Introduction Dating Violence
Dating violence, which includes physical, sexual, psychological, or emotional
aggression, is a serious social problem affecting the well-being of college students (CDC,
2018). In college life, romantic relationships become longer and more serious, with
couples sometimes deciding to live together (Arnett, 2000). Dating violence entails
significant negative effects such as low academic achievement and various adverse health
outcomes, including physical injuries and psychological distress (Buelna, Ulloa, &
Ulibarri, 2009; Paat & Markham, 2016). Dating violence perpetration has also been
found to be a strong predictor of marital violence in later life (White, Merrill, & Koss,
2001). Thus, it is important to examine relevant correlates of aggression to prevent future
occurrence of dating violence among people in this age group – as well as those in later
life (Paat & Markham, 2016).
Research has revealed that one of the consistent predictors of dating violence
perpetration is early exposure to violence (O’Keefe, 1998; Eriksson & Mazerolle, 2015).
However, little is known about the particular correlates or mechanisms that facilitate
dating violence perpetration among college students. Thus, this dissertation aims to
address these gaps, using three empirical studies to examine the relevant correlates or
mechanisms that lead from early exposure to violence to dating violence perpetration
among college students.
1.1 Study aims
The aims of the proposed research are as follows:
Aim 1. To examine the relationship between three forms of violent socialization
during childhood and attitudes supporting interpersonal violence (Chapter 2).
Aim 2. To examine the moderating effect of pro-violence messages in the
relationship between early exposure to violence and later attitudes supporting violence, as
well as dating violence perpetration (Chapter 3).
Aim 3. To examine the possible mediating effect of attitudes supporting violence
and substance use in the relationship between early exposure to violence and dating
violence perpetration (Chapter 4).
1.2 Research questions
I will achieve the aims proposed for this dissertation research by answering the
following research questions.
Aim 1 research questions (Chapter 2).
Research Question 1: Do multiple forms of early violent socialization (advised
violence, witnessed violence, and victimized violence) from family members produce
different associations with acceptance of interpersonal violence?
Research Question 2: Do multiple forms of early violent socialization (advised
violence, witnessed violence, and victimized violence) from non-family members
produce different associations with acceptance of interpersonal violence?
Aim 2 research questions (Chapter 3).
Research Question 1: Are pro-violence messages positively related to attitudes
approving violence and/or to dating violence perpetration?
Research Question 2: Do pro-violence messages moderate the association
between exposure to violence during childhood and later attitudes approving violence, as
well as dating violence perpetration?
Aim 3 research questions (Chapter 4).
Research Question 1: Are there any significant direct and indirect pathways (i.e.,
mediation) from early exposure to violence to dating violence perpetration through
attitudes toward violence and substance use?
Research Question 2: Does a student’s gender moderate the direct or indirect
pathways from early exposure to violence to attitudes toward violence?
1.3 Overall research design and methodological approach
The overall research design of this dissertation is cross-sectional, and it includes
secondary data analysis. The dataset used is 4,533 U.S. college students – a subsample of
the International Dating Violence Survey (Status, 1999). The International Dating
Violence Survey was conducted to investigate risk factors and protective factors related
to dating violence perpetration and victimization among college students in 32 countries
(Straus, 2011). Prior to conducting any analysis, several data management tasks,
including missing data analysis and measurement testing, were completed. Univariate and
bivariate analyses are completed for all three studies to understand data distribution and
variable correlations and examine group differences. Advanced statistical approaches
include ordinary least squares regression models (Chapter 2 & Chapter 3) and path
analysis (Chapter 4). In addition, I used robust standard errors to control the cluster
effects of nested data (universities) (Chapter 3 &4) and confirmed that the data set was
balanced, and the results could be generalized through cross validation (Chapter 3).
CHAPTER 2
WHERE DO THEY LEARN VIOLENCE? THE ROLE OF
MULTIPLE FORMS OF VIOLENT SOCIALIZATION IN CHILDHOOD
1
Abstract
Early violent socialization from family and community contexts has consistently
been found to be associated with individuals' attitudes toward interpersonal violence in
adulthood. The specific sources and processes underlying children’s learning of violence,
1
Kim, J., Lee, B., and Farber, N. Submitted to Children and Youth Services Review,
10/10/2018.
however, are poorly specified and unelaborated. The current study examined how
multiple types of violent socialization during childhood differently influenced
individuals' acceptance of interpersonal violence among college students (N =3,930). The
findings of hierarchical multiple regression analyses indicated that advised violence,
witnessed violence, and victimized violence each contributed to college students’
acceptance of violence. The verbal advice toward violence from family and community
had stronger associations with on students’ acceptance of interpersonal violence than
experiences related to childhood violence victimization and witnessing violence. Given
that existing studies on violent socialization have largely focused on witnessing and
childhood victimization, our findings add the new knowledge that family and community
members’ coaching that includes pro-violence messages plays a significant role in young
college students’ acceptance of interpersonal violence. Our findings suggested that
educational interventions related to communicating skills and attitudes toward violence
that create a safe and stable environment for children may be beneficial for parents,
teachers, community members, and professionals who work closely with children. The
results may be also helpful in informing professional practitioners in colleges when they
develop social awareness programs for college students to help change their attitudes
toward violence.
1. Introduction
Interpersonal violence, including dating violence and sexual assault are serious
and widespread problems on college campuses (Alegría-Flores, Raker, Pleasants,
Weaver, & Weinberger, 2017; Kaukinen, 2014). A National Intimate Partner and Sexual
Violence Survey (2010) found that 47% of females and 39% of males experience
physical violence, sexual violence, and/or stalking by an intimate partner between the
ages of 18 and 24 (Black et al., 2011). Although differences in these estimates may be
due to variance in the definition of dating violence, research consistently shows that 20%
to 50% of college couples report being the victim of at least one act of dating violence
(Baker, Stith, & Trauma, 2008; Shorey, Cornelius, & Bell, 2008).
Many researchers are increasingly examining causes of interpersonal violence
(Banyard, Cross, & Modecki, 2006). They have determined that one possible factor is a
belief system supporting the use of interpersonal violence (Coker et al., 2015; Herrenkohl
& Jung, 2016; Herrero, Rodríguez, & Torres, 2017) and such beliefs are significantly
associated with experiences of violence in childhood (Bryant & Spencer, 2003; Flood &
Pease, 2009). Social learning theories suggest that early experiences of violence infuse
cognitive beliefs condoning the use of interpersonal violence and increase the likelihood
that individuals will engage in violence (Eriksson & Mazerolle, 2015). This suggests
further that early learning of violence may prompt not only victimization and perpetration
of violence but also individuals’ approval of violence later in life as an adult (Delaney,
2015; Gage, 2016; Lichter & McCloskey, 2004).
Childhood socialization toward violence may take on multiple forms, including
verbal reinforcement of violence, observing/witnessing violence, and being the victim of
violent acts from others such as family and neighbors (Delaney, 2015; Mattingly &
Straus, 2008; Owens & Straus, 1975). These events may coincide; however, each form of
violence may have a different effect on individuals (Bacchini, Affuso, & Aquilar, 2015).
Existing studies mainly focus on the effect of a particular form of violent socialization
such as physical abuse or witnessing family violence (Jennings et al., 2014; Steel,
Watkins, & DiLillo, 2017). The effects of various forms of violent socialization on
individual’s belief system have rarely been assessed simultaneously. Our study aims to
fill this gap by examining multiple forms of violent socialization from family and
community members to see if there are different effects among them on college students’
acceptance of violence.
1.1. Literature review
Socialization is defined as individuals learning and internalizing social skills,
knowledge, behavior patterns and values of a culture to adequately interact with other
people and more generally, to function in the larger society. In particular, a child’s
socialization occurs through various social institutions. Picked up from family, peers,
television, and school, learned habits are then often transmitted from one generation to
the next (Maccoby, 1984). This section explains the process of violent socialization
through social learning theory and the various forms of violent socialization.
Social learning theory & multiple forms of violent socialization
Akers’ social learning theory includes four types of processes: differential
association, definitions, differential reinforcement, and imitation. Differential association
means that individuals interact with certain “associations,” including family members,
friends, and neighborhood; and, that those individuals can promote favorable or
unfavorable definitions of illegal or law-abiding behaviors (Akers, 1996; Akers, Krohn,
Lanza-Kaduce, & Radosevich, 1979). Several factors modulate the strength of the
association’s influence on an individual’s behavior. These include how early (Priority),
how long (Duration), how frequent (Frequency) and how close (Intensity) the associated
influence is present (Akers, 1996). According to Akers et al. (1979) and Pratt et al.
(2010), people learn attitudes, meanings and hence, “definitions” for certain behaviors as
good or bad from their environment (Akers et al., 1979; Pratt et al., 2010). If a person
possesses a positive or neutral definition for a certain behavior, the person will be more
likely to engage in the behavior. Differential reinforcement refers to how the balance
between expected and actual reward or punishment influences personal behavior (Akers,
1996; Akers et al., 1979). The more certain behaviors are reinforced by rewards, the more
such behaviors are repeated, and the reverse when a behavior is associated with a
punishment (Akers, 1996; Akers et al., 1979). Finally, imitation refers to engaging in
certain behavior after observing some models of the behavior (Akers & Jennings, 2009;
Pratt et al., 2010).
Straus, Mouradian, and DeVoe (1999) specified violent socialization as a result of
three processes: advised violence, witnessed violence, and victimized violence (Straus,
Mouradian, & DeVoe, 1999). Advised violence refers to have verbal coaching to instigate
an aggressive response or retaliation to (cope with) a situation. Witnessed and victimized
violence refer to experiences observing violence and experiencing violence as victims.
This study adapted the conceptualization of multiple violent socialization defined by
Straus et al. (1999).
Advised violence
Children might learn that violent behavior is an appropriate response to solving
problems through family and community members’ justification for use of violence
(Chen, Flores, Shetgiri, & studies, 2016; Herrenkohl & Jung, 2016). Straus et al. (1999)
called such violent socialization “advised violence,” meaning gaining positive feedback
on violent behavior. This type of socialization is learned through modeling and
reinforcement (Bandura, 1978). Previous studies supported the association, that is,
encouragement by parents or other adult relatives toward aggressive peer interactions
may have considerable effects on a child’s willingness toward violence and retaliatory
attitudes (Chen et al., 2016; Copeland-Linder et al., 2007; Johnson, Finigan, Bradshaw,
Haynie, & Cheng, 2011; Orpinas, Murray, & Kelder, 1999; Solomon, Bradshaw, Wright,
& Cheng, 2008). For example, youth’s perceptions of parental attitudes toward fighting
are significantly associated with their own aggressive behavior more: fights at school,
greater injuries in fight, and being more likely to carry a weapon than among youth who
did not perceive their parent as supporting fighting (Orpinas et al., 1999). Specifically,
youths who are more aggressive perceived that their parents are more supportive of
fighting and interpret their parent’s supportive attitude as approval for their aggressive
behaviors (Orpinas et al., 1999).
Such advised violence might be also experienced by children through proviolence
messages or community norms. For example, disadvantaged communities may have
informal rules or expectation of violent retribution for disrespect and perceived attacks as
self-protection (Anderson, 1999). Such community norms could be conveyed to youth
through their interaction with peers and adults in communities. In particular, the attitudes
of violence among peers play a critical role in on children’s views of violence, providing
models for behavior and shaping individuals’ norms and values of violence (Bernburg &
Thorlindsson, 2005; Harding, 2009; Herrenkohl & Jung, 2016; Zimmerman
& Messner, 2011).
Witnessed violence
As another process of violent socialization, children can be violently socialized
through vicarious observations (Bandura, 1978). When children observe rewarding of
violent behaviors, they tend to have a positive assessment of violence. (Akers, 1996;
Bandura, 1978). For example, if children observe that their father’s violence leads to
mother’s changed behaviors such as submission toward the father, they might think
violence is a useful tool to achieve what they want. In this context, witnessing violence
leads to aggressive behavior through the mediation of such social learning mechanisms
(Schwartz & Proctor, 2000). Several studies have found that children who grow up in
families in which they have witnessed interparental violence are more likely to imitate
and tolerate these behaviors than are children from nonviolent homes (Allwood & Bell,
2008; Lichter & McCloskey, 2004; O'Keefe, 1998). Outside of the immediate family
context, youth who witness violence in their neighborhood might then conclude that
violent behavior is a valid way to resolve problems in their community context
(Anderson, 1999; Sampson, Raudenbush, & Earls, 1997).
Victimized violence
Being a victim of family and community violence are associated with children’s
acceptance of violence and aggressive behaviors (Fosco, DeBoard, & Grych, 2007;
Temple, Shorey, Tortolero, Wolfe, & Stuart, 2013). In particular, the effect of violent
socialization in the context of physical child abuse perpetuating further violent offenses
has been studied for decades – a concept dubbed the “Cycle of Violence” (Fagan, 2005;
Milaniak & Widom, 2015; Widom & Maxfield, 2001). Several empirical studies support
the Cycle of Violence theory by examining the effect on the victim’s involvement of
various types of offense as well as violent offense (Fagan, 2005; Widom & Maxfield,
2001; Wolfe, Wekerle, Scott, Straatman, & Grasley, 2004). A meta-analysis by Wolfe et
al. (2003) examined 41 studies which found associations between a child’s exposure to
domestic violence including child abuse and his or her internal (depression and anxiety)
and external behavioral problems (delinquency and aggression). This study concluded
that children who were exposed to domestic violence have a high rate of either type of
emotional and behavioral problems. Other studies also showed that childhood
maltreatment predicted attitudes for justifying dating violence and aggressive behaviors
toward dating partners (Manchikanti Gómez, 2011; Wolfe et al., 2004).
Researchers have suggested that witnessing violence and being a victim of
violence, which are considered exposure to violence, have strong associations with
children’s aggressive behaviors. However, some researchers have made a distinction
between the effects of witnessed violence and victimized violence. They argue that victim
experiences might not be associated with children’s positive attitude of violence and
aggressive behaviors. For example, Schwartz and Proctor (2000) found that violent
victimization was associated with negative emotional dysregulation while witnessed
violence was connected to aggressive action (Schwartz & Proctor, 2000). They suggested
that direct victimization might include the central mediational processes that would
contribute to disruptions in emotional regulation instead of positive assessment of
violence. In this context, we assumed that witnessing violence and being the victim of
violence might play different roles in cognitive beliefs about violence.
1.2. The current study
Several gaps in the existing literature indicate the importance of our study. Past
research on violent socialization has focused predominantly on a particular type of
violent socialization (for example, witnessing family violence), rather than exploring
multiple forms of violent socialization from family and community context
simultaneously. Examining the effects of a particular form alone does not fully capture
violent socialization. Thus, our study makes an important contribution by exploring each
role and the different effects of multiple forms of early violent socialization. In addition,
few studies have focused on the relationship between early violent socialization and
acceptance of violence in young adulthood. Considering the possibility that an attitude of
violence might be a significant mechanism in the relationship between violence
experience and actual violence perpetration, it is important to identify the effect of violent
socialization on acceptance of violence. Thus, our study aims to examine college
students’ cognitive beliefs about interpersonal violence and the effect of multiple forms
of early violent socialization. Based on theoretical frameworks and previous empirical
studies, two main research questions were addressed in this study:
Research Question 1: Do multiple forms of early violent socialization (advised
violence, witnessed violence, and victimized violence) from family members produce
different associations with acceptance of interpersonal violence?
Research Question 2: Do multiple forms of early violent socialization (advised
violence, witnessed violence, and victimized violence) from non-family members
produce different associations with acceptance of interpersonal violence?
2. Method
2.1 data
The data for this study were obtained from the International Dating Violence
Study (IDVS). The IDVS was conducted between 2001 and 2006 in 32 countries in seven
world regions, including Africa, Asia, Australia, Europe, Latin America, the Middle East,
and North America (Straus, 2010). The IDVS contains a convenience sample of
undergraduates enrolled in university courses (mostly criminology, sociology,
psychology, or family studies) taught by a designated research consortium member.
Across the participating nations, the study yielded a total sample of 17,404 individuals.
However, only the 14,252 respondents who reported being in a relationship that lasted at
least 1 month are used in the current study. After removing approximately 4% of these
cases because of implausibly high numbers of sexual assaults/injuries or inconsistent
responses and using imputation for missing data (Medeiros, 2007), the final sample size
was 13,659 (Hines & Straus, 2007).
2.2 sample
This study utilized data from 4,533 U.S. college students in the IDVS. U.S.
college students in the sample were from 17 universities in the U.S. (including Illinois
State University, Indiana State University, Grambling University, University of Southern
Mississippi, Jackson State University, University of New Hampshire, John Jay College,
University of Cincinnati, Dickinson College, University of Tennessee, University of
Texas at El Paso, Texas Tech University, Rice University, Stephen F. Austin State
University, Utah State University, and Howard University). Focusing on young adults,
this study restricted the analysis to students 18 to 23 years old. After excluding the cases
with age range 34 to 45 and missing values, a total of 3,930 students were included in the
analysis.
2.3. Measures
Dependent Variable. Acceptance of violence. Acceptance of violence was
measured using ten items from the Violence Approval Scale of the Personal and
Relationship Profile (PRP) (Straus, Hamby, Boney-McCoy, & Sugarman, 1999). It was
designed to assess the extent to which respondents endorse the use of physical violence in
various social contexts. Among ten items, four were about approval of violence within
the family such as a husband slapping his wife’s face; three were about stereotypical male
violence, such as the view that a man should not walk away from a fight; and three were
about sexual aggression, such as whether there were circumstances when it would be
acceptable for a man to force his wife to have sex (Mattingly & Straus, 2008). The
response categories for all scales were from 1 (strongly disagree) to 4 (strongly agree).
The range of this scale was 10 to 40 points with higher scores indicating higher levels of
acceptance of violent behaviors. The alpha coefficient for this scale was .70 with the
overall IDVS student sample (Straus, 2011) and .70 for the current study as well.
Independent variables. Advised violence was conceptualized as respondents’
experience of how much they received pro-violence advice during childhood. Advised
violence from family was measured based on participants’ response to the question,
“When I was a kid, my father or mother told me to hit back if someone hit me or insulted
me”. Advised violence from non-family person was measured based on the question,
“When I was a kid, people (adults or kids) who were not part of my family told me to hit
back if someone hit me or insulted me.” The response categories to these questions ranged
from 1 (strongly disagree) to 4 (strongly agree) and high scores indicated higher levels of
exposure to violence during childhood.
Witnessed violence was conceptualized as respondents’ experience of how much
they witnessed violence during childhood. Witnessed violence from family was measured
based on participants’ response to the question, “When I was a kid, I saw my mother or
father kick, punch, or beat up their partner”. Witnessed violence from non-family person
was measured based on the question, “When I was a kid, I often saw kids who were not
in my family get into fights and hit each other.” The response categories to these
questions ranged from 1 (strongly disagree) to 4 (strongly agree) and high scores
indicated higher levels of exposure to violence during childhood.
Victimized violence was conceptualized as respondents’ experience of how much
they experienced violence during childhood. Victimized violence from family was
measured based on participants’ response to the question, “When I was less than 12 years
old, I was spanked or hit a lot by my mother or father”. Victimized violence from
nonfamily person was measured based on the questions, “When I was a kid, people
(adults or kids) who were not part of my family pushed, shoved or slapped me, or threw
things at me”. The response categories to these questions ranged from 1 (strongly
disagree) to 4 (strongly agree) and high scores indicated higher levels of exposure to
violence during childhood.
Covariates. We controlled for age, gender, family’s socioeconomic status, and
parent’s marital status in all analyses. In particular, family’s socioeconomic status was a
computed z score based on the mean site distribution of socioeconomic status derived
from the following variables: father’s education, mother’s education, and family income
(Hines & Straus, 2007).
2.4. Analysis
All analyses were conducted using SAS v9.1.3. Descriptive statistics (univariate
and bivariate analyses) were used to describe the distribution of the dependent variables,
independent variables, and covariates. Comparisons were conducted using Pearson’s
Chisquare tests for categorical variables and analysis of variance (ANOVA) tests for
continuous variables. Hierarchical multiple regression analysis was conducted to test the
association between early violent socialization and acceptance of violence (RQ1 & RQ2).
3. Results
3.1. Sample characteristics
Descriptive statistics of the sample (N=3,930) can be found in Table 2.1. College
students included in the study sample were more often girls than boys (68% vs. 32%).
The mean age was 20.1 years (SD = 0.03), and most students lived in families of slightly
higher-than-average socioeconomic status (SES) (M =0.06, SD = 0.02) with two parents
rather than a single parent (66% vs. 34%).
3.2. Acceptance of interpersonal violence and early violent socialization
Table 2.2 shows the acceptance of interpersonal violence and early violent
socialization. For the overall population, acceptance of interpersonal violence was 19.3
(SE = .06). Comparing male students and female students, male students were more
Table
17
2.1: Sample characteristics by student gender, International Dating Violence Study
(IDVS), % or Mean (SE)
Total Male Female Range Skew
Note. a. The variables are dummy coded and can be interpreted as proportions, b. t-test, c.
Chi-square test, d. p < 0.05, e. Parents never married to each other and not living together
or parents never married to each other and living together.
approving of interpersonal violence (p <.05). In the family context, students reported that
they experienced advised violence (1.98, SE=.02), victimized violence (1.33, SE=.01),
and witnessed violence (1.93, SE=.01) in that order. Male students reported experiencing
more advised violence and victimized violence in the family during childhood than
female students (p < .05). In the non-family context, students reported that they
experienced witnessed violence (2.50, SE=.01), advised violence (2.26, SE =.02), and
victimized violence (1.50, SE=.01). Similarly, male students experienced more advised
violence, witnessed violence, and victimized violence from non-family members
including neighbors and peers than female students (p < .05).
Gender a
Age (in years) b,d
20.1
(0.03)
20.3
(0.04)
20.0
(0.13)
18 - 23
0.56
Family SES b,d
0.06
(0.02)
0.17
(0.03)
0.01
(0.02)
-3.1 - 2.9
-0.31
Parents’ marital status a,c
Married
66%
68%
65%
-
Divorced or separated
25%
25%
26%
-
Separated by death
4%
4%
4%
-
Other e
5%
4%
5%
-
(
N
=
3,930)
(
N
=
1,277)
(
N
=
2,653)
-
32
%
68
%
Table
18
3.3. Effect of early violent socialization on the acceptance of interpersonal violence
Hierarchical multiple regression tested whether each form of early violent
socialization (advised violence, witnessed violence, victimized violence from family and
2.2: Bivariate associations between focal variables and gender, Mean (SE)
Total (N=
3,930)
Male
(N=
1,277)
Female
(N=
2,653)
Range
Skew
DV: Acceptance of
Interpersonal Violence b,d
19.3
(0.06)
20.8
(0.12)
18.6
(0.08)
10 - 36
0.13
IV: Violent Socialization
Family
Advised violence b,d
1.98
(0.02)
2.22
(0.03)
1.87
(0.02)
1 - 4
0.54
Witnessed violence b
1.33
(0.01)
1.33
(0.02)
1.33
(0.02)
1 - 4
2.34
Victimized violence b,d
1.93
(0.01)
2.07
(0.03)
1.88
(0.02)
0.66
Non-family
Advised violence b,d
2.26
(0.02)
2.62
(0.03)
2.08
(0.02)
1 - 4
0.12
Witnessed violence b,d
2.50
(0.01)
2.75
(0.02)
2.38
(0.02)
1 - 4
-0.08
Victimized violence b,d
1.50
(0.01)
1.81
(0.03)
1.35
(0.01)
1 - 4
1.52
Note. a. The variables are dummy coded and can be interpreted as proportions, b. t-test,
c. Chi-square test, d. p < 0.05.
non-family members during childhood) predicted the acceptance of interpersonal
violence (Table 2.3).
Model 1 [F (6, 3923) = 57.71; p < .001] contained demographic control variables,
including gender, age, family SES, and parents’ marital status, and explained 8% of the
variance in acceptance of violence. As can be seen, all demographic variables
Table
19
significantly predicted the acceptance of interpersonal violence. That is, students who are
male, younger age, have lower family SES, and live with single parents are more likely to
accept interpersonal violence than those who are not (p ≤ .01).
2.3: Hierarchical multiple regression on the acceptance of interpersonal violence (N
= 3,930)
Gender
Female
-.263(.137)***
-.179(.125)***
-.143(.129)***
Age (in years)
-.041(.038)**
-.045(.034)**
-.046(.034)**
Family SES
-.056(.066)***
-.009(.059)
.001(.059)
Parents’ marital status
Divorced or
separated
.060(.327)***
.034(.293)*
.029(.290)*
Separated by death
.041(.150)**
-.019(.138)
-.022(.136)
Others a
.092(.310)***
.024(.282)
.014(.279)
Family
Advised Violence
.353(.063)***
.280(.070)***
Witnessed Violence
.051(.079)***
.043(.079)**
Victimized Violence
.181(.064)***
.155(.065)***
Non family
Advised Violence
.114(.074)***
Witnessed Violence
.088(.071)***
Victimized Violence
.008(.075)
R2
.081
.265
.285
Δ R2
.081
.184
.019
F (df)
57.715
(6, 3923)
157.390 (9,
3920)
129.842
(12, 3917)
Δ F
57.715***
327.885***
34.935***
*p < 0.05, **p < 0.01, ***p < 0.001. Note. a. Parents never married and are not living
together, or parents never married but are living together.
Model 2 [F (9, 3920) = 157.39; p < .001], which contained demographic variables
and family forms of early violent socialization (advised violence, witnessed violence, and
Model 1
β
(
)
SE
Model 2
β
)
SE
(
Model 3
β
(
SE
)
Table
20
victimized violence), accounted for an 18.4% increase in the variance of the acceptance
of interpersonal violence compare to Model 1 (R2 = 0.265, Δ R2 = 0.184). As shown in
Table 2.2, all types of early violence socialization from family were significantly and
positively associated with students’ acceptance of interpersonal violence (p ≤ .001).
Model 3 [F (12, 3917) = 129.84; p < .001], which contained independent variables
that were contained in Models 1 and 2 and non-family forms of early violent socialization
21
(advised violence, witnessed violence, and victimized violence), accounted for a 2%
increase in the variance of the acceptance of interpersonal violence compare to Model 2
(R2 = 0.285, Δ R2 = 0.019). Advised violence and witnessed violence experiences from
non-family persons were significantly and positively associated with students’ acceptance
of interpersonal violence (p ≤ .001); however, victimized violence experience was not
significantly associated with students’ acceptance of interpersonal violence. All variables
of violent socialization from family still remained significant predictors.
4. Discussion
The current study aimed to examine multiple forms of violent socialization from
family and neighbors to see if there were different effects among them on college
students’ acceptance of violence. Overall, our findings showed that the multiple forms of
early violent socialization had a significant impact on college students’ beliefs about
violence. This finding was consistent with social learning theories which suggested that
experiences of violence in childhood influence individuals’ approval of violence (Bryant
& Spencer, 2003; Flood & Pease, 2009; O'Keefe, 1998; Tontodonato & Crew, 1992).
Additionally, compared to violent socialization from non-family members, socialization
processes through family members largely influenced college students’ cognitive beliefs
regarding using violence. This means, not surprisingly, that family context played a
robust role in constructing children’s beliefs and attitude of violence. Given that if a
person has an early, long, frequent, and close association, the association’s effect on the
person’s behavior is stronger (Akers, 1996), it was understandable why family had an
influential role on children’s socialization than neighbors.
22
However, the findings showed that multiple forms of violent socialization had
different effects on college students’ acceptance of violence. Specifically, socialization
processes through verbal advice about violence from both family and non-family
members were stronger predictors of students’ acceptance of interpersonal violence than
experiences related to childhood violence victimization and witnessing violence. Our
findings add the new knowledge that early advised violence from family and community
members was a stronger predictor of young college students’ acceptance of interpersonal
violence than witnessed violence and victimized violence.
Violent socialization from family and acceptance of violence
The first research question we addressed was whether multiple forms of violent
socialization from family members had a specific association with acceptance of
interpersonal violence among college students. Our findings showed that all kinds of
socialization processes (advised violence, witnessed violence, and victimized violence
from family members) were significantly related to the acceptance of interpersonal
violence among college students (p<.001); however, each association had a different level
of strength. For example, advised violence had a crucial impact on acceptance of violence
in comparison to the other types of socialization. This result supported the concept that
parents’ values, attitudes, and behaviors strongly influenced children’s aggressive
attitudes and behaviors (Duman & Margolin, 2007; Snyder & Patterson, 1995). In some
situations, including provocations, parents or family members might encourage their
children to respond aggressively. Also, some parents might think it is good for their
children, particularly boys, to learn to fight. Such verbal permission and encouragement
regarding violence might encourage children to accept or condone violence against
23
others, which in turn may lead to permission for interpersonal violence in adulthood. Our
findings were consistent with prior research confirming the relationship between parental
verbal directions toward violence and children’s aggressive behaviors (Copeland-Linder
et al., 2007; Orpinas et al., 1999; Solomon et al., 2008).
In addition, our findings indicated that students who witnessed violence during
childhood and had the childhood experience of being a violence victim were more likely
to accept interpersonal violence than those who did not. Witnessing violence early in life
influenced not only attitudes toward violence (Lichter & McCloskey, 2004) but also
subsequent behavioral outcomes (Roberts, Gilman, Fitzmaurice, Decker, & Koenen,
2010; Schwartz & Proctor, 2000). Our findings were consistent with other studies
reporting a strong association between childhood witnessing of intimate partner violence
and adulthood s toward attitude and perpetration of partner violence (Gage, 2016;
Roberts et al., 2010; Wolfe, Crooks, Lee, McIntyre-Smith, & Jaffe, 2003). Numerous
studies also have reported that the early experience of being a victim of family violence
was a crucial predictor of violence in adulthood (Ehrensaft et al., 2003; Gover et al.,
2011) because abusive parents, family members, and community members model
aggressive attitudes and perceptions as well as rationalizations that justify violence for
children.
Violent socialization from non-family and acceptance of violence
The second research question was whether multiple form of violent socialization
from non-family members had a specific association with acceptance of interpersonal
violence among college students. Compared to the effect of family members, the effect of
socialization from non-family members on acceptance of violence was weaker but we
24
still see there were significant associations. Similarly, the finding indicated that advised
violence from non-family member was a higher predictor of students’ acceptance of
interpersonal violence than experiences related to direct and indirect violent experience
(p<.001). This result implied that even though children did not experience direct or
indirect violence in the communities, there was enough possibility that community
residents’ pro-violence attitude largely influenced children’s norms of violence (Harding,
2017). The finding also showed that witnessed violence from non-family members was
associated with students’ acceptance of interpersonal violence (p<.001). These findings
were consistent with the previous studies confirming the effect of violent environmental
context on children’s attitude of violence (Anderson, 1999).
In contrast, the childhood experience of being a victim of violence from
nonfamily members was not significantly associated with the acceptance of interpersonal
violence. This finding might contradict the findings of previous studies. Studies have
reported that the early experience of being a victim of community violence was a crucial
predictor of violence in adulthood (Eitle, Turner, & Delinquency, 2002; Gorman-Smith &
Tolan, 1998) because community members modeled aggressive attitudes and perceptions
as well as rationalizations that justify violence for children. There might be alternative
explanations to understanding our results. First, when we examined the correlation
between victimized violence and acceptance of violence, there was a positive correlation
between two variables (r =. 207, p ≤ .001). However, in the regression analysis, when we
included three forms of violent socialization in the model, victimized violence was not
significantly associated with the acceptance of violence. This might mean that victimized
violence is associated with an acceptance of violence, however, considering the effect of
25
other forms (i.e. advised violence and witnessed violence) together on acceptance of
violence, the effect of victimized violence might be reduced or disappear. Considering
children’s experiences of victimized violence (1.5, SE=.01) is also less than other types of
violent socialization, witnessing violence (2.25, SE=.01) and advised violence (2.26,
SE=.02), experiences of victimization might not have a strong effect on their acceptance
of violence than form. Lastly, studies argue that violent victimization is more strongly
associated with disruptions in emotion regulation such as depression and PTSD than
individual’s attitude toward violence or aggressive behavior (Schwartz & Proctor, 2000).
Schwartz and Proctor (2000) suggested that direct victimization may include the central
mediational processes that would contribute to disruptions in emotion regulation unlike
witnessed violence which is most likely to lead to learned positive evaluations of violent
behavior. Thus, the effect of victimization might be weaker than other forms of violent
socialization on acceptance of violence.
5. Implications
Our findings, consistent with social learning theory, indicated that college
students’ attitude toward violence might be learned and reinforced through environmental
contexts. Thus, it is critical to provide early intervention efforts to prevent the effect of
childhood socialization to violence on the potential attitudes toward and behaviors of
interpersonal violence. This study highlights the need for social workers, teachers,
policymakers, and other individuals who work closely with children, their family, and
community members to be trained to recognize forms of early socialization to violence.
Direct verbal advice about violence and exposure to violence in a family and
neighborhood environment can serve as lessons and models for the children’s attitudes
26
toward and behaviors of violence, which later affect their acceptance of interpersonal
violence in adulthood. Therefore, educational intervention programs to create a safe and
stable environment for children may help them avoid violence and abuse when they
become adults. Parents should take a special responsibility for the discourse on violence.
Verbal skills and creating a non-violent environment for children is an effective way to
prevent potential violent attitudes and behaviors among college students. In addition,
children living in neighborhoods where they have seen or experienced other members of
the community intervening or discouraging the use of violence may feel that their
community is more likely to intervene, support nonviolent behaviors, or provide sources
of support and protection (Garthe, Gorman‐Smith, Gregory, & E. Schoeny, 2018). These
findings imply that family and non-family members can play an important role in
creating and teaching children to value a safe community that does not justify violence.
6. Limitations and suggestions for future studies Our study has
several limitations. First, this study used a cross-sectional exploratory research design
and the associations between variables are not causal. It seems valuable to develop this
research further with a longitudinal study in order to explain causal relationships among
the variables including early violent socialization (advised violence, witnessed violence,
and victimized violence) and the acceptance of violence. In addition, the sample was
collected through convenience sampling, which may impact generalizability given
nonrandomization. Specifically, the sample was selected from students enrolled mostly in
criminology, sociology, psychology, and family studies courses; thus, it cannot accurately
represent all college students. Also, gathering data from these departments could explain
why more female students participated in this study. For female college students with
27
slightly below average socioeconomic status, their early violence socialization could be
different from male college students from different socio- economic backgrounds. Future
studies should be tested using a more representative sample with more representative
characteristics of the college student population in different contexts. Furthermore, the
questions regarding early violent socialization and acceptance of violence were focused
on physical violence. However, it is important to consider that violence includes diverse
forms including psychological abuse. Thus, our measurements might not fully measure
the constructs of violent socialization and acceptance of violence. Also, the measurement
of violence socialization needs to measure its prevalence and severity and so future
studies should further be considering these limitations of measurement. Lastly, even
though family and community have interactive effects on children’s socialization, our
study did not take into account that complex dynamic, particularly the influence of family
on socialization of coping with community violence. Future studies may incorporate the
interactive dynamic of family and community context on children’s socialization process.
CHAPTER 3
EARLY EXPOSURE TO VIOLENCE AND DATING VIOLENCE
PERPETRATION: THE ROLE OF PRO-VIOLENCE MESSAGES
2
2
Kim, J. To be submitted to Journal of Interpersonal Violence.
28
Abstract
Dating violence is a serious and prevalent problem among college-aged couples.
Empirical studies based on social learning theories have found a strong relationship
between early exposure to violence and dating violence perpetration in later life.
However, substantially less research has examined the factors that reinforce learning of
violence. The purpose of the present study was to examine whether pro-violence
messages moderate the effect of exposure to violence on attitudes toward violence and
dating violence perpetration among U.S. college students (N =3,302). Findings showed
that pro-violence messages are associated with attitudes toward violence among both
male and female students, but that such messages are associated with increased dating
29
violence perpetration only among female students. In addition, pro-violence messages
moderate the relationship between exposure to violence and attitudes accepting violence
among only female students; however, the messages also moderated the effect of
exposure to violence on dating violence perpetration among male students. Implications
for future research and policy are discussed.
1. Introduction
Due to high rates of dating violence among college students and the associated
negative outcomes, researchers are increasingly turning their attention to the causes and
correlates of dating violence perpetration. Dating violence studies have found that
exposure to violence during childhood is strongly associated with beliefs supporting
violence and dating violence perpetration (Black et al., 2015; Eriksson & Mazerolle,
2015; Ramsay, Steeves, Feng, & Farag, 2017). Such the reason, researchers have
explored which factors reinforce violence acquisition in individuals who witness violence
and/or become victims of violence (Thomas, Caldwell, Assari, Jagers, & Flay, 2016).
Some studies have revealed that pro-violence messages from parents and peers,
such as support for fighting and tolerant attitudes toward violence, are significantly
30
related to children’s attitudes toward violence and aggressive behaviors (Garthe, Gorman‐
Smith, Gregory, & E. Schoeny, 2018; Walters, 2017). However, only a few studies have
examined how pro-violence messages adjust or reinforce the effects of childhood
exposure to violence on attitudes supporting violence and dating violence perpetration
(Garthe et al., 2018). In addition, considering that the learning process of violence differs
according to gender (Debowska, Boduszek, Jones, Willmott, & Sherretts, 2017; Spencer,
Morgan, Bridges, Washburn-Busk, & Stith, 2017), research is needed to explore the
mechanism of violence acquisition through focusing on gender differences. Thus, the
current study investigates the possibility that pro-violence messages moderate the
relationship between exposure to violence, and attitudes toward violence and dating
violence perpetration among male and female college students.
2. Literature review
Social learning theory and pro-violence messages
Social learning theory suggests that children imitate violent behaviors through
direct and indirect exposure to violence, such as witnessing violence and direct
victimization (Akers & Jennings, 2009; Bandura, 1978). The theory also emphasizes that
learning can be facilitated by acquiring attitudes of certain behaviors – whether or not
those behaviors are appropriate (Akers & Jennings, 2009). For example, family and
community members’ supportive messages for fighting might play a critical role in
facilitating violence learning (Copeland-Linder et al., 2007; Solomon, Bradshaw, Wright,
& Cheng, 2008). Social learning theory (Akers, 1996) also postulates that observed or
experienced behaviors may be reinforced through reward or punishment. That is, the
more often certain behaviors are reinforced by rewards, the more such behaviors are
31
repeated; and in the same way, certain behaviors are eliminated through punishment
(Akers, 1996). Violence can be reinforced through parental support for fighting and
praise for aggressive peer interactions from family, peers, and community members
(Johnson, Finigan, Bradshaw, Haynie, & Cheng, 2011; Orpinas, Murray, & Kelder, 1999;
Wright & Fagan, 2013).
The effect of pro-violence messages on violence attitudes and violent behaviors
Even though many researchers have found a strong association between early
exposure to violence and aggressive behaviors (Eriksson & Mazerolle, 2015; Foshee,
Bauman, & Linder, 1999; Gover, Kaukinen, & Fox, 2008), few studies have examined
how pro-violence messages such as support for fighting, predict or adjust an individual’s
attitude toward violence and violence perpetration. Existing studies have focused mainly
on the role of parental and peer support for fighting in producing general aggressive
patterns among adolescents. These studies reveal that encouragement by parents or other
adult relatives for aggressive peer interactions may have considerable effects on
children’s willingness to engage in violence and adopt retaliatory attitudes (Chen, Flores,
Shetgiri, & studies, 2016; Copeland-Linder et al., 2007; Solomon et al., 2008; Wright &
Fagan, 2013). For example, youths who believe that their parents support fighting are
more likely to get into fights at school, to receive greater injuries during fights, and to
carry weapons (Orpinas et al., 1999). Two longitudinal studies also examined the role of
caregivers’ messages/coaching supporting violence and nonviolence on adolescents’
aggression. They found that perceived parental support of fighting is related to higher
frequencies of aggressive behavior, and perceived parental support of nonviolent
32
responses to conflict is related to lower levels of aggression (Farrell et al., 2012; Kliewer,
Parrish, Taylor, Jackson, Walker, & Shivy et al., 2006).
In addition to parents and family members, children might also receive support for
aggressive coping from their peers or from community norms such as tolerance of
violence (Thomas et al., 2016; Walters, 2017; Wright & Fagan, 2013). For example,
disadvantaged communities may have informal rules or expectations of violent
retribution for disrespect, and individuals in these communities may view their violent
reaction toward others as self-protection (Anderson, 1999). Such community norms can
be conveyed to youths through interactions with peers and adults in their communities.
Peers’ attitudes toward violence also play a critical role in children’s views of violence,
providing models for behavior and shaping their norms and values regarding violence
(Bernburg & Thorlindsson, 2005; Harding, 2009; Herrenkohl & Jung, 2016; Zimmerman
& Messner, 2011).
Regarding dating violence perpetration, few studies have examined the effect of
support for fighting, including pro-violence messages, on dating violence perpetration,
and these studies show mixed findings (Garthe et al., 2018). In one longitudinal study,
Garthe et al. (2018) found that parental support for fighting did not predict changes in
dating violence perpetration across time within a sample of adolescents. In addition, only
a few studies have examined the effects of parental and peer support for fighting on
individuals who have witnessed violence or who have been victimized (Copeland-Linder
et al., 2007; Thomas et al., 2016). In a study of African-American adolescents, Thomas et
al. (2016) examined how witnessing physical violence is related to violent behavior and
how peer and parent expectations about violent or nonviolent behaviors adjust the
33
linkage. Thomas et al. (2016) found that parents’ expectations of nonviolence were not
directly related to violent behaviors among these adolescents; rather, parents’ protective
roles strengthened boys’ self-efficacy in avoiding violence. However, expectations of
violence from the boys’ peers directly predicted violent behaviors.
Gender differences
Although findings are mixed (Holt, Buckley, & Whelan, 2008), most studies
suggest that exposure to violence may affect boys and girls differently. In general, boys
who have been exposed to violence are more likely to demonstrate externalizing
behaviors, including violent behaviors; while girls tend to display more internalizing
behaviors, such as emotional distress or depression (Evans, Davies, & DiLillo, 2008; Holt
et al., 2008). In terms of attitudes toward violence, studies have found that male students
who experienced violence within the family or from outside are more likely than female
students to approve of and adopt violent behaviors (Bacchini, Affuso, & Aquilar, 2015;
Mattingly & Straus, 2008).
Such gender differences can be seen in research on parental and peer support for
fighting. Copeland-Linder et al. (2007) examined the factors associated with retaliatory
attitudes among adolescents who had been assaulted by peers. They found that aggressive
peer interactions influenced the retaliatory attitudes of boys, but did not influence the
retaliatory attitudes of girls. This could mean that pro-violence messages are more likely
to influence boys than girls. Meanwhile Garthe et al. (2018) found that for adolescents of
both genders, perceptions of parental support for fighting were positively associated with
the perpetration of physical dating violence. However, for females alone, perceptions of
34
parental support for nonviolent responses to conflict were negatively associated with the
perpetration of physical dating violence.
The current study
To my knowledge, no studies have examined the moderating effect of supportive
messages for fighting in the relationship between exposure to violence and imitated acts
of violence. Because existing studies have consistently found that norms supporting
violence play a critical role in individuals’ attitudes and violent behaviors, it is possible
that receiving pro-violence messages may moderate the relationship between exposure to
violence during childhood and later dating violence perpetration. In the current study, I
examine whether pro-violence messages moderate the relationship between early
exposure to violence and later attitudes toward violence and dating violence perpetration.
Based on the theoretical and empirical evidence reviewed above, I will address the
following hypotheses:
Hypothesis 1: Pro-violence messages, controlling other covariates, will be
positively related to attitudes approving violence as well as dating violence perpetration
among male and female college students.
Hypothesis 2: Pro-violence messages, controlling for other covariates, will
moderate the association between exposure to violence during childhood and attitudes
approving violence as well as dating violence perpetration among male and female
college students. That is, exposure to violence during childhood would be more strongly
associated with attitudes approving violence and increased dating violence perpetration
among those high in receiving pro-violence messages compared to those low in receiving
pro-violence messages.
35
3. Method
3.1 Data and sample
The data for this study comes from the International Dating Violence Study
(IDVS). The IDVS was conducted to investigate risk factors and protective factors related
to dating violence perpetration and victimization among college students (Straus,
2011). Study participants were selected through convenience sampling from 32 countries.
Data were collected from 17,404 student respondents, and most participants were
enrolled in university courses (criminology, sociology, psychology, or family studies).
All study procedures were approved by the University of New Hampshire Human
Subjects Board (Graham, Jensen, Givens, Bowen, & Rizo, 2019).
The sample for this study was composed of 4,533 U.S. college students from 17
universities in the IDVS. As this study focused on emerging adults and dating violence, I
restricted the sample to U.S. college students aged 18-23 years old who had engaged in a
heterosexual dating relationship lasting more than one month. Thus, I excluded cases of
respondents whose ages ranged from 33 to 45 (n=601), those with no dating relationships
or no relationships of at least 1 month’s duration (n=343), those having only same-sex
dating relationships (n=93), and those who were married (n=121). After removing cases
with missing values regarding family income (missing data < 2.0%), the final analytical
sample for our study included 3,302 participants.
3.2 Measures
Except for the CTS 2 scale for dating violence perpetration, all scales come from
the Personal and Relationships Profile (PRP) in IDVS (Straus, Hamby, Boney-McCoy, &
Sugarman, 1999). Commonly, the response categories of PRP scales range from 1
(strongly disagree) to 4 (strongly agree).
36
Independent variable. Exposure to violence. This variable was assessed using
the witnessed and victimized violence subscales within the Violence Socialization Scale
of the PRP (Straus, Hamby, Boney-McCoy, & Sugarman, 1999). A total of six items
consist of direct and indirect experiences of violence by family members and non-family
members, including physical punishment and witnessing violence. Higher scores indicate
higher exposure to violence during childhood. In this study, the internal consistency
reliability (Cronbach’s alpha) with the U.S. sample was 0.66.
Moderating variable. Pro-violence messages. This variable was measured using
the advised violence subscales within the Violence Socialization Scale of the PRP
(Straus, Hamby, Boney-McCoy, & Sugarman, 1999). The two items ask whether
respondents’ parents or community members gave the respondents pro-violence messages
during childhood (i.e., “When I was a kid, my father or mother told me to hit back if
someone hit me or insulted me” and “When I was a kid, people [adults or kids] who were
not part of my family told me to hit back if someone hit me or insulted me.) Higher
scores indicate a higher number of pro-violence messages from parents and other
community members. Reliability analysis found an alpha of .69 (Cronbach’s alpha).
Dependent variables. Dating Violence Perpetration. The physical assault scale
from the revised Conflict Tactics Scales (CTS2) was used to measure this variable. The
twelve items range from minor physical aggression (e.g., pushing or grabbing) to severe
physical aggression (e.g., punching or kicking). Each item has seven response categories:
1 = once, 2 = twice, 3 = 3-5 times, 4 = 6-10 times, 5 = 11-20 times, 6 = more than 20
times, and 0 = this has never happened in the current relationship. Higher scores indicate
37
higher frequency of physical violence toward dating partners. The internal consistency
reliability (Cronbach’s alpha) with the U.S. sample was 0.82.
Attitudes toward violence. This variable was measured using the Violence
Approval scale of the PRP (Straus et al., 1999). The scale includes 10 items which ask
respondents’ opinions concerning the acceptability of interpersonal violence in situations
such as family violence and sexual violence (e.g., whether it is acceptable for a man to
force his wife to have sex, or for parents to slap teens who talk back or repeatedly get into
trouble). Higher scores indicate higher levels of acceptance of violence. Internal
consistency with the U.S. sample was 0.70.
Control variables. Six control variables were incorporated into the analyses.
Three of the control variables were demographic measures: age (in years), family SES,
and type of household (two parents vs. single parents). Family SES was created by
summing three variables: father’s education, mother’s education, and family income. The
summed values were standardized using the sample mean and sample standard deviation.
Thus, the family SES indicates the level of standard deviation – if any – by which each
student is above or below the mean for U.S college students. Since previous research has
shown that relationships with parents and interactions with peers are associated with the
violent socialization process (Henry, Tolan, & Gorman-Smith, 2001), I controlled factors
relevant to parents and peers. Thus, the fourth control variable was positive parenting,
which came from the positive parenting scale of the PRP. Six items measured the degree
to which parents were loving and supportive, and properly supervised their children.
Internal consistency with the U.S. sample was 0.83. The fifth control variable was
delinquent peers. This variable came from a subscale, delinquent peers, of the social
38
integration scale in the PRP. Two items assess whether respondent have friends who
commit crimes. Internal consistency with the U.S. sample was 0.75. Lastly, previous
studies have pointed out that dating violence research should focus on current dating
relationship dynamics (O'Keefe, 1997). Thus, I controlled the effect of current
relationship dynamics with regard to dating violence perpetration by using the
Relationship Conflict variable from the PRP. A total of nine items asked respondents
about areas of disagreement with their partners regarding relationships with others,
habits, sexual relationships, and time spent together. The response categories for these
questions ranged from 1 (strongly disagree) to 4 (strongly agree), and higher scores
indicated higher levels of conflict. Internal consistency with the U.S. sample was 0.79.
3.3 Data analysis
The analysis occurred in two stages. First, chi-square tests and independent
sample t-tests were used to determine any significant differences among study variables
when comparing females and males. For the second stage of the analysis, the sample was
stratified by gender. Multivariate ordinary least square (OLS) regression was then
conducted to test the association between early exposure to violence and attitudes toward
violence and dating violence perpetration in each gender group. The dependent variable
of dating violence perpetration in particular was positively skewed; therefore, the scores
on the dating violence scales were logarithmically transformed (natural log) before
analyses to normalize distributions. In addition, the moderating effect of pro-violence
messages was examined by introducing and testing an interaction term between exposure
to violence and pro-violence messages in the OLS model. I also grand-mean centered the
continuous predictors to aid interpretation and avoid multicollinearity with the interaction
39
terms (Aiken, West, & Reno, 1991). Finally, since correlated errors were a concern due to
the nested nature of the data—the 3,302 college students were clustered in 17
universities––I estimated all of the models using robust standard errors clustered at the
university level. This was done using Stata’s vce(cluster) command, which generates
robust standard errors that are unbiased in the context of clustered data by generalizing
the Huber/White/sandwich estimate of variance to account for clustering (Lokshin &
Sajaia, 2004). All analyses were conducted using Stata 15.0.
4. Results
4.1. Descriptive statistics
Table 3.1 presents descriptive statistics for the two gender groups, as well as the
overall sample. College students included in the study sample were more often girls than
boys (68% vs. 32%). The mean age was 20.1 years (SE = 0.03), and most students lived
with two parents rather than a single parent (66% vs. 34%). Significant bivariate gender
differences in the study variables were noted. Female students were more likely than
male students to perpetrate dating violence (p <.01). However, male students were more
likely than female students to approve of interpersonal violence (p <.001) and were more
likely to have been exposed to violence during childhood (p <.001). Male students also
received more pro-violence messages than female students (p <.001).
4.2. Regression analyses for male students
Table 3.2 provides the results of OLS regression analyses in the male student
group. Model 1 and Model 2 show the findings of regression with regard to attitudes
toward violence. In Model 1, exposure to violence during childhood was associated with
40
male students’ attitudes approving of violence (b = 0.84, p <.001) controlling for other
covariates. In Model 2, pro-violence messages and the interaction term (exposure to
violence x pro-violence messages) were included. Exposure to violence was still
associated with attitudes toward violence (b = 0.25, p <.001), and pro-violence messages
were associated with male students’ approval of violence (b = 0.72, p <.001). However,
no interaction effect was found.
Model 3 & 4 show the results of regression analysis on dating violence
perpetration. In Model 3, early exposure to violence was not associated with dating
Table 3.1 Descriptive statistics for the variables and gender differences, International Dating Violence Survey (IDVS), % or
Mean (SE)
Gender (Female)
Age (years) 20.05 0.03 20.30 0.05 19.93 0.03 18-24.5 5.91***
Family SES 0.00 0.02 0.13 0.03 -0.06 0.02 -1.8-3.6 5.13***
CVs Two parents (vs. single parent) 66.29 68.79 65.15
Positive parenting 18.42 0.05 17.92 0.10 18.65 0.06 6-24 -6.43***
Delinquent peers 4.96 0.03 5.50 0.05 4.71 0.03 2-8 13.16***
Relationship conflict 17.95 0.08 18.73 0.13 17.59 0.09 9-34 6.96***
Dating violence perpetration 76.50 2.70 62.13 4.35 83.06 3.39 0-1200 -
3.60** (DVP)
DVs Log transformed DVP 1.61 0.04 1.40 0.07 1.70 0.05 0-7.09 -3.31*
Attitude towards violence 19.34 0.07 21.05 0.13 18.56 0.08 10-36
Moderating V
Pro-violence messages
4.25
0.03
4.89
0.05
3.96
0.03
2-8
IV
Exposure to violence
10.50
0.05
11.46
0.09
10.06
0.06
6-24
12.52***
Total
(N=3,302)
Male
(N=1,035)
Female
(N=2,267)
Range
t
-
val
ue
31.34
68.66
*p < 0.05, **p < 0.01, ***p < 0.001.
43
violence perpetration. In Model 4, early exposure to violence and pro-violence messages
were not associated with dating violence perpetration. Even though there was no overall
effect of either exposure to violence or pro-violence messages, crossover interaction was
noted (b = 0.03, p <.01). This means that the effect of exposure to violence on dating
violence perpetration may decline, depending on the value of pro-violence messages. As
shown in Figure 3.1, the effect of early exposure to violence on later dating violence
perpetration tends to be stronger on participants who received higher levels of
proviolence messages as children. In addition, the slope appears steeper in the group of
students who received higher levels of pro-violence messages than in the group who
received lower levels of pro-violence messages.
Figure 3.1 Interaction effect of exposure to violence and pro-violence messages in the
male group (N=1,035).
44
This means that children who faced higher frequencies of exposure to violence
increasingly strengthened the likelihood that they perpetrated dating violence in later
years among students who received pro-violence messages.
Table 3.3 indicates the results of OLS regression analyses in the female student
group. In Model 1, exposure to violence during childhood was associated with female
students’ attitudes approving of violence (b = 0.42, p <.001) controlling for other
covariates. In Model 2, exposure to violence and pro-violence messages were associated
with attitudes accepting violence (respectively, b = 0.22, p <.001; b=0.85, p <.001). An
interaction effect was found showing that the more often a student was exposed to
violence during childhood, the greater is the predicted probability that they will approve
of violence if they also received pro-violence messages in childhood (Figure 3.2).
Figure 3.2 Interaction effect of exposure to violence and pro-violence messages in the
female group (N=2,267).
10
15
20
25
30
35
6
10
14
18
22
24
Early exposure to violence
-1
SD Proviolence messages
Mean Proviolence messages
SD Proviolence messages
+1
45
However, students who experienced early exposure to violence, but who did not receive
pro-violence messages, were less likely than the former group to approve of violence (b
=
-0.04, p <.01).
Model 3 shows a pattern found among female students but not among male
students: for female students, early exposure to violence is associated with dating
violence perpetration (b=0.09, p <0.01). In Model 4, after adding pro-violence messages,
early exposure to violence is still associated with increased dating violence perpetration
(b = 0.06, p <.05); and pro-violence messages are also associated with dating violence
perpetration (b = 0.14, p <.05). No interaction effect was found.
4.3. Additional analysis
For the purpose of cross validation, I randomly divided the original data set
(n=3,302) into two subsets (holdout sample and exploratory sample) and compared the
results of analyses, including OLS regression. The randomization process was conducted
using the commands “Select case – Random sample of cases” by SPSS 22.0. Through
this cross validation procedure, I found that the results of the holdout sample and
exploratory sample were similar, meaning our data set was balanced and the results could
be generalized.
5. Discussion
The present study has examined the moderating role of pro-violence messages in
the relationship between early exposure to violence and later attitudes toward violence
and dating violence perpetration. Overall, our findings showed that pro-violence
46
messages are associated with attitudes supporting violence among male and female
students, but that such messages are associated with increased dating violence
Table 3.2. OLS regression on attitudes toward violence and dating violence perpetration among male students (N = 1.035)
Attitudes toward violence
Dating violence perpetration
Model 1
Model 2
Model 3
Model 4
b
RSE
Beta
b
RSE
Beta
b
RSE
Beta
b
RSE
Beta
Exposure to violence (ETV)
0.84***
0.08
0.84
0.25***
0.04
0.18
0.04
0.03
0.05
0.02
0.03
0.03
Pro-violence messages (PM)
0.72***
0.08
0.28
-0.05
0.05
-0.03
Age (years)
-0.10
0.08
-0.10
-0.15
0.07
-0.10*
0.04
-0.07
-0.10*
0.04
-0.07
Family SES
-0.12
0.11
-0.12
-0.06
0.11
-0.01
-0.15
0.09
-0.06
-0.16
0.08
-0.06
Two-parent household
-0.12
0.30
-0.12
-0.03
0.30
0.00
-0.20
0.14
-0.04
-0.19
0.15
-0.04
Positive parenting
-0.17***
0.03
-0.17
-0.11**
0.04
-0.09
0.01
0.02
0.01
0.00
0.02
0.00
Delinquent peers
0.34**
0.08
0.34
0.27**
0.08
0.10
0.07
0.05
0.05
0.07
0.05
0.04
Relationship conflict
0.14***
0.01
0.24
0.14***
0.01
0.25
ETV x PM
-0.03
0.02
-0.04
0.03**
0.01
0.07
Δ R2
0.17
0.02
0.08
0.01
*p < 0.05, **p < 0.01, ***p < 0.001. Note. Relationship conflict is only a control variable in the regression analysis on dating
violence perpetration.
Table 3.3 OLS regression on attitudes toward violence and dating violence perpetration among female students (N = 2,267)
Model 1 Model 2 Model 3 Model 4
b
RSE
Beta
b
RSE
Beta
b
RSE
Beta
b
RSE
Beta
Exposure to violence (ETV)
0.42***
0.03
0.32
0.22***
0.03
0.17
0.09**
0.02
0.11
0.06*
0.02
0.08
Pro-violence messages (PM)
0.85***
0.06
0.36
0.14*
0.05
0.09
Age (years)
-0.07
0.07
-0.03
-0.07
0.06
-0.03
-0.03
0.04
-0.02
Family SES
-0.40***
0.08
-0.10
-0.14*
0.06
-0.05
-0.15*
0.06
-0.06
-0.12*
0.05
-0.05
Two-parent household
0.13
0.19
0.02
-0.14
0.08
-0.03
-0.12
0.09
-0.02
Positive parenting
-0.11***
0.02
-0.09
-0.10***
0.02
-0.08
0.00
0.01
0.01
0.01
0.01
0.01
Delinquent peers
0.22
0.05
0.09
0.16**
0.04
0.07
0.07*
0.03
0.05
0.06
0.03
0.04
Relationship conflict
0.14***
0.01
0.25
0.14***
0.01
0.25
ETV x PM
-0.04*
0.02
-0.05
0.01
0.01
0.01
Δ R2 0.18 0.09 0.11 0.01
*p < 0.05, **p < 0.01, ***p < 0.001. Note. Relationship conflict is only a control variable in the regression analysis on
dating violence perpetration.
Attitudes
toward violence
Dating violence perpetration
50
perpetration only among female students. In addition, pro-violence messages moderate
the relationship between exposure to violence and attitudes accepting violence among
only female students; however, the messages also moderated the effect of exposure to
violence on dating violence perpetration among male students.
These findings are partially consistent with the view that parental and peer
support for aggression plays a significant role in cognitive belief and aggressive
behaviors (Copeland-Linder et al., 2007; Johnson et al., 2011; Solomon et al., 2008).
However, the findings clearly show that pro-violence messages have a different effect on
college students’ cognition (attitude toward violence) and their actual behaviors (dating
violence perpetration). Furthermore, the effect of pro-violence messages varies according
to gender. Thus, our findings add new knowledge to the study of violent socialization:
messages from parents, peers, and other community members which support violent
coping will have different effects on individuals’ cognition and actual violent behavior,
depending on the individuals’ gender.
The first hypothesis I addressed was that receiving pro-violence messages from
parents and other community members would be positively related to attitudes supporting
violence and increased dating violence perpetration. The findings support Hypothesis 1
and showed that receiving pro-violence messages is positively associated with attitudes
supporting violence in both male and female students controlling for
demographic/background characteristics. This result supports the concept that parents’
values, attitudes, and behaviors strongly influence children’s aggressive attitudes and
behaviors (Duman & Margolin, 2007; Snyder & Patterson, 1995). For instance, parents or
family members might encourage their children to respond aggressively in certain
51
situations, including provocation. Some parents might also think it is good for their
children, particularly boys, to learn how to fight. Such verbal permission and
encouragement regarding violence might encourage children to accept or condone
violence against others, which in turn may lead to perpetuation of interpersonal violence
in adulthood.
Meanwhile receiving pro-violence messages is positively associated with dating
violence perpetration in the female student sample, but not in the male sample. This
finding seems to contradict some previous studies that suggest parental tolerance of
violence and peer support for fighting are more closely related to aggression in male
adolescents than in females – see Walters et al. (2017), Garthe et al (2018), and
Copeland-Linder et al (2007). One possible explanation for these results is that there
might be other more proximate risk factors (e.g., current relationship factors) which lead
to dating violence perpetration among boys. O'Keefe (1999) argued that contextual
factors (previous background) are not the only factors which should be considered in
studies of dating violence, but that current situational factors such as alcohol use,
relationship distress, and satisfaction should be considered as well.
Overall, among both male and female college students, receiving pro-violence
messages appears more strongly related to college students’ pro-violence cognition
supporting violence rather than actual aggressive behavior toward against dating partners.
Even though receiving verbal support of fighting during childhood significantly predicts
acceptance of violence in later life, the finding showed that such verbal reinforcement
does not have a strong effect on actual aggressive behavior.
52
The second hypothesis was that receiving pro-violence messages would moderate
the pathway from exposure to violence to attitudes toward violence and dating violence
perpetration. To begin, I examined the association between exposure to violence and
individual’s’ attitudes toward violence, and I found a moderating effect of pro-violence
messages – but only in the female students’ model. That is, female students who
experienced early exposure to violence were more likely to have attitudes approving
violence when they received high levels of pro-violence messages. However, in the male
students’ model, even though pro-violence messages predicted male students’ attitudes
toward violence, the association did not differ according to the presence or absence of
pro-violence messages.
I next explored the moderating effect of pro-violence messages in the association
between exposure to violence and dating violence perpetration. Although the direct
impact of exposure to violence and pro-violence messages on dating violence
perpetration were not significant, I found a moderating effect in the male students’ model.
That is, male students who were exposed to violence as children tended to show
increased dating violence perpetration when they also received higher levels of
proviolence messages. This means that as children face higher frequencies of exposures
to violence, the effects of exposure to violence on dating violence become stronger for
those who received a high level of proviolence messages than those who not. As
discussed above, a moderating effect of pro-violence messages was partially found.
The current study extends theories of violence – in particular, social learning
theory – in two ways. First, it shows that parents, peers, and community members’ norms
supporting or tolerating violent coping may be a significant risk factor in learning and
53
reinforcing beliefs that justify the use of violence and actual violent behaviors. However,
existing studies applying social learning theory have focused largely on the effect of
exposure to violence on children’s aggressive behaviors. We need to consider the
mechanisms or conditioning factors that adjust or reinforce the relationship between early
exposure to violence and later aggressive behaviors. Second, theories and empirical
studies of violence have shown that exposure to violence and verbal reinforcement of
violence might be common risk factors in developing beliefs that justify violence and in
actual violence perpetration. However, our study points out that exposure to violence
predict or moderate cognition alone or behaviors alone. This study also showed that
proviolence messages might be more closely related to college students’ cognition than to
actual dating violence perpetration. Thus, this finding suggested that we need to explore
what risk factors have stronger effects on individual’s cognition and actual behaviors in
violence learning process.
The findings from the current study have important clinical implications. First,
this study provides initial evidence that families and community members can help
reduce the likelihood of dating violence perpetration in later life through nonviolent
advice and communication with their children. Thus, interventions focused on preventing
pro-violence norms by training in healthy parenting/coaching skills to help children avoid
violence appear to be worthwhile targets for community programs aimed at parents,
peers, and other community members. In addition, practitioners developing dating
violence programs should consider college students’ personal backgrounds, including
their genders and violence histories, and should seek understanding of how student
54
attitudes toward violence (cognition) and dating violence perpetration (behaviors) are
developed and enhanced.
In interpreting the results of the present study, it is important to note its
limitations. First, the cross-sectional data precluded causal interpretations, despite the
causal assumptions suggested by social learning theory. Future research using
longitudinal designs are necessary to rule out alternative interpretations of the current
findings. A second limitation of the present study lies in its measurements. In particular,
exposure to violence and attitudes toward violence cover different types of general
violence patterns rather than specific types of violence exposure or attitudes toward
particular type of violence (i.e. male-to-female violence). This measurement might give
different results in comparison to previous studies which have focused on particular types
of violence. However, despite these limitations, the present study extends our
understanding of the role of pro-violence messages in moderating the relationship
between exposures to violence and dating violence perpetration. In doing so, these
findings provide an empirical basis for further research aimed at alleviating dating
violence perpetration among college students through changing norms support violence.
CHAPTER 4
FACTORS MEDIATING THE LINK BETWEEN EARLY EXPOSURE
TO VIOLENCE AND DATING VIOLENCE PERPETRATION
3
3
Kim, J. To be submitted to Journal of Interpersonal Violence
55
Abstract
Research indicates that children exposed to violence are at a greater risk for engaging in
violent behavior as adults. However, the pathway between exposure to violence during
childhood and perpetration of dating violence in young adulthood remains unclear.
Utilizing a sample of U.S. college students, the current study examines whether attitudes
toward violence and substance use mediate the relationship between childhood exposure
to violence and dating violence perpetration. The results of structural equation modeling
confirm the mediation models and suggest that the direct and mediating pathways are
partially moderated by gender. Implications for future research and policy are discussed.
Keywords: exposure to violence, dating violence perpetration, attitude towards violence,
substance use
56
1. Introduction
Dating violence is a serious and prevalent problem among college-aged couples.
A National Intimate Partner and Sexual Violence Survey (2010) found that 47% of
females and 39% of males first experience physical violence, sexual violence, and/or
stalking by an intimate partner between the ages of 18 and 24 (Black et al., 2011).
Research has also shown that 34% of female students report having experienced
psychological violence, such as being insulted or humiliated by a dating partner, during
the past 12 months (Barrick, Krebs, & Lindquist, 2013).
Although estimates of dating violence vary depending on the definition of dating
violence and the measurements used to study it, research consistently shows that between
20% and 50% of college students report being the victim of at least one type of dating
violence (Barrick et al., 2013; Kaukinen, Gover, & Hartman, 2012; Shorey, Cornelius,
Bell, & behavior, 2008). Dating violence entails a risk of various adverse health
outcomes, including physical injuries and psychological trauma (Buelna, Ulloa, &
Ulibarri, 2009; Paat & Markham, 2016). Thus, it is important to examine studies that help
guide interventions and prevent future occurrences of dating violence within this age
group.
A number of studies have delved into the causal factors of intimate partner
violence perpetration, and they reveal that early exposure to violence is one of the most
consistent predictors of later violence (Eriksson & Mazerolle, 2015; O'Keefe, 1998).
However, little is known about how exposure to violence during childhood leads to the
perpetration of dating violence, and still questions remain about the mechanisms that may
underlie this relationship. Furthermore, compared to research of adolescents and adults,
57
the link between exposure to violence and violence perpetration among college students
has not been extensively explored, even though dating violence is more prevalent in
young adulthood than in other life stages (Buelna et al., 2009).
The current study aims to examine what factors may mediate the relationship
between exposure to violence during childhood and dating violence perpetration among
college students. Several theoretical frameworks have been proposed to explain these
underlying mechanisms. Social learning theory (Akers & Jennings, 2009; Bandura, 1978)
emphasizes that indirect and direct exposure to violence leads children to view violence
as normative or acceptable in relationships – which, in turn, contributes to the likelihood
of violent behavior. According to this perspective, the impact of early exposure to
violence and dating violence in young adulthood can be assessed by measuring attitudes
toward violence. The general strain theory (Agnew, 1992) and alcohol myopic model
(Steele & Josephs, 1990) guide our understanding of the ways in which people who are
exposed to violence may manage their stress through substance use – which may, in turn,
lead them to engage in delinquent or violent acts.
Exposure to violence and dating violence: attitudes toward violence as a mediator
Social learning theories posit that behaviors are learned through observation and
imitation of others (Bandura, 1978). During the process, people also learn attitudes or
definitions for certain behaviors as good or bad (Akers & Jennings, 2009). If a person
possesses a positive definition of a certain behavior, the person will be more likely to
engage in that behavior. As social learning theory suggests, children who are exposed to
violence might acquire a belief system justifying the use of interpersonal violence
58
(Eriksson & Mazerolle, 2015; Su, Mrug, & Windle, 2010). Such exposure to violence
increases the risk of perpetrating violence directly through modeling and imitation, or
indirectly through attitudes toward violence (Su et al., 2010). Thus, an individual’s
attitudes, approval, and perception of violence may mediate the relationship between the
social learning of violence in childhood and the perpetration of violence later on.
Although social learning theories emphasize the mediating role of individual’s cognition
or beliefs, little is known about the mediating role of attitudes toward violence in the
pathway from childhood exposure to violence to dating violence perpetration. Some
existing studies support the theory that approval of violence which develops from early
exposure to violence is associated with intimate partner violence perpetration (Clarey,
Hokoda, & Ulloa, 2010; Foshee, Bauman, & Linder, 1999; O'Keefe, 1998; Reyes et al.,
2015). For example, one longitudinal study of adolescents found that the pathway from
witnessing and directly experiencing family violence to physical dating aggression
perpetration is mediated by the belief that physical dating violence is acceptable (Reyes
et al., 2015). However, some studies contradict these findings, demonstrating that
attitudes toward violence does not mediate the relationship between exposure to violence
and dating violence perpetration (Debowska, Boduszek, Jones, Willmott, & Sherretts,
2017; Eriksson & Mazerolle, 2015).
Meanwhile, other studies have shown that particular types of exposure to violence
are more likely to be associated with acceptance of violence and violence perpetration
(Debowska et al., 2017; Foshee et al., 1999). For example, Foshee et al (1999) found that
only victimization, not witnessing violence, was strongly associated with acceptance of
59
dating violence perpetration. In another study, Debowska et al (2107) found that violent
victimization of children by family members created significant positive associations
with male physical domestic violence and social norms regarding physical violence
against females. These findings indicate different effects of exposure to violence and the
need to explore type-specific etiology and effects regarding exposure to violence (Reyes,
Foshee, Niolon, Reidy, & Hall, 2016).
The pathways which lead from early exposure to violence, through attitudes
toward violence, and on to actual violence perpetration seem to differ according to
gender. Studies have found that males who witness or are victimized by family and
nonfamily violence are more likely than females to approve of violent behaviors
(Bacchini, Affuso, & Aquilar, 2015; Debowska et al., 2017; Mattingly & Straus, 2008;
Ramsay,
Steeves, Feng, & Farag, 2017; Spencer, Morgan, Bridges, Washburn-Busk, & Stith,
2017). In addition, research has suggested that boys who were exposed to violence
demonstrate more externalizing behaviors, including violent behaviors, while girls tend to
display more internalizing behaviors, such as emotional distress and depression (Black et
al., 2015; Evans, Davies, & DiLillo, 2008). These findings suggest a need for further
research of gender differences among these pathways.
Exposure to violence and dating violence: substance use as a mediator
Substance use may be an important mediator in the pathway from exposure to
violence during childhood to violence perpetration in adulthood. General strain theory
suggests that exposure to strain increases the potential for young adults to engage in
delinquent behaviors such as violence (Agnew, 1992; Morash & Moon, 2007). Agnew
60
(1992) explained that, in response to the strain of a stressful event, adolescents may try to
manage their negative emotions through alcohol and drug abuse – which may, in turn,
lead them to engage in delinquent or violent acts. Studies have also reported that
individuals abuse alcohol and other drugs to cope with or block negative emotions or
memories of trauma (Crimmins, Cleary, Brownstein, Spunt, & Warley, 2000; Harrison,
Fulkerson, & Beebe, 1997).
Meanwhile, intimate partner violence research has identified substance use,
particularly alcohol abuse, as a trigger for aggressive behaviors (Reyes, Foshee, Bauer, &
Ennett, 2011; Roudsari, Leahy, & Walters, 2009; Temple, Shorey, Tortolero, Wolfe, &
Stuart, 2013). Violence perpetration can occur during various phases of alcohol use,
including acute intoxication and alcohol-seeking behavior (Shorey, Stuart, & Cornelius,
2011). The alcohol myopic model (Steele & Josephs, 1990) suggests that acute
intoxication, most notably with alcohol, reduces an individual’s ability to detect social
cues and leads to a risk of aggressive behaviors (Shorey et al., 2011). Specifically, alcohol
creates a “myopic” perspective which distorts perceptual cues. This distorted cognitive
function may then facilitate aggression by focusing an individual’s attention on
provocation rather than on inhibitory cues in situations of potential conflict (Giancola,
2002; Reyes, Foshee, Bauer, & Ennett, 2014).
This association is supported by other research showing that substance abuse
increases intimate partner violence (Follingstad, Bradley, Laughlin, & Burke, 1999;
Luthra & Gidycz, 2006; Shorey et al., 2018; Shorey et al., 2011; Temple et al., 2013). For
example, in one longitudinal study, Temple, Shorey, Fite, Stuart, and Le (2013) found that
alcohol use was significantly associated with subsequent physical violence perpetration
61
against a dating partner. Shorely et al (2018) also found that marijuana use positively
associated with all forms of IPV (psychological, physical, and sexual violence) among
men arrested for domestic violence.
Empirical evidence regarding the effect of substance abuse on aggression indicates that
gender may be a significant moderator in this pathway. Although there are mixed findings
(Shorey et al., 2011), many studies have found that male drinking patterns are more likely
than female patterns to combine binge drinking with violent behavior (Boles & Miotto,
2003; Hove, Parkhill, Neighbors, McConchie, & Fossos, 2010; Luthra & Gidycz, 2006).
Consistent with alcohol research, research on abuse of other drugs – including both
marijuana and hard drugs – suggests that the relationship between drug abuse and dating
violence is moderated by gender (Brady & Randall, 1999).
Current Study
As discussed above, although dating violence is a significant issue among college
students, the factors that mediate the link between exposure to violence and violence
perpetration have yet to be thoroughly examined. Furthermore, some previous studies
have focused on exposure to physical violence and physical dating violence; but other
types of violence are serious issues as well, and we need to explore the underlying
mechanisms of these types of violence. Based on these research gaps, the purpose of this
study is (1) to examine models in which attitudes toward violence and substance use
mediate the relationship from early exposure to violence (physical and sexual violence) to
dating violence perpetration (physical violence and psychological violence), and (2) to
examine whether these models vary by gender (See Figure 4.1). Based on the theoretical
framework, the following research questions are explored:
62
1) Are there any significant direct and indirect pathways (i.e., mediation)
from early exposure to violence to dating violence perpetration through attitudes
toward violence and substance use?
2) Does a student’s gender moderate the direct or indirect pathways from
early exposure to violence to attitudes toward violence?
Figure 4.1. Hypothesized mediation model linking violence exposure with college
students’ dating violence perpetration. Note. Direct links from exposure to violence to
dating violence perpetration are not depicted for simplicity.
2. Method
2.1 Data
The data for this study was obtained from the International Dating Violence Study
(IDVS). The IDVS was conducted between 2001 and 2006 and focused on college and
university students from 32 countries (Straus, 2011). Study participants were selected
through convenience sampling, and most participants were enrolled in university courses
(criminology, sociology, psychology, or family studies) taught by a designated research
63
consortium member. Data were collected from 17,404 student respondents, 14,525 of
whom reported having been in an intimate relationship of at least 1 month.
2.2 Sample
This study utilized data from 4,533 U.S. college students in the IDVS. The sample
included students from 17 universities in the U.S. As this study focused on emerging
adults and dating violence, it included only U.S. college students aged 18-23 years old
who had engaged in a heterosexual dating relationship lasting more than one month.
Thus, I excluded cases of respondents whose ages ranged from 33 to 45 (n=601), those
with no dating relationships or no relationships of at least 1 month’s duration (n=343),
those having only same-sex dating relationships (n=93), and those who were married
(n=121). Cases with missing values regarding family income (missing data < 2.0%) were
removed through listwise deletion. The final analytical sample for our study included
3,302 participants with a mean age of 20.01 years (SE = 0.03). The students included in
the study sample were more often girls than boys (69% vs. 31%). Slightly more male
students than female students lived in families having above-average socioeconomic
status (SES) (Male=0.13, Female = -0.06).
2.3 Measures
Physical Violence Perpetration. This variable was measured using the physical
assault scale of the revised Conflict Tactics Scales (CTS2). Twelve items were assessed
by individuals’ self-report of physical aggression toward intimate partners, including
minor physical aggression (e.g., pushing or grabbing) and severe physical aggression
(e.g., punching or kicking). For each item, participants chose their answers from seven
frequency categories: 1 = once, 2 = twice, 3 = 3 to 5 times, 4 = 6 to 10 times, 5 = 11 to 20
64
times, 6 = more than 20 times, and 0 = this has never happened in the current
relationship. The CTS2, for which previous studies have established construct validity
and moderate to high reliability, measures violent behaviors (Straus, 2008). In this study,
the internal consistency reliability (Cronbach’s alpha) with the U.S. sample was 0.82.
Psychological Violence Perpetration. Eight items from the psychological aggression scale
of the CTS2 were used to assess self-reported psychological violence perpetration against
a dating partner. This scale includes minor aggression (e.g., insulting or swearing) and
severe psychological aggression (e.g., threatening to hit or throw something). For each
item, participants chose their answers from seven frequency categories: 1 = once, 2 =
twice, 3 = 3 to 5 times, 4 = 6 to 10 times, 5 = 11 to 20 times, 6 = more than 20 times, and
0 = this has never happened in the current relationship. The internal consistency
reliability with the U.S. sample was 0.76.
Exposure to physical violence. This variable was measured using the witnessed
and victimized violence subscales within the Violence Socialization Scale of the Personal
and Relationships Profile (PRP) (Straus, Hamby, Boney-McCoy, & Sugarman, 1999).
This scale includes six items, which question the kinds of violence exposure respondents
have experienced during childhood. These items consist of direct and indirect experiences
of violence from family members and non-family members, such as physical punishment
and witnessing violence. The response categories to these questions range from 1
(strongly disagree) to 4 (strongly agree) with higher scores indicative of higher exposure
to physical violence. In this study, the internal consistency reliability with the U.S.
sample was 0.66.
65
Exposure to sexual violence. This variable was measured using the Sexual Abuse
History Scale of the PRP (Straus et al., 1999). The scale includes eight items, concerning
individuals’ experiences of sexual abuse by family and nonfamily members during
childhood. For example, respondents were asked whether family members or nonfamily
members looked at or touched the respondents’ private parts. The response categories to
these questions ranged from 1 (strongly disagree) to 4 (strongly agree) with higher scores
indicative of higher exposure to sexual violence. In this study, the internal consistency
reliability with the U.S. sample was 0.79.
Attitude toward violence. Attitude toward violence was measured using the
Violence Approval portion of the PRP (Straus et al., 1999). This scale assesses the extent
to which respondents endorse the use of physical violence in various social contexts. The
scale includes 10 items, which focus on approval of interpersonal violence such as family
violence and sexual violence. For example, respondents were asked whether there were
circumstances in which it would be acceptable for a man to force his wife to have sex, or
for parents to slap a teen who talked back or repeatedly got into trouble. The response
categories for all scales ranged from 1 (strongly disagree) to 4 (strongly agree) with
higher scores indicative of higher acceptance of violence. Internal consistency with the
U.S. sample was 0.70.
Substance use. To obtain a measure of substance use, the Substance Abuse scale
of the PRP was used (Straus et al., 1999). Eight items assessed students’ excessive use of
alcohol or other mind-altering drugs. For example, respondents were asked whether there
were times when they couldn’t remember what happened the night before because of
drinking. The response categories to these questions ranged from 1 (strongly disagree) to
66
4 (strongly agree), and higher scores indicated higher levels of substance use. Internal
consistency with the U.S. sample was 0.81.
Control variables. Demographic controls of gender, age in years, and family SES
were included. Family SES was created by summing three variables: father’s education,
mother’s education, and family income. The summed values were standardized by the
sample mean and sample standard deviation. Thus, the family SES indicates the number
of the standard deviation – if any – by which each student was above or below the mean
for U.S college students. Additionally, I controlled the effect of current relationship
dynamics with regard to dating violence perpetration by using the Relationship Conflict
variable from the PRP. A total of nine items asked respondents about areas of
disagreement with their partners about relationships with others, habits, sexual
relationships, and time spent together. The response categories to these questions ranged
from 1 (strongly disagree) to 4 (strongly agree), and higher scores indicated higher levels
of conflict. Internal consistency with the U.S. sample was 0.79.
3. Data Analysis
Data were analyzed with Mplus version 6.6 (Muthen &, 1998-2010) using
structural equation modeling (SEM). SEM allows simultaneous assessments of the
magnitude and significance of relationships among the exogenous (i.e., predictors) and
endogenous (i.e., mediators and outcome) variables included in the model. In this case,
the relationship between the predictor (i.e., exposure to violence), mediators (i.e., attitude
towards violence and substance use), and dating violence perpetration among college
students were assessed.
67
Because of the endogenous variables, dating violence perpetration was highly
skewed, and the data were considered to be severely non-normal. For this reason, all
analyses were performed using MLR (maximum likelihood estimation with robust
standard errors) that are robust with regard to non-normality (Asparouhov & Muthén,
2005; Maydeu-Olivares, 2017). Another factor which I took into consideration was that
the college students were clustered within schools, a fact which may violate the
assumption of independence. I analyzed the data accounting for such clustering with the
use of the Mplus complex sampling option (Muthén & Muthén, 2019).
All causal paths in the model were adjusted for the covariates (age, gender, family SES,
and relationship conflict). The statistical significance of mediation effects was evaluated
using the Model Indirect statement in Mplus. This analysis utilizes Chi-square values
(non-significant), the root mean square error of approximation (RMSEA < 0.05), the
comparative fit index (CFI > 0.90), the Tucker–Lewis Index (TLI > 0.90), and the
standardized root-mean-square residual (SRMR < 0.05), as primary fit indices (Kline,
2010) All path coefficients are presented standardized.
4. Results
4.1. Descriptive statistics
Table 4. 1 shows the mean and t-test results of gender difference in the variables
of our study. Female students scored significantly higher than male students on both
physical and psychological violence perpetration toward dating partners (p < .05).
However, male students were more likely than female students to have experienced
physical violence during childhood (p < .05). Male students also tended to more
approving of interpersonal violence and substance use than female students (p < .05).
68
These gender difference results also indicate that the proposed structural equational
model should be tested separately for males and females, rather than including gender as
a covariate within a single model.
Table 4.1. Descriptive statistics for the variables and gender differences, Mean (SE)
Total (N=3,302)
Male
(N=1,035)
Female
(N=2,267)
Range
t-value
DV
Physical VP
1.78
0.07
1.45
0.12
1.93
0.09
0-72
-2.99*
Psychological VP
6.48
0.17
5.54
0.19
6.91
0.14
0-48
-5.47*
IVs
Exposure to PV
10.49
0.07
11.46
0.09
10.06
0.06
6-24
12.51*
Exposure to SV
10.46
0.06
10.60
0.11
10.40
0.07
8-32
1.52
Mediators
Attitudes toward Violence
19.34
0.07
21.05
0.13
18.56
0.08
10-40
16.65*
Substance use
13.44
0.08
14.84
0.15
12.81
0.09
8-32
12.44*
Note. VP = violence perpetration; PV = physical violence; SV = sexual violence. *p<0.5 or lower
71
4.2. Correlations among observed variables in the models
To examine the relationships and influences, we computed the correlations
between the variables in the models (see Table 4.2). The results demonstrated that
variables related to dating violence perpetration were positively associated with attitudes
toward violence and substance use overall (p < .05), but physical violence perpetration
had no particular association with substance use. Exposure to physical and sexual
violence were positively associated with attitudes toward violence and substance use (p <
.05).
4.3. Testing the Mediation Model
Structural equation modeling was employed to test whether attitudes toward
violence and substance use mediated the association between violence exposure and
dating violence perpetration. First, we tested a full model in which early exposure to
physical and sexual violence was linked both directly and indirectly with dating violence
perpetration through both mediators (see Figure 4.1). In this and all following models,
early exposure to physical violence and sexual violence were allowed to covary with each
other and all control variables could also covary. In addition, covariations among the two
mediators were included.
This full model was just-identified (df = 0), so model fit indices indicated perfect
fit. In this mediation model, of the four direct effects, only the direct path, exposure to
physical violence with physical violence perpetration, remained significant. In the model,
all other paths were significant except for the path linking exposure to physical violence
with substance use and the path linking substance use with physical violence
perpetration.
Table 4.2. Bivariate correlations among the variables (N=3,302).
Variables
1
2
3
4
5
6
7
8
9
10
1. Gender (Female)
1.00
2. Age (Years)
-0.10*
1.00
3. Family SES
-0.09*
-0.04*
1.00
4. Conflict
-0.12*
0.04*
-0.02
1
5. Physical DV
0.05*
-0.02
-0.11*
0.25*
1
6. Psychological DV
0.09*
-0.01
-0.04*
0.39*
0.58*
1
7. Exposure to PV
-0.21*
0.04*
-0.20*
0.29*
0.20*
0.17*
1
8. Exposure to SV
-0.03
0.03
-0.10*
0.21*
0.14*
0.12*
0.38*
1
9. Attitudes toward
-0.28*
-0.01
-0.12*
0.34*
0.20*
0.21*
0.40*
0.26*
1
10. Substance use
-0.21*
0.01
0.17*
0.22*
0.03
0.12*
0.14*
0.22*
0.21*
1
Note: Gender was coded 0 for male and 1 for female; Family SES (socio-economic status) was standardized; VP =
violence perpetration; PV = physical violence; SV = sexual violence *p<0.5 or lower
74
A reduced model was formulated by eliminating all nonsignificant paths in the
full model. Model fit indices suggested a good fit of the reduced model, χ2(df=4,
n=3,302) = 11.032, p = 0.0262; comparative fit index = 0.998, root mean square error of
approximation = 0.023, standardized root mean square residual = 0.008. Because the
reduced model fit the data well and was more parsimonious than the full model, all
following analyses were based on the reduced model. The reduced model with all
standardized path coefficients is depicted in Figure 4.2.
Figure 4.2. Reduced mediation model linking early exposure to violence with college
students’ dating violence perpetration. Note: All paths were adjusted for students’ age,
gender, family SES, and dating relationship conflict; VP = violence perpetration. *p<.05,
**p<.01, ***p < .001.
The first research question addressed the direct and indirect pathways by which
exposure to violence influenced dating violence. As expected, exposure to both physical
violence and sexual violence was significantly associated with increased approval of
violence (respectively, β = 0.35, p < .001; β = 0.08, p < .001). However, only exposure to
sexual violence was associated with substance use (β = 0.21, p < .01). Attitude toward
75
violence as a mediator was in turn related to physical and psychological violence
perpetration (respectively, β = 0.11, p < .001; β = 0.10, p < .001). Substance use,
however, was only associated with psychological violence perpetration (β = 0.05, p <
.05), not physical violence perpetration.
Analyses of indirect effects indicated statistically significant mediation between
exposure to physical violence and physical violence perpetration (β = 0.03, p < .001) and
between exposure to physical violence and psychological violence perpetration (β = 0.03,
p < .001) – in both cases via attitude toward violence. The association between exposure
to sexual violence and two types of dating violence perpetration was also mediated by
attitude toward violence (both, β = 0.01, p < .001). In fact, attitude toward violence was a
significant mediator in all mediation pathways, but substance use only mediated in the
relationship between exposure to sexual violence and psychological violence perpetration
(β = 0.01, p < .05).
4.4 Testing the Moderating Effects of Gender
To examine whether gender moderates the mediation pathways of exposure to
violence on dating violence perpetration, the total sample was divided into two
subgroups: a male student group (n = 1,035) and a female student group (n = 2,267). The
moderation hypothesis was evaluated with multigroup modeling by testing equivalence
of the model across the male and female groups. Specifically, we compared the fit of a
constrained model (all paths fixed as equal for both groups) with the fit of an
unconstrained model (all paths allowed to vary across the groups). As suggested by a
significant chi-square difference, Δχ2= 65.17, df= 19, p < .001, the model differs
according to student gender.
76
Standardized path coefficients for both groups are depicted in Figure 4.3.
Followup tests of invariance for individual path estimates across groups were conducted
by freeing one path at a time and comparing each new model with the fully constrained
model. One path was found to be unequal across genders, and it is indicated by bold lines
in Figure 4.3. Substance use is shown to predict psychological violence perpetration only
in the female student group, not in the male student group, Δχ2(1) = 17.96, p < .001.
Figure 4.3 Multigroup analysis based on gender. Note: Bold lines represent paths which
differ significantly across gender; VP = violence perpetration. *p<.05, **p<.01, ***p <
.001.
77
5. Discussion
Although a wide range of studies has posited an association between exposure to
violence during childhood and violent behaviors in later life, little research has
investigated the mechanisms that might underlie or explain this relationship.
The current study has examined the association between early exposure to
violence and dating violence perpetrated by college students. It tested whether this
relationship was mediated by attitudes toward violence and/or by substance abuse, and
whether the effects were moderated by the students’ gender. Overall, the study findings
show that attitude toward violence is a significant mediator in the relationship between
childhood exposure to violence and dating violence perpetration. However, substance
abuse is a significant mediator only in the relationship between exposure to sexual
violence and psychological violence perpetration. Significant differences emerge in the
pathway from substance use and psychological dating violence according to students’
gender.
Different patterns of mediation emerged for each predictor. First, the finding
highlights the significance of attitude toward violence as a common mediator in the
effects of exposure to violence – whether physical or sexual – on physical and
psychological violence perpetration. This finding reinforces earlier work demonstrating
that exposure to violence in childhood is strongly linked to partner violence perpetration
when the perpetrators hold attitudes condoning the use of violence (O’Keefe, 1998).
These findings are also consistent with social learning theory, which posits that through
early exposure to violence, interpersonal violence becomes instilled as an acceptable
78
behavior in an individual’s mind; and that, in turn, can lead to perpetration of violence in
adulthood (Eriksson & Mazerolle, 2015). Even though we can clearly see the role of
cognitive mechanism in facilitating behavior, we still need to examine the dynamics of
specific types of cognitive beliefs in the social learning process that lead to the use of
violence.
The study’s findings also show that substance abuse mediates the effects of
exposure to sexual violence on psychological violence perpetration. However, substance
abuse has no mediating effect in other pathways – including the pathway from exposure
to physical violence to physical dating violence perpetration. This finding, in particular,
contradicts previous studies demonstrating that substance abuse is associated with
physical dating violence (Foshee et al., 1999; Reyes et al., 2011; Roudsari et al., 2009).
The difference in our study’s findings might be related to differences in measurements.
Prior studies used substance abuse or binge drinking scales to measure the severity of
drug and alcohol abuse (Rapoza & Drake, 2009; Roudsari et al., 2009). However, this
study assessed substance abuse through general patterns of alcohol and drug use rather
than through specific symptoms of abuse. Thus, the effects of substance use might be
assessed differently in our study than in other studies. Meanwhile, there may be other
explanations for our finding that substance abuse is related to psychological violence
rather than physical violence, but for now the question remains unanswered. Further
research is needed to explore the ways in which substance abuse relates to particular
types of violence perpetration. Although our findings regarding the pathways that involve
substance abuse are mixed, the results are still partly congruent with general strain theory,
which suggests that exposure to sexual violence may provoke substance abuse as a
79
coping mechanism in dealing with trauma – and that coping with trauma may involve
committing psychological violence.
Regarding gender differences, I found only one link – from substance abuse to
psychological violence perpetration – which varied by gender. In this link, substance
abuse was a significant predictor of psychological dating violence by female students, but
not by male students. This pattern differs from the findings of some previous studies
which suggest that male drinking patterns are more likely to combine binge drinking with
aggressive behavior (Boles & Miotto, 2003; Hove et al., 2010; Luthra & Gidycz, 2006).
Meanwhile, I found no gender differences in any other paths leading from exposure to
violence to dating violence perpetration. In this way, this study varies from previous
studies which state that male students who were exposed to violence are more likely to
approve of violence than their female counterparts (Bacchini et al., 2015; Mattingly &
Straus, 2008). As previously noted, I found no gender difference in the path between
exposure to violence and attitudes toward violence. One possible reason for the lack of
gender moderation may be that this study measured attitudes toward interpersonal
violence, including family violence, male violence, and sexual aggression, while previous
studies focused on attitudes toward violence against women or on intimate partner
violence (Bacchini et al., 2015; Flood & Pease, 2009). Attitudes toward interpersonal
violence might vary along with attitudes toward gender-based violence or intimate
partner violence.
6. Limitations
There are several limitations to the current study. First, since the study was
crosssectional, it did not allow for causal prediction of variables. Longitudinal studies are
80
needed in order to examine the long-term effects of exposure to violence during
childhood on violence perpetration in later life. Second, the study sample was
conveniently drawn from only 17 universities across the U.S. The study sample consisted
of undergraduates enrolled mostly in the social sciences such as criminology, sociology,
psychology, and family studies. Random sampling and more diverse samples from other
majors will increase the generalizability. In addition, the survey questions regarding
exposure to violence and attitudes towards violence focused on physical violence. Future
research should consider other forms of violence in order to examine all association
between exposure to violence and attitudes toward violence.
7. Implications
This study is an integrated theory-driven examination of multiple mediators in the
relationship between exposure to violence and dating violence perpetration. Expanding
our knowledge of the factors that lead to dating violence in college settings is critical in
order to develop interventions which will contribute to the security and wellbeing of
young adults. The preceding results identify attitude toward violence as the primary
mediator in the relationship between early exposure to violence and later dating violence
perpetration by male and female students. This study also suggests a possible link
between substance use and dating violence perpetration. Dating violence prevention
programs should target multiple risk factors for aggression, such as positive beliefs about
aggressive behavior, substance abuse, gender, and a history of exposure to violence.
Target-based intervention through screening of individual risk factors for aggression
would provide college students, high-risk groups in particular, with effective educational
and skill-building materials (Shorey et al., 2011). In addition, although existing
81
interventions focus largely on physical dating violence, practitioners should also consider
how substance abuse and acceptance of violence influence psychological violence in
intimate relationships.
82
CHAPTER 5
CONCLUSION
Overall, this dissertation aimed to examine relevant correlates in the relationship
between early exposure to violence and U.S. college students’ attitudes toward violence
as well as U.S. college students’ dating violence perpetration, using International Dating
Violence Study (IDVS) data.
The first study (Chapter 2) examined how several types of violent socialization
during childhood differently influenced U.S. college students’ acceptance of interpersonal
violence. The findings showed that advised violence, i.e., verbal endorsement of
violence, had stronger associations with students’ acceptance of interpersonal violence
than childhood experiences of violent victimization and witnessed violence. Given that
existing studies of violent socialization have focused largely on childhood witnessing of
violence and violent victimization, our findings add the new knowledge that family and
community members’ pro-violence messages play a significant role in college students’
acceptance of interpersonal violence.
The second study examined the moderating effect of pro-violence messages in the
relationship between early exposure to violence and later acceptance of interpersonal
violence, as well as dating violence perpetration. The results showed that pro-violence
messages moderated the relationship between exposure to violence and attitudes
accepting violence among female students, and the messages moderated the effects of
exposure to violence on dating violence perpetration among male students. These
83
findings clearly show that pro-violence messages have a different effect on college
students’ cognition (attitude toward violence) than on their actual behaviors (dating
violence perpetration), and that the effects of pro-violence messages vary according to
gender. Thus, our findings add new knowledge to the study of violent socialization:
messages from parents, peers, and other community members which support violent
coping will have different effects on individuals’ cognition than on actual violent
behavior.
The third study examined the possible mediating effect of pro-violence attitudes
and substance use in the relationship between early exposure to violence and later dating
violence perpetration. The findings highlight the significance of attitude toward violence
as a common mediator in the effects of exposure to violence – whether physical or sexual
– on physical and psychological violence perpetration. However, substance abuse was a
significant mediator only in the relationship between exposure to sexual violence and
psychological violence perpetration. Furthermore, of all the pathways from exposure to
violence to dating violence perpetration, the only gender difference found was in
substance use, which predicted psychological violence perpetration in the female student
group but not in the male student group. This study expands our knowledge of the factors
that lead to dating violence in college settings, and thus is critical in determining the
appropriate interventions to prevent dating violence perpetration by young adults.
Limitations
This dissertation has limitations. First, the study was cross-sectional; thus, it did
not allow for causal prediction of variables. Longitudinal studies are necessary to
examine the long-term effects of exposure to violence during childhood on violence
84
perpetration in later life. Second, the study sample was conveniently drawn from 17
universities across the U.S., and participants consisted of undergraduates enrolled mostly
in the social sciences. Thus, future researchers must be careful in applying this study’s
findings to college samples which have different characteristics. The final limitation of
the present study lies in its measurements. In particular, exposure to violence and
attitudes toward violence cover different types of general violence patterns rather than
specific types of violence exposure or attitudes toward particular type of violence (e.g.,
male-to-female violence). This measurement might give different results in comparison to
previous studies which focused on particular types of violence. In addition, some
measurements, such as exposure to violence (Chapter 3) do not possess strong reliability
(Cronbach’s alpha = 0.66). Future studies should be conducted using measurement items
with enhanced reliability and validity.
Implication for social work research and practice
This dissertation indicates that early intervention is critical in preventing the
effects of childhood socialization to violence on attitudes toward violence and actual
interpersonal violence in early adulthood. Thus, the findings highlight a need for training
school social workers, teachers, policymakers, and other individuals who work closely
with children to recognize signs of early socialization to violence. In addition,
practitioners who work with communities should seek to change attitudes that support
violence. Parents, in particular, should take responsibility for discourse with their
children about how to cope with conflict. Enhancement of verbal skills and preservation
of non-violent environments for children are good ways to prevent potential violent
attitudes and behaviors among college students.
85
Interventions for college students might include social awareness programs
covering content such as pro-violence norms and the dangers of tolerating aggression. In
addition, dating violence prevention programs should target multiple risk factors for
dating violence perpetration, including pro-violence attitudes, substance abuse, and
individuals’ histories of exposure to violence. Screening participants for their individual
risks of violence perpetration would allow target-based prevention programs. Finally,
although existing dating violence interventions focus largely on physical violence,
practitioners should also consider how other types of violence might be prevented and
reduced.
