medio- groups were considered invariant when the model
cre fit, between .05 and .08 an acceptable fit, .01 with the equivalence constraints associated with
and .05 a close fit, and .00 an exact fit (Hu & each level of invariance does not differ significantly
Bentler, 1999). Fit statistics demonstrated that from the previous level of invariance. If the change
strong factorial invariance held for all constructs, in CFI is <.01, then the assumption of invariance
and that all constructs were measured equivalently holds (Cheung & Rensvold, 2002). Although the
across all three time points in the four analyses (see groups passed the invariance tests up to the level of
Table 2 and Tables S3 and S4 in supporting infor- strong invariance, the test examining the homogene-
mation). ity of the variance–covariance matrix across groups
FAMILY VIOLENCE, BULLYING, AND SUBSTANCE USE 343

TABLE 2
Fit Indices for the Multiple Group Analyses of Family Violence—Fighting Perpetration—AOD Across Gender

RMSEA
Model v2 df p Δ v2 Δdf p RMSEA 90% CI CFI Δ CFI TLI ΔTLI Pass?

Null 22435.49 792 <.001


Measurement model estimatesa
Configural invariance 1933.53 522 <.001 .07 .064 .071 .970 .960 Yes
Weak invariance 2110.58 552 <.001 177.05 30 .00 .07 .066 .072 .967 .003 .958 .002 Yes
Strong invariance 2255.77 582 <.001 145.19 30 .00 .07 .067 .073 .965 .002 .958 .000 Yes
Latent model estimatesa
Homogeneity of variances/ 2949.42 648 <.001 693.65 66 .00 .08 .075 .081 .952 .013 .939 .019 No
covariances
Longitudinal structural modelb
Final SEM 2293.580 595 <.001 37.81 13 .00 .07 .067 .073 .964 .001 .958 .000 Yes

Note. SEM = structural equation models, Δ = the change in value as compared to the previous test, RMSEA = Root Mean Square
Error of Approximation, CFI = Comparative Fit Index, TLI = the Tucker-Lewis Index.
a
Evaluated with the CFI difference test.
b
Evaluated with the chi-squared difference test.

indicated that the variables examined were not petration on the association between family vio-
invariant across gender (see Table 2 and Table S3 in lence on substance use (Figure 1).
supporting information). Consequently, two-group Additionally, several paths were equivalent
analyses were conducted examining the relation across gender. Autoregressive paths from all vari-
between variables separately for boys and girls. ables at Wave 1 to Wave 2 were equivalent across
gender. Cross-lagged paths from fighting perpetra-
Structural equation models. Next, SEM using tion to substance use over both time lags were
the Cole and Maxwell (2003) framework were used equivalent across gender. Additionally, the cross-
to test our mediation hypotheses. Mediation was lagged path from family violence at Wave 2 to
concluded when the indirect effect was significant. fighting perpetration at Wave 3 was equivalent
The indirect effect is the product of the path from across gender.
the independent variable to the mediator variable Next, analyses examining the bivariate longitudi-
and the path from the mediator variable to the nal associations among pairs of the three constructs
dependent variable. were conducted to examine how each of the variables
Fighting perpetration. Prior to fitting the SEM was related to each other outside the mediation
testing the mediation hypothesis, SEMs examining framework. Here, too, the transactional associations
bivariate longitudinal associations among family present in the bivariate longitudinal analyses
violence, fighting perpetration, and substance use dropped out for both gender and an additional auto-
were conducted to examine the associations regressive path from fighting perpetration at Wave 1
between each pair of these variables. These results to fighting perpetration at Wave 3 was found.
indicated positive transactional associations Bullying perpetration. Similar results were found
between all pairs of the variables for both girls and for the mediator of bullying perpetration. Results
boys. A two-group mediation model examining the indicated positive transactional associations
mediating role of fighting perpetration on the asso- between all pairs of the variables for both girls and
ciation between family violence and substance use boys. Bullying perpetration mediated the relation
was fit across gender. This analysis indicated that between family violence and substance use only for
fighting perpetration mediated the relation between boys (see Table S3 in supporting information). The
family violence and substance use only for boys overall indirect effect (.09) was significant at the
(see Table 2). The overall indirect effect, although p < .05 level, indicating mediation (Figure S1 in
small (.03), was significant at the p < .05 level, indi- supporting information). Positive predictive paths
cating mediation (Figure 1). Positive predictive were found among the variables for the girls. How-
paths were found among the variables for the girls. ever, bullying perpetration at Wave 2 did not pre-
However, family violence at Wave 1 did not pre- dict substance use at Wave 3, and consequently, the
dict fighting perpetration at Wave 2, precluding hypothesized meditational role of bullying perpetra-
the hypothesized meditational role of fighting per- tion on the association between family violence on
344 ESPELAGE, LOW, RAO, HONG, AND LITTLE

Family Violence
.67 Family
.59/.83 Family
W1 Violence W2 Violence W3

.17/-.01(ns) .31
.43 .31 .17

.12 Fighting
.42 Fighting
.41/.57 Fighting
Perpetration Perpetration Perpetration .11
W1 W2 W3
.10

.22 .10 .13 .12


.17

Substance use .78 .30/.59


Substance use Substance use
W1 W2 W3

FIGURE 1 Mediational model testing the mediating role of fighting perpetration on the relation between family violence and AOD
use for boys and girls.
Note. Values for boys are in front of the slash (/), and values for the girls are behind the slash; if there is a single value, then this path
is equal across groups. All effects have p < .001 unless indicated otherwise; Model fit: v2 (595, n = 1,132) = 2293.58; Root Mean Square
Error of Approximation (RMSEA) = .07(.067; .073); NNFI = .0958; Comparative Fit Index (CFI) = .964; R2 for substance use at Wave 3 is
.22 for boys and .41 for girls.

substance use was not found (Figure S1 in support- the equality constraints. Results showed that plac-
ing information). ing equality constraints on the mediating pathways
Further analyses indicated that the initial associ- did not significantly change the fit of the model [v2
ations between the three variables are equivalent (2, 1.225)] indicating that the mediating influence
across genders, but different associations emerge as of fighting and bullying did not differ significantly
the adolescents got older. The transactional associa- from each other for the male sample (see Table S4
tions between variables found in the three-variable in Supporting information). This suggests that
model dropped out for both gender. although bullying and fighting perpetration are
Comparing bullying and fighting perpetra- unique constructs, they operate similarly as media-
tion. Additional analyses were conducted to tors for the relation between family violence and
examine whether bullying and fighting perpetra- substance use.
tion uniquely mediated the association between
family violence and substance use for boys. Adapt-
DISCUSSION
ing the multiple group structural model tests of
equivalence across groups, a two-group structural Fighting and bullying can be broadly construed as
model was tested to examine two sets of variables. maladaptive social interactions (or social dynamic)
One set of variables examined the mediating role that are influenced, maintained, or mitigated by
of fighting perpetration on the relation between relationships in the school, peer, and familial con-
family violence and substance use with the male texts. Despite this, there are few studies that inves-
sample, and the other set of variables examined the tigated family context of bullying particularly in
mediating role of bullying perpetration on the rela- the United States (e.g., Wang, Iannotti, & Nansel,
tion between family violence and substance use, 2009), and the majority of the existing studies on
also with the male sample. As expected from the the topic have been cross-sectional. Our research is
previous sets of analyses, results indicated that predicated on a SIL model in which family vio-
fighting perpetration and bullying perpetration lence serves as an important context for under-
separately mediated the relation between family standing the relation between fighting and bullying
violence and substance use. Equality constraints perpetration and substance use. As such, this study
were then placed on the mediating pathways to addresses the paucity of longitudinal studies of
examine whether the model with these constraints family context and bullying in the United States,
would significantly differ from the model without as well as the significant dearth of longitudinal
FAMILY VIOLENCE, BULLYING, AND SUBSTANCE USE 345

investigations on bullying perpetration and sub- (i.e., bullying and fighting). Given that adolescent
stance use. girls report higher rates of internalizing symptoms
Our results suggest that fighting and bullying (i.e., depression, anxiety; Broderick & Korteland,
perpetration are important mechanisms linking 2002), a more relevant mediator between family
family violence and substance use only for male violence and alcohol and drug use might be
adolescents. That is, fighting and bullying perpetra- depression for girls and should be considered in
tion served as indirect mechanisms between family subsequent studies.
violence at Wave 1 and substance use at Wave 3. These data augment the aggression literature by
These findings are consistent with previous suggesting bullying (name-calling, rumor-spread-
research that has linked exposure to familial vio- ing) perpetration is an explanatory variable
lence as a precursor to alcohol and drug use (Dube between family violence and substance use among
et al., 2006; Gilbert et al., 2009; Marcenko et al., a sample of male adolescents. Although aggression
2000). The connection between familial violence is multidimensional, findings overall suggest that,
and aggression at school is not a new finding and when examined conjointly with AOD use and fam-
not even surprising, given the intergenerational ily violence, bullying and fighting operate some-
transmission of aggression and the SIL framework what similarly (for males and females). However, it
of youth aggression (Reid et al., 2002; Slomkowski remains important to not assume redundancy
et al., 2001). For males, both forms of aggression, among these behaviors in future work, as the
whether manifested in fighting or verbal and rela- specific antecedents and consequences may vary
tional aggression, seem to be a part of a cluster of across contexts. It is imperative that researchers
“problem behavior,” insofar as substance use and investigating the family context of bullying and
bullying show reciprocal relations and operate in substance abuse examine not only violence involv-
similar context (i.e., a spurious relation). ing parents but also that involving siblings. A note-
Two potential explanations could be drawn from worthy strength of this study was the assessment
these findings. First, for males, fighting and bully- of violence between and among parents and chil-
ing perpetration could be simply another manifesta- dren. In addition, future studies are needed to flesh
tion of an antisocial tendency that is a precursor to out the secondary mechanisms (i.e., more proximal
alcohol and drug onset (Farrington & Ttofi, 2011). determinants) explaining the links between bully-
Second, although not accounted for in this model, it ing and substance use, ideally taking the forms of
is plausible that for males, both behaviors stem bullying into account, as well as individual level
from a similar peer environment, escalating both (e.g., race, ethnicity, socioeconomic status, gender)
behaviors, as evidenced by the cascading strength and macro level contextual factors (e.g., culture).
of the relations between aggression and AOD use The current study has several strengths, includ-
over time. After all, youth who are prone to engage ing a large, diverse sample, and multiple waves of
in fighting and bullying perpetration might have data collection (albeit over a relatively shorter win-
difficulty with self-control, which makes socializing dow of time). The overreliance on cross-sectional
with a conventional peer group difficult (Schwartz data in the literature has prevented the field from
& Proctor, 2000). These youth are prone to rejection understanding how bullying and fighting perpetra-
and exclusion by their mainstream peers (Dishion, tion may mediate associations between family
Patterson, & Stoolmiller, 1991), and consequently violence and substance use. This study was also
resort to affiliating with other aggressive peers who unique given the comprehensive assessment of
might be engaging in deviant behaviors, including family violence. Students reported on the general
alcohol and drug use (Farmer et al., 2010; Sijtsema, level of conflict in their family, but also reflected
Lindenberg, & Veenstra, 2010). upon aggression among children in the family.
For females, we found a direct effect from expo- Finally, this sample included students in middle
sure to family violence (Wave 1) to greater alcohol school, a time in which alcohol and drug use
and drug use over time, although the absence of onsets, and when engagement in other antisocial
mediation in both models suggests that these are behavior more generally increases.
related but somewhat independent processes when Despite this, there are also a few limitations.
examined over time. For females, it seems the cou- Foremost is the reliance on youth self-report rather
pling of these behaviors over time is not as evident than multiple reports, which could diminish the
as it is for males; insofar that family violence is validity and reliability (i.e., sampling) of reported
influencing growth in AOD use for females, but behavior. Second, our measure of family violence
somewhat independently of aggressive behavior does not distinguish between direct and indirect
346 ESPELAGE, LOW, RAO, HONG, AND LITTLE

violence exposure, which could lead to underesti- Bolger, K. E., & Patterson, C. J. (2001). Developmental
mation of relations among observed variables. Fur- pathways from child maltreatment to peer rejection.
ther, the sibling conflict items assumed that the Child Development, 72, 549–568. doi:10.1111/14678624.
student participants had either a sibling or a child 00296
Bowes, L., Arseneault, L., Maughan, B., Taylor, A., Caspi,
in the home, but the survey did not ask how many
A., & Moffitt, T. E. (2009). School, neighborhood, and
siblings were in the home. Third, these data were
family factors are associated with children’s bullying
collected in one community located in a Midwest- involvement: A nationally representative longitudinal
ern urban setting and these findings need to be study. Journal of the American Academy of Child Adoles-
replicated in other communities to enhance the cent Psychiatry, 48, 545–553. doi:10.1097/CHI.
ecological validity of our findings. Lastly, although 0b013e31819cb017
this is a conservative test, the mediational path- Broderick, P. C., & Korteland, C. (2002). Coping style
ways are nevertheless small in magnitude, and and depression in early adolescence: Relationships to
warrant replication. These limitations aside the cur- gender, gender role, and implicit beliefs. Sex Roles, 46,
rent findings represent an important initial step in 201–213. doi:10.1023/A:1019946714220
enhancing youth violence scholarship through the- deBry, S. C., & Tiffany, S. T. (2008). Tobacco-induced
neurotoxicity of adolescent cognitive development
ory-driven, longitudinal research.
(TINACD): A proposed model for the development of
Taken together, the findings from this study sup-
impulsivity in nicotine dependence. Nicotine & Tobacco
port the social information learning framework in Research, 10, 11–25.
understanding the relations among family violence, Burke, J. D., Loeber, R., & Birmaher, B. (2002). Opposi-
fighting, bullying, and substance use in a sample of tional defiant disorder and conduct disorder: A review
early adolescents. As boys are exposed to greater of the past 10 years, part II. Journal of the American
levels of family conflict, they engage in increasing Academy of Child & Adolescent Psychiatry, 41, 1275–1293.
levels of aggression directed toward their peers at Caffaro, J. V., & Conn-Caffaro, A. (2005). Treating sibling
school, and in turn, this aggression is associated abuse families. Aggression and Violent Behavior, 10,
with greater levels of AOD use over time. Preven- 604–623. doi:10.1016/j.avb.2004.12.001
tion programs that focus on reduction of aggression Centers for Disease Control and Prevention. (2009). Bul-
lying among middle school and high school students
might therefore have an impact in reducing AOD
—Massachusetts, 2009. MMWR 2011, 60, 465–495.
use among male adolescents. Girls who are exposed
Cheung, G. W., & Rensvold, R. B. (2002). Evaluating
to high levels of family violence report greater alco- goodness-of-fit indexes for testing measurement invari-
hol and drug use over time, suggesting that aggres- ance. Structural Equation Modeling, 9, 233–255. doi:10.
sion is not a conduit from family violence to AOD 1207/S15328007SEM0902_5
use. It is imperative that future research examines Cluver, L., Bowes, L., & Gardner, F. (2010). Risk and pro-
other potentially malleable mediators of the family tective factors for bullying victimization among AIDS-
violence-AOD association for girls. affected and vulnerable children in South Africa. Child
Abuse and Neglect, 34, 793–803.
Cole, D. A., & Maxwell, S. E. (2003). Testing mediational
REFERENCES models with longitudinal data: Questions and tips in
the use of structural equation modeling. Journal of
Achenbach, T. M. (1991). Manual for the youth self-report
Abnormal Psychology, 112, 558–577. doi:10.1037/0021-
and 1991 profile. Burlington, VT: University of Vermont,
843X.112.4.558
Department of Psychiatry.
Conger, R., & Rueter, M. (1996). Siblings, parents, and
Baldry, A. C. (2003). Bullying in school and exposure to
peers: A longitudinal study of social influences in ado-
domestic violence. Child Abuse and Neglect, 27, 713–732.
lescent risk for alcohol use and abuse. In G. Brody
doi:10.1016/S0145-2134(03)00114-5
(Ed.), Sibling relationships: Their causes and consequences.
Bandura, A. (1986). Social foundations of thought and action:
Advances in applied developmental psychology (pp. 1–30).
A social cognitive theory. Englewood Cliffs, NJ: Prentice-
Norwood, NJ: Ablex.
Hall.
Cook, C. R., Williams, K. R., Guerra, N. G., Kim, T. E., &
Bauer, N. S., Herrenkohl, T. I., Lozano, P., Rivara, F. P.,
Sadek, S. (2010). Predictors of bullying and victimiza-
Hill, K. G., & Hawkins, J. D. (2006). Childhood bully-
tion in childhood and adolescence: A meta-analytic
ing involvement and exposure to intimate partner vio-
investigation. School Psychology Quarterly, 25, 65–83.
lence. Pediatrics, 118, 235–242. doi:10.1542/peds.2005-
doi:10.1037/a0020149
2509
Crick, N. R. (1996). The role of relational aggression,
Berthold, K. A., & Hoover, J. H. (2000). Correlates of bul-
overt aggression, and prosocial behavior in the predic-
lying and victimization among intermediate students
tion of children’s future social adjustment. Child Devel-
in the Midwestern USA. School Psychology International,
opment, 67, 2317–2327. doi:10.2307/1131625
21, 65–78. doi:10.1177/0143034300211005
FAMILY VIOLENCE, BULLYING, AND SUBSTANCE USE 347

Criss, M., & Shaw, D. S. (2005). Sibling relationship qual- lence, and psychological outcomes during adolescence.
ity as a context for delinquency training in low-income Psychology of Violence, 2, 313–324. doi:10.1037/a0027386
families. Journal of Family Psychology, 19, 592–600. Espelage, D. L., & Stein, N. D. (2006). The University of
doi:10.1037/0893-3200.19.4.592 Illinois and Wellesley College: Student behavior survey.
Dishion, T. J., & Andrews, D. W. (1995). Preventing esca- Champaign, IL: University of Illinois at Urbana-Cham-
lation in problem behaviors with high-risk young ado- paign, Department of Educational Psychology.
lescents: Immediate and 1-year outcomes. Journal of Farmer, T. W., Petrin, R. A., Robertson, D. L., Fraser, M.
Consulting and Clinical Psychology, 63, 538–548. doi:10. W., Hall, C. M., Day, S. H., & Dadisman, K. (2010).
1037/0022-006X.63.4.538 Peer relations of bullies, bully-victims, and victims:
Dishion, T. J., Nelson, S. E., & Bullock, B. M. (2004). Pre- The two worlds of bullying in second-grade class-
mature adolescent autonomy: Parent disengagement rooms. The Elementary School Journal, 110, 364–392.
and deviant peer process in the amplification of prob- doi:10.1086/648983
lem behavior. Journal of Adolescence, 27, 515–530. doi:10. Farrell, A. D., Kung, E. M., White, K. S., & Valois, R.
1016/j.adolescence.2004.06.005 (2000). The structure of self-reported aggression, drug
Dishion, T. J., Patterson, G. R., & Stoolmiller, M. L. (1991). use, and delinquent behaviors during early adoles-
Family, school, and behavioral antecedents to early ado- cence. Journal of Clinical Child Psychology, 29, 282–292.
lescent involvement with antisocial peers. Developmental doi:10.1207/S15374424jccp2902_13
Psychology, 27, 172–180. doi:10.1037/0012-1649.27.1.172 Farrington, D. P., & Ttofi, M. M. (2011). Bullying as a
Donovan, J. E., Jessor, R., & Costa, F. (1991). Adolescent predictor of offending, violence and later life out-
health behavior and conventionality–unconventional- comes. Criminal Behaviour and Mental Health, 21, 90–98.
ity: An extension of Problem-Behavior Theory. Health doi:10.1002/cbm.801
Psychology, 10, 2–61. doi:10.1037/0278-6133.10.1.52 Furlong, M., Casas, J. M., Corral, C., Chung, A., & Bates,
Downs, W. R., & Harrison, L. (1998). Childhood maltreat- M. (1997). Drugs and school violence. Education and
ment and the risk of substance problems in later life. Treatment of Children, 20, 263–280.
Health and Social Care in the Community, 6, 35–46. Gelles, R. J., & Straus, M. A. (1988). Intimate violence.
Dube, S. R., Miller, J. W., Brown, D. W., Giles, W. H., New York, NY: Simon & Schuster.
Felitti, V. J., Dong, M., & Anda, R. F. (2006). Adverse Gilbert, R., Widom, C. S., Browne, K., Fergusson, D.,
childhood experiences and the association with ever Webb, E., & Janson, J. (2009). Burden and consequences
using alcohol and initiating alcohol use during adoles- of child maltreatment in high-income countries. The
cence. Journal of Adolescent Health, 38, 444. doi:10.1016/ Lancet, 373, 68–81. doi:10.1016/S0140-6736(08)61706-7
j.jadohealth.2005.06.006 Graham, J. W., Cumsille, P. E., & Elek-Fisk, E. (2003).
Duncan, R. D. (1999). Peer and sibling aggression: An Methods for handling missing data. In J. A. Schinka &
investigation of intra- and extra-familial bullying. Jour- W. E. Velicer (Eds.), Research methods in psychology
nal of Interpersonal Violence, 14, 871–886. doi:10.1177/ (Vol. 2, pp. 87–114). New York, NY: Wiley.
088626099014008005 Haynie, D. L., & McHugh, S. (2006). Sibling deviance: In
DuRant, R. H., Smith, J. A., Kreiter, S. R., & Krowchuk, the shadows of mutual and unique friendship effects.
D. P. (1999). The relationship between early age of Criminology, 41, 355–392. doi:10.1111/j.1745-9125.2003.
onset of initial substance use and engaging in multiple tb00991.x
health risk behaviors among young adolescents. Hemphill, S. A., Kotevski, A., Herrenkohl, T. I., Bond, L.,
Archives of Pediatrics & Adolescent Medicine, 153, 286. Kim, M. J., Toumbouro, J. W., & Catalano, R. F. (2011).
Ensminger, M. E., Juon, H. S., & Fothergill, K. (2002). Longitudinal consequences of adolescent bullying per-
Childhood and adolescent antecedents of substance petration and victimisation: A study of students in Vic-
use in early adulthood. Addiction, 97, 833–844. toria, Australia. Criminal Behaviour and Mental Health,
Espelage, D. L., Bosworth, K., & Simon, T. R. (2000). 21, 107–116. doi:10.1002/cbm.802
Examining the social context of bullying behaviors in Holt, M. K., Kantor, G. K., & Finkelhor, D. (2009). Par-
early adolescence. Journal of Counseling and Development, ent/child concordance about bullying involvement
78, 326–333. doi:10.1002/j.1556-6676.2000.tb01914.x and family characteristics related to bullying and peer
Espelage, D. L., & Holt, M. L. (2001). Bullying and vic- victimization. Journal of School Violence, 8, 42–63. doi:10.
timization during early adolescence: Peer influences 1080/15388220802067813
and psychosocial correlates. Journal of Emotional Abuse, Houbre, B., Tarquinio, C., Thuillier, I., & Hergott, E.
2, 123–142. doi:10.1300/J135v02n02_08 (2006). Bullying among students and its consequences
Espelage, D. L., Holt, M. K., & Henkel, R. R. (2003). on health. European Journal of Psychology of Education,
Examination of peer-group contextual effects on 21, 183–208. doi:10.1007/BF03173576
aggression during early adolescence. Child Develop- Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit
ment, 74, 205–220. doi:10.1111/1467-8624.00531 indexes in covariance structure analysis: Conventional
Espelage, D. L., Low, S., & De La Rue, L. (2012). Rela- criteria versus new alternatives. Structural Equation
tions between peer victimization subtypes, family vio- Modeling, 6, 1–55. doi:10.1080/10705519909540118
348 ESPELAGE, LOW, RAO, HONG, AND LITTLE

Jimerson, S. R., & Furlong, M. J. (Eds.). (2006). The hand- Ortega, L., & Espelage, D.L. (under review). A cluster
book of school violence and school safety: From research to analytic view of poly-victimization and psychosocial
practice. Mahwah, NJ: Lawrence Erlbaum. functioning among Hispanic adolescents. Psychology of
Johnston, L. D., O’Malley, P. M., Bachman, J. G., & Violence.
Schulenberg, J. E. (2012). Monitoring the future national Reid, J. B., Patterson, G. R., & Snyder, J. J. (Eds.). (2002).
results on adolescent drug use: Overview of key findings, Antisocial behavior in children and adolescents: A develop-
2011. Ann Arbor: Institute for Social Research, the Uni- mental analysis and a model for intervention. Washington,
versity of Michigan. DC: American Psychological Association.
Kim, M. J., Catalano, R. F., Haggerty, K. P., & Abbott, R. Roche, K. M., Ghazarian, S. R., Little, T. D., & Leventhal,
D. (2011). Bullying at elementary school and problem T. (2011). Understanding links between punitive par-
behaviour in young adulthood: A study of bullying, enting and adolescent adjustment: The relevance of
violence and substance use from age 11 to age 21. context and reciprocal association. Journal of Research
Criminal Behaviour and Mental Health, 21, 136–144. on Adolescence, 21, 448–460. doi:10.1111/j.1532-7795.
doi:10.1002/cbm.804 2010.00681.x
King, K. M., & Chassin, L. (2007). A prospective study of Roscoe, B., Goodwin, M. P., & Kennedy, D. (1987). Sib-
the effects of age of initiation of alcohol and drug use ling violence and agonistic interactions experienced by
on young adult substance dependence. Journal of early adolescents. Journal of Family Violence, 2, 121–137.
Studies on Alcohol and Drugs, 68, 256–265. doi:10.1007/BF00977037
Little, T. D., Cunningham, W. A., Shahar, G., & Wid- SAS Institute Inc. (2008). SAS/STAT 9.2 user’s guide. Cary,
aman, K. F. (2002). To parcel or not to parcel: Explor- NC: SAS Institute Inc.
ing the question, weighing the merits. Structural Schermelleh-Engel, K., Moosbrugger, H., & Muller, H.
Equation Modeling: A Multidisciplinary Journal, 9, (2003). Evaluating the fit of structural equation models:
151–173. doi:10.1207/S15328007SEM0902_1 Tests of significance and descriptive goodness-of-fit
Little, T. D., Slegers, D. W., & Card, N. A. (2006). A non- measures. Methods of Psychological Research, 8, 23–74.
arbitrary method of identifying and scaling latent vari- Schwartz, D., & Proctor, L. (2000). Community violence
ables in SEM and MACS models. Structural Equation exposure and children’s social adjustment in the school
Modeling: A Multidisciplinary Journal, 13, 59–72. doi:10. peer group: The mediating roles of emotion regulation
1207/s15328007sem1301_3 and social cognition. Journal of Consulting and Clinical
Luengo, J., Garcıa, S., & Herrera, F. (2010). A study on Psychology, 68, 670–683. doi:10.1037/0022-006X.68.4.670
the use of imputation methods for experimentation Sijtsema, J. J., Lindenberg, S., & Veenstra, R. (2010). Do
with radial basis function network classifiers handling they get what they want or are they stuck with what
missing attribute values: The good synergy between they can get? Testing homophily against default selec-
RBFNs and event covering method. Neural Networks, tion for friendships of highly aggressive boys. The
23, 406–418. doi:10.1016/j.neunet.2009.11.014 TRAILS study. Journal of Abnormal Child Psychology, 38,
Luk, J. W., Wang, J., & Simons-Morton, B. G. (2010). Bul- 803–813.
lying victimization and substance use among U.S. ado- Skara, S., Pokhrel, P., Weiner, M. D., Sun, P., Dent, C.
lescents: Mediation by depression. Prevention Science, W., & Sussman, S. (2008). Physical and relational
11, 355–359. doi:10.1007/s11121-010-0179-0 aggression as predictors of drug use: Gender differ-
Marcenko, M. O., Kemp, S. P., & Larson, N. C. (2000). ences among high school students. Addictive behaviors,
Childhood experiences of abuse, later substance use, 33, 1507. doi:10.1016/j.addbeh.2008.05.014
and parenting outcomes among low-income mothers. Slomkowski, C., Rende, R., Conger, K. J., Simons, R. L.,
American Journal of Orthopsychiatry, 70, 316–326. doi:10. & Conger, R. D. (2001). Sisters, brothers, and delin-
1037/h0087853 quency: Evaluating social influence during early and
Moretti, M. M., Obsuth, I., Odgers, C. L., & Reebye, P. middle adolescence. Child Development, 72, 271–283.
(2006). Exposure to maternal vs. paternal partner vio- doi:10.1111/1467-8624.00278
lence, PTSD, and aggression in adolescent girls and Snyder, J., Bank, L., & Burraston, B. (2005). The conse-
boys. Aggressive Behavior, 32, 385–395. doi:10.1002/ab. quences of antisocial behavior in older male siblings
20137 for younger brothers and sisters. Journal of Family Psy-
Mustanoja, S., Luukkonen, A. H., Hakko, H., Rasanen, P., chology, 19, 643–653. doi:10.1037/0893-3200.19.4.643
Saavala, H., Riala, K., et al. (2011). Is exposure to Snyder, J. J., Reid, J. B., & Patterson, G. R. (2003). A
domestic violence and violent crime associated with social learning model of child and adolescent antisocial
bullying behaviour among underage adolescent psy- behavior. In B. B. Lahey, T. E. Moffitt, & A. Caspi
chiatric inpatients? Child Psychiatry and Human Develop- (Eds.), The causes of conduct disorder and juvenile delin-
ment, 42, 495–506. doi:10.1007/s10578-011-0222-9 quency (pp. 27–48). New York, NY: Guilford Press.
Olweus, D. (1993). Bullying at school: What we know and Sourander, A., Helstela, L., Helenius, H., & Piha, J.
what we can do. Oxford, UK: Blackwell. (2000). Persistence of bullying from childhood to ado-
Olweus, D. (2002). Bullying at school. Malden, MA: Black- lescence: A longitudinal 8-year follow-up study. Child
well. Abuse & Neglect, 24, 873–881.
FAMILY VIOLENCE, BULLYING, AND SUBSTANCE USE 349

Sullivan, T. N., Farrell, A. D., & Kliewer, W. (2006). Peer Zucker, R. A., Wong, M. M., Clark, D. B., Leonard, K. E.,
victimization in early adolescence: Association Schulenberg, J. E., Cornelius, J. R., . . . Putler, L. I.
between physical and relational victimization and drug (2006). Predicting risky drinking outcomes longitudi-
use, aggression, and delinquent behaviors among nally: What kind of advance notice can we get? Alco-
urban middle school students. Development and Psycho- holism: Clinical and Experimental Research, 30, 243–252.
pathology, 18, 119–137. doi:10.1017/S095457940606007X doi:10.1111/j.1530-0277.2006.00033.x
Taiwo, T., & Goldstein, S. (2006). Drug use and its associ-
ation with deviant behaviour among rural adolescent
students in South Africa. East African Medical Journal, Supporting Information
83, 500–506.
Tharp-Taylor, S., Haviland, A., & D’Amico, E. J. (2009). Additional Supporting Information may be found
Victimization from mental and physical bullying and in the online version of this article at the
substance use in early adolescence. Addictive Behaviors, publisher’s website:
34, 561–567. doi:10.1016/j.addbeh.2009.03.012
Thornberry, T. P., Krohn, M. D., Lizotte, A. J., Smith, C. Table S1. Loading, intercept values, residuals, and
A., & Tobin, K. (2003). Gangs and delinquency in develop- R2 values for each indicator from the strong metric
mental perspective. Cambridge, UK: Cambridge Univer-
invariance model for girls.
sity Press.
Table S2. Loading, intercept values, residuals, and
Vakalahi, H. F. (2001). Adolescent substance use and
family-based risk and protective factors: A literature R2 values for each indicator from the strong metric
review. Journal of Drug Education, 31, 29–46. doi:10. invariance model for boys.
2190/QP75-P9AR-NUVJ-FJCB Table S3. Fit indices for the multiple group analy-
Voisin, D. R., & Hong, J. S. (2012). A mediational model ses of family violence—bullying perpetration—
linking witnessing intimate partner violence and bully- AOD across gender.
ing behaviors and victimization among youth. Educa- Table S4. Fit indices for the multiple group analy-
tional Psychology Review, 24, 479–498. doi:10.1007/ ses—comparing the mediating role of bullying
s10648-012-9197-8 perpetration and fighting perpetration.
Wang, J., Iannotti, R. J., & Nansel, T. R. (2009). School Figure S1. Mediational model testing the mediating
bullying among adolescents in the United States: Phys-
role of bully perpetration on the relation between
ical, verbal, relational, and cyber. Journal of Adolescent
family violence and AOD use for boys and girls.
Health, 45, 368–375. doi:10.1016/j.jadohealth.2009.03.021

You might also like