Co-Occurrence of Witnessed Parental Violence and Child Physical Abuse from a National Sample of Adolescents

Abstract

This study examined the relations between witnessed parental violence (PV) and child physical abuse (CPA) over a 1 year period among a nationally representative sample of 2,514 adolescents, ages 12–18. History of witnessed PV (Wave 1) prospectively predicted new experiences (controlling for abuse history) and first experiences of CPA reported at Wave 2. Conversely, history of CPA predicted new experiences of PV, but not first experiences. For adolescents who reported witnessed PV and CPA, witnessed PV preceded CPA in 70 % of cases. Most common configuration was single-perpetrator of violence. Additional perpetrator and sequencing configurations are reported. Study findings addressed several limitations in the literature by including use of adolescent report, longitudinal design, inclusion of perpetrator identity, and a nationally representative sample.

Introduction

Literature on interpersonal violence has indicated that a large proportion of children witness physical parental violence (PV) in their lifetimes, with prevalence estimates ranging from 8 to 37 % (Appel and Holden 1998; Bourassa 2007; Finkelhor et al. 2013; Zinzow et al. 2009). For example, results from a nationally representative sample of adolescents (Kilpatrick et al. 2000) found that 8.9 % had witnessed some form of physical parental violence, which equates to 2.3 million adolescents in the United States (U.S.) (Zinzow et al. 2009). Of these events, the most frequent form of PV involved witnessing a parent punch, hit, or kick another parent (about 1 in 16 adolescents). Further highlighting the prevalence of PV, in a community-based sample in the U.S., McCloskey and Walker (2000) found that 20 % of school-age children reported that they had seen their mother slap their father, and in their national study of violence exposure among children and youth, Finkelhor et al. (2013) found that 34.5 % of the oldest youth in the sample (ages 14–17) had witnessed some type of family physical violence in their lifetimes, and 28.3 % reported PV, specifically.

Researchers have given particular attention to examining correlates of witnessed PV, given that exposure has been linked to detrimental outcomes for children and adolescents, including internalizing and externalizing problems, social problems, posttraumatic stress disorder, and major depression (Buka et al. 2001; Howells and Rosenbaum 2008; Kitzmann et al. 2003; Osofsky 2003; Zinzow et al. 2009). In addition to conferring increased risk for mental health difficulties, witnessed PV has been consistently associated with child maltreatment, including both physical and sexual abuse (Appel and Holden 1998). While PV is associated with a range of child maltreatment types, it is most commonly associated with child physical abuse (CPA), and prevalence estimates on the co-occurrence of PV and CPA have ranged from 4 to 100 % (Appel and Holden 1998; Edelson 2001), depending on the sample and the source of the abuse report (e.g., child protective services; clinical vs. general population samples).

Studies also have indicated that exposure to multiple forms of violence is significantly worse than exposure to a single form of violence. For example, Bourassa (2007) examined consequences of PV alone and also combined with direct victimization, specifically CPA, and found that adolescents who had experienced both were more likely to report internalizing (21 % versus 37 %) and externalizing (6 % versus 30 %) symptoms. One study, which investigated rates of PV among mothers involved with social services for perpetrating neglect or physical abuse towards their children, found that when mothers were exposed to PV, they were significantly more likely to be re-referred to social services for repeated perpetration, and were re-referred almost twice as quickly than maltreating mothers who did not report PV (Casanueva et al. 2009). Although studies have examined the co-occurrence and consequences of witnessed PV and CPA, a number of questions remain regarding the nature of the relation, including mutual influence between the two and patterns of co-occurrence within the home. Appel and Holden (1998) developed a conceptual framework, which was expanded by Jouriles et al. (2008) for understanding the co-occurrence of PV and CPA. This framework outlined and reviewed existing literature for both unidirectional and bidirectional models of influence. Unidirectional models included single perpetrator (partner and child are both passive victims of the abuser), sequential perpetrator (abused partner responds to victimization by abusing child), and dual perpetrator (one partner is abused, and both partners abuse the child). Bidirectional models included marital violence (both partners are violent towards one another, and one or both may abuse the child) and family dysfunction (all family members are violent towards each other). Early research primarily investigated the single perpetrator model, with at least one study investigating each other configuration of family violence in Appel and Holden (1998). Since then, research has continued to investigate sequelae of dual exposure to PV and CPA (Kitzmann et al. 2003), yet less is known about patterns of violence precedence within families (i.e., rates at which PV precedes CPA and vice versa) or rates of the above described models of co-occurrence based on perpetrator and victim characteristics. Two investigations, one of clinically-referred adolescents and one community sample of families with young children, have estimated relative frequencies of these PV/CPA configurations within families, with marital violence configurations accounting for approximately 50 % of homes with PV/CPA co-occurrence and <20 % representing sole perpetrator and sequential perpetrator configurations in both of these studies (Mahoney et al. 2003; Slep and O’Leary 2005). In sum, research has continued to substantiate overall findings that both perpetrators and victims of PV are at increased risk of physically abusing their children (Dixon et al. 2007; Lutenbacher 2002; Casanueva et al. 2005; Coohey 2004). These findings necessitate a more nuanced understanding of the complex patterns of CPA/PV co-occurrence to best inform prevention and intervention programming efforts.

A complex and transactional ecological framework is useful in understanding the co-occurrence of multiple forms of family violence over time and mechanisms of influence between violence types within families (Bronfenbrenner 1992; Straus and Gelles 1990). At the individual and family levels, conceptualizations exist for different configurations of PV/CPA co-occurrence; for example, the aggressive individual as the perpetrator of both forms of family violence (McCloskey 2001), and stress spill-over from the marital/partner relationship to the parent–child relationship as a cause of sequential perpetration (i.e., PV victims as perpetrators of CPA; Margolin and Gordis 2003). Additionally, it is plausible to assume that a PV victim as well as the child witness may believe parental authority has been compromised by PV. This could potentially impact the child’s response to their parent’s discipline efforts and/or cause the caregiver to engage in coercive or abusive parenting as a means of gaining back this perceived loss of control. Relatedly, many perpetrators of PV explicitly undermine their partner’s parenting as an abusive practice (LaPierre 2010). Thus, this undermining of parental authority may increase maladaptive parent–child interactions and an overreliance on harsh punishment, highlighting potentially reciprocal influences between violence in the parental and parent–child dyads. At broader contextual levels, social factors, such as poverty, income/employment, cumulative stress, and social support, may potentiate violence within at-risk families (Gewirtz and Edleson 2007; Herrenkohl and Herrenkohl 2007). The present study represents a first step in linking epidemiological findings regarding the co-occurrence of violence types within homes, and research investigating mechanisms linking violence types within families by taking a more detailed perspective on CPA and PV events reported by a nationally representative sample of adolescents.

To this end, the goals of the present study are to investigate the longitudinal relations between history of witnessed PV and CPA over a 1 year period in a nationally representative sample of adolescents, and to examine frequencies of different configurations of PV and CPA co-occurrence. Given the high incidence of PV and CPA, as well as continued variation in estimates of co-occurrence, the current study focuses specifically on the PV-CPA relation.

This investigation will build on limitations of the current literature in a number of ways. First, information has typically relied on child protection case records, which tend to either over- or under-estimate the actual prevalence of violence (Chaffin and Valle 2003; Edelson 1999); or on parental self-report, which tends to be of questionable validity because parents may be reluctant to admit violence due to the potential consequences (Ammerman 1998). Second, many studies investigating the co-occurrence of PV and CPA have been cross-sectional in nature, rather than examining the longitudinal associations (Appel and Holden 1998; Hartley 2002). In addition, previous studies have typically relied on selective or idiosyncratic samples, with limited generalizability, such as women involved in social services due to physical abuse perpetration, or women residing in domestic violence shelters (Bowker et al. 1988; Shipman et al. 1999). Although a few nationally representative samples have examined the relations among CPA and PV, these studies either were conducted over two decades ago and may not indicate current rates of PV and CPA (Edelson 1999; Gelles and Straus 1988; Straus et al. 1980), are cross-sectional in nature (Zinzow et al. 2009), or consider only some configurations of violence (e.g., female victims of PV who perpetrate CPA; Casanueva et al. 2009). Finally, many studies investigating the prevalence of the co-occurrence of PV and CPA have not taken into account the identity of the perpetrator (Kohl et al. 2005), considered whether one form of family violence precedes the other, or have focused solely on male perpetrated violence.

Aside from contributing to existing literature, further understanding of temporal relations between witnessed PV and CPA and configurations of co-occurrence in homes with both PV and CPA is needed to inform prevention and intervention efforts, especially since the co-occurrence of PV and CPA is often not adequately assessed and/or overlooked by child protective services (Kohl et al. 2005; Casanueva et al. 2009). The following hypotheses were derived from our review of the literature: (1) Prior exposure to PV would be associated with increased odds of experiencing any CPA incident from Wave 1 to Wave 2 after controlling for history of CPA, as well as increased odds of experiencing a first CPA (new onset) experience during the 1 year period. (2) Prior exposure to CPA would similarly be associated with increased odds of any witnessed PV incident from Wave 1 to Wave 2, controlling for history of PV. (3) In contrast, we did not hypothesize that history of CPA would be predictive of new-onset PV during the 1 year period. (4) For adolescents reporting both CPA and witnessed PV, age of onset for PV would, on average, precede age of onset for CPA. (5) A range of perpetrator configurations would be present, representing both male and female perpetration of both PV and CPA, as well as CPA perpetrated by both the perpetrators and victims of PV although no more specific hypotheses were set regarding rates of these configurations.

Method

Overview

Data for the present investigation were utilized from the National Survey of Adolescents-Replication (NSA-R). The NSA-R was initiated in 2005 as an epidemiological study of youth ages 12 to 17 years using computer-assisted telephone interviewing technology. The primary goals of the NSA-R were to investigate population prevalence of traumatic events and major life stressors; investigate population prevalence of trauma-related mental health conditions; and examine risk factors associated with exposure to traumatic events and mental health outcomes.

Participants

The NSA-R included a national household probability sample and an oversample of urban-dwelling adolescents. A total of 6,694 households were contacted during recruitment, in which parents completed a brief structured interview and were asked to identify at least one eligible adolescent. Of these, 1,268 (18.9 %) parents refused adolescent participation; 188 (2.8 %) adolescents refused to be interviewed following parental consent; 119 (1.8 %) adolescent interviews were initiated but not completed; and 1,505 (22.5 %) identified eligible adolescents were unreachable or not available for interview. Thus, 3,614 cases resulted in complete adolescent interviews at Wave 1, including 2,459 in the national cross-section and 1,155 urban-dwelling adolescents. Of these adolescents, 69.6 % completed the follow-up assessment at Wave 2 (mean length of time between Wave 1 and 2 = 15.44 months, SD = 4.60 months). To examine attrition of the sample over time, effect sizes were calculated to compare participants who completed versus those who did not complete the Wave 2 assessment. Effect sizes were small (Cohen 1988) for all variables, with one statistically significant difference emerging on assessment of witnessed parental violence (completers were slightly less likely to report PV at Wave 1; Cramer’s V = 0.06, small effect).

Because adolescents were oversampled in urban areas, cases were weighted to maximize representativeness of the sample to the 2005 U.S. adolescent population. A weight was created to restore the urban cases to their true proportion of the urban/suburban/rural variable, based on 2005 U.S. Census estimates. Next, weights were created to adjust the weight of each case based on age and sex. We generated sample frequencies by age cohort and sex and compared this distribution to the U.S. Census estimates, and weightings were assigned to each sex × age group cell within the sample. This procedure resulted in weighted sample distributions that closely approximated 2005 U.S. Census estimates. The weighted sample of 2,514 participants at Wave 1 included 1,279 boys and 1,236 girls, ages 12–18 (M = 14.47, SD = 1.69). Sixty-eight percent (N = 1,721) self-identified as Caucasian, 13 % (N = 331) as African American, 10 % (N = 257) as Hispanic, 5 % (N = 127) as other racial/ethnic backgrounds.

Measures

A highly structured telephone interview with specially trained interviewers was used to collect information regarding a variety of topics, including demographic characteristics, parental violence, and child physical abuse. To increase accuracy of responses, the interview included behaviorally specific terminology and introductory statements to orient adolescents to questions (Kilpatrick et al. 2000). Wave 1 responses were based upon lifetime prevalence of victimization, while Wave 2 assessed victimization incidents that occurred since the Wave 1 interview.

Demographic Characteristics

Demographic characteristics including adolescent age at time of interview, gender, and race/ethnicity, were assessed with standard questions used by the U.S. Bureau of the Census (1988) and coded into dummy-coded variables (e.g., Caucasian, African American, Hispanic, or Other). Yearly family income was coded in the following ranges: <$20,000, $20,000–$50,000, and > $50,000.

Parental Violence (PV)

PV was assessed using yes/no responses to the following: “Have you ever seen or heard:” (a) “one of your parents punch or hit the other one with their fist or kick them real hard;” (b) “one of your parents choke the other;” (c) “one of your parents beat up the other so that they were hurt pretty bad;” (d) “one of your parents hit the other with an object like a bat, pan, or lamp and they were hurt pretty bad;” or (d) “one of your parents threaten the other with a gun, knife, or other weapon?” Participants reported on their age at the time, as well as the perpetrator and victim identity for up to three instances of PV.

Child Physical Abuse (CPA)

CPA was assessed using yes/no responses to items from the physical abuse and physical assault modules of the interview. Physical abuse items asked, “Has a parent or other adult in charge of you ever:” (a) “locked you in a closet or tied you up;” (b) “thrown you across the room or against a wall, car, floor, or other hard surface and you were hurt pretty badly;” (c) “slapped or spanked you so hard that it caused bruises, cuts or welts, or you had to see a doctor or nurse because of it,” (d) “beaten you up, hit you with a fist, or kicked you hard;” and/or (e) “grabbed you around the neck and choked you?” Participants reported on their age at the time and the identity of the perpetrator for up to three instances. Physical assault items were also included when participants indicated that the perpetrator of the physical assault was a mother, father, stepmother, stepfather, mother’s boyfriend, or father’s girlfriend. Physical assault items asked, “Has anyone—including family members, dating partners, or friends—ever:” (a) “Attacked you with a gun, knife, or some other weapon, regardless of when it happened or whether you ever reported it to the police;” (b) physically attacked you without a weapon, but you thought they were trying to kill or seriously injure you;” and/or (c) “ever beat you up, attacked you, or hit you with something like a stick, club, or bottle so hard that you were hurt pretty bad?” In addition to identity of the perpetrator, participants also reported their age at the time of the event for physical assault items. Instances were combined across physical abuse items and physical assault items (perpetrated by a caregiver) and were collectively considered incidents of CPA.

Procedures

Recruitment of participants began following approval by the Institutional Review Board of the Medical University of South Carolina. Sample selection and interviewing were conducted by Abt-SRBI, a New York – based survey research firm. To conduct the initial probability sample, Abt-SRBI used a multistage, stratified, area probability, random-digit-dialing six-stage sampling procedure.

Interviews were conducted in English using computer-assisted telephone interviewing technology. First, interviewers provided parents with a brief description of the study and interview topics, informed them that the adolescent could refuse to answer any questions or terminate the interview at any time, and obtained parental permission to contact the adolescent. Then, adolescent assent to participate was obtained before the interview began. Adolescent participants received a check for $10 for their participation in each interview.

Two steps were taken to increase adolescents’ comfort to answer interview questions freely and privately. First, interviewers asked whether the adolescent was in a situation where privacy could be assured and he/she could answer in an open and honest manner. If the adolescent indicated that he/she could not, the interviewer offered to reschedule the interview. Second, the interview was primarily designed with closed-ended questions (i.e., yes or no) to reduce the likelihood of negative consequences if others were listening.

Several steps were taken to recontact participants for the Wave 2 assessment, including (1) contacting participants at the telephone number on file, (2) obtaining three additional telephone numbers from past residences using ChoicePoint Credit Information Bureau, (3) sending letters to the last known address, and (4) recontacting ChoicePoint after 3 months to obtain updated telephone numbers.

Results

Table 1 presents the sample characteristics and frequencies for all independent variables in the study. Results indicated that 11.9 % of adolescents reported CPA, and 7.9 % of adolescents reported PV at Wave 1; at Wave 2, 2.7 % reported new onset CPA and 1.3 % reported new onset PV since Wave 1. Adolescents reporting PV at Wave 1 indicated that 39.0 % of these incidents were perpetrated by the mother or mother figure (e.g., stepmother) and 54.0 % by the father or father figure (i.e., stepfather or mother’s boyfriend); 7.0 % of adolescents reported ‘don’t know’ or refused to disclose the perpetrator. Adolescents reporting CPA at Wave 1 indicated that 37.7 % of these incidents were perpetrated by a father or father figure (e.g., stepfather, mother’s boyfriend), 30.9 % by a mother or mother figure (e.g., stepmother), 24.5 % by another adult (e.g., grandparent, uncle, aunt); and 6.9 % of adolescents refused to disclose their relationship to the perpetrator. Chi-square analyses indicated that adolescents reporting PV at Wave 1 were significantly more likely to disclose new CPA events at Wave 2, χ ^{2 (1)} = 50.58, p < 0.001; additionally, adolescents reporting CPA at Wave 1 were significantly more likely to disclose new PV events at Wave 2, χ ^{2 (1)} = 35.88, p < 0.001.

Table 1.

Frequencies for all independent variables

Predictor	n	%
Age (Wave 1)
11–13	835	33.2
14–15	864	34.4
16–18	816	32.4
Race/Ethnicity
Caucasian	1721	68.4
African American	331	13.2
Hispanic	257	10.2
Other	127	5.1
Gender
Male	1279	50.8
Female	1236	49.2
Income
< $20,000	234	9.3
$20,000–$50,000	694	27.6
> $50,000	1308	52.0
Witnessed parental violence (Wave 1)
No	2315	92.1
Yes	199	7.9
Witnessed parental violence (Wave 1–2)
No	2481	98.7
Yes	33	1.3
Physical abuse (Wave 1)
No	2215	88.1
Yes	299	11.9
Physical abuse (Wave 1–2)
No	2446	97.3
Yes	68	2.7

Note. Weighted n = 2514

A series of four separate logistic regression analyses were conducted to investigate longitudinal associations between CPA (any new experience, first experience) and witnessed PV (any new experience, first experience). SPSS (version 20) was used for analyses to account for complex survey design and sample weighting as described. Table 2 reports the odds ratios (ORs), 95 % confidence levels surrounding the ORs, and the significance level for the logistic regression analyses predicting CPA (any new experience, first experience), and Table 3 reports the same information for logistic regression analyses predicting witnessed PV (any new experience, first experience).

Table 2.

Logistic regression results: prediction of wave 2 physical abuse

Predictor	Any physical abuse (W1–W2)			New onset physical abuse (W1–W2)
	OR	95 % CI	p-value	OR	95 % CI	p-value
Age
11–13	1.00	–	0.50	1.00	–	0.47
14–15	1.49	0.75–2.45		1.67	0.70–3.94
16–18	1.39	0.69–2.79		1.19	0.47–3.06
Race/Ethnicity
Caucasian	1.00	–	0.46	1.00	–	0.96
African American	0.82	0.34–1.95		1.31	0.50–3.47
Hispanic	1.52	0.73–3.16		1.09	0.35–3.37
Other	1.67	0.62–4.47		1.12	0.24–5.16
Gender
Male	1.00	–	0.45	1.00	–	0.57
Female	1.23	0.72–2.10		1.23	0.61–2.48
Income
< $20,000	1.00	–	0.80	1.00	–	0.40
$20,000–$50,000	1.36	0.53–3.47		1.76	0.47–6.64
> $50,000	1.33	0.54–3.28		1.43	0.38–5.39
Physical abusea
No	1.00	–	<0.001	–	–	–
Yes	4.83	2.68–8.71
Witnessed parental violencea
No	1.00	–	0.002	1.00	–	0.01
Yes	2.85	1.48–5.48		3.09	1.27–7.49

Note. ^a prior to W1

Table 3.

Logistic regression results: prediction of wave 2 partner violence

Predictor	Any witnessed partner violence (W1–W2)			New onset witnessed partner violence (W1–W2)
	OR	95 % CI	p-value	OR	95 % CI	p-value
Age
11–13	1.00	–	0.97	1.00	–	0.26
14–15	1.07	0.41–2.81		0.25	0.05–1.33
16–18	0.97	0.36–2.63		0.64	0.18–2.31
Race/Ethnicity
Caucasian	1.00	–	0.38	1.00	–	0.23
African American	2.20	0.81–6.01		3.97	1.05–15.04
Hispanic	1.86	0.66–5.20		2.59	0.52–12.92
Other	0.95	0.14–6.38		0.00	0.00	–
Gender
Male	1.00	–	0.76	1.00	–	0.94
Female	1.13	0.51–2.51		0.96	0.31–2.97
Income
<$20,000	1.00	–	0.83	1.00	–	0.97
$20,000–$50,000	0.94	0.29–3.08		0.88	0.17–4.67
>$50,000	1.23	0.40–3.82		0.80	0.15–4.22
Witnessed parental violencea
No	1.00	–	<0.001	–	–	–
Yes	10.41	4.35–24.91
Physical abusea
No	1.00	–	0.03	1.00	–	0.81
Yes	2.62	1.10–6.25		1.22	0.25–6.00

Note. ^a prior to W1

Does Witnessed PV Predict CPA?

In the first logistic regression, Wave 1 witnessed PV was investigated as a predictor of new experiences of CPA (between Waves 1 and 2), controlling for Wave 1 CPA (lifetime CPA history) and demographic characteristics (age, ethnicity, gender, family income) entered in a single step. In the prediction of CPA at Wave 2, not surprisingly, the strongest predictor was reported history of CPA at Wave 1 (OR = 4.83). However, even after controlling for Wave 1 CPA, witnessed PV at Wave 1 more than doubled the likelihood of Wave 2 CPA (OR = 2.85). Age, racial/ethnic status, gender, and family income were not significant predictors of Wave 2 CPA.

Does Witnessed PV Predict New Onset CPA?

The second logistic regression investigated Wave 1 PV as a predictor of new experiences of CPA that were first CPA experiences (i.e., youth reported their very first CPA experience between Wave 1 and 2) controlling for demographic characteristics. In the prediction of new onset CPA, history of witnessed PV at Wave 1 was associated with increased odds (OR = 3.09) of reporting new experience of CPA at Wave 2. No demographic predictors emerged.

Does CPA Predict Witnessed PV?

The third and fourth logistic regressions investigated the reverse: CPA as a predictor of subsequent PV experiences of PV between Wave 1 and 2, controlling for history of PV and demographics, as well as very first experiences of PV between Wave 1 and 2 controlling for demographics. In the prediction of witnessed PV at Wave 2, the strongest predictor was reported history of witnessed PV at Wave 1 (OR = 10.41). After controlling for Wave 1 PV, prior CPA more than doubled the likelihood of Wave 2 PV (OR = 2.62). Age, racial/ethnic status, gender, and family income were not significant predictors of Wave 2 witnessed PV.

Does CPA Predict New Onset Witnessed PV?

In the prediction of new onset (first experience) of witnessed PV at Wave 2, neither history of CPA nor demographic predictors were associated with increased likelihood of new onset PV.

Configurations of Comorbid CPA and Witnessed PV

Across all adolescents, average age of onset for witnessed PV was 8.6 years and average age of onset for CPA was 10.3 years. Additional descriptive statistics were conducted on the subgroup of adolescents who reported both witnessed PV and CPA to better characterize adolescents’ reports of family co-occurrence. By Wave 2, 98 (3.90 %) adolescents reported a history of CPA and witnessed PV; 60 had complete data regarding age of onset and perpetrator identity and reported CPA perpetrated by mother/mother figure or father/father figure. Seventy percent of adolescents reported an earlier age of onset for witnessed PV; 23 % reported that CPA occurred first; and 7 % reported that both occurred first within the same year (i.e., same age of onset).

Regarding configurations of family violence, single perpetrator CPA/PV was most common (50 %), followed by sequential perpetration (32 %; victim of witnessed PV is perpetrator of CPA). Less common were dual perpetration (10 %; both parents perpetrated CPA, only one parent perpetrated PV) and family violence (8 %; both parents perpetrated PV and one or both parents perpetrated CPA). For configurations characterized by single perpetrator DV or CPA, sex of perpetrator was also of interest. For single perpetrator configurations, 44 % of perpetrators were mothers/mother figures and 56 % were fathers/father figures. Sequential perpetrator configurations were very similar. Forty-five percent of adolescents within this configuration reported that their mother/mother figure was the perpetrator of the witnessed PV and a father figure was the perpetrator of the CPA. Conversely, 56 % of adolescents reported that their father figure was the perpetrator of the witnessed PV and a mother figure was the perpetrator of the CPA. In dual perpetration configurations (where both parents engaged in CPA), the witnessed PV was perpetrated by the father figure in 80 % of cases.

Discussion

The primary purpose of this study was to examine the prospective relation between PV and CPA among a nationally representative sample of adolescents across a 1 year period. As hypothesized, adolescents who reported witnessing violence between their parents at the first assessment were significantly more likely to report physical abuse exposure 1 year later. This finding was consistent across two sets of analyses 1) analyses that controlled for history of CPA and 2) analyses that considered only first incidents of CPA over the 1 year period. Thus, history of witnessed PV conferred risk for subsequent CPA during adolescence, both for teens who had previously experienced CPA and those who had not. History of CPA was also investigated as a risk factor for subsequent witnessed PV during adolescence. Results indicated that history of CPA increased risk for witnessed PV over the 1 year period, above and beyond the effect of history of witnessed PV. Yet, history of CPA did not increase the risk for new onset witnessed PV. In other words, history of witnessed PV and history of CPA increased the risk that an adolescent would experience the other form of family violence over the next 1 year period. Yet, for teens that had not experienced CPA at Wave 1, witnessed PV increased the risk they would experience first incident of CPA by Wave 2, whereas history of CPA did not increase the risk of a first experience of witnessed PV by Wave 2. Consistent with this pattern of results, for families characterized by both PV and CPA by Wave 2, age of onset for PV was, on average, earlier than age of onset for CPA.

Both the theoretical and empirical literature on the co-occurrence between PV and CPA have largely focused on PV as a temporal predecessor and risk factor for subsequent CPA, rather than the opposite (i.e., CPA as a risk factor for PV; for exception see Emery et al. 2013). The results of the present study were consistent with this conceptualization, in that PV preceded CPA in 70 % of cases of co-occurrence, and that PV predicted new onset of CPA prospectively, whereas CPA did not predict new onset of PV over the 1 year period. Yet, history of CPA did confer additional risk for witnessed PV over the subsequent year, controlling for history of PV. Furthermore, a substantial number of teens (about ¼ of those reporting both CPA and witnessed PV) reported that CPA preceded witnessed PV. One interpretation of this pattern of results is that both forms of family violence (PV and CPA) play a role in maintaining the other form of family violence over time within violent families. However, PV contributes significantly to the onset of CPA, whereas in the present study, CPA did not contribute to the onset of PV. These results are consistent with ecological, transactional conceptualizations that PV likely precedes CPA in cycles of family violence (Appel and Holden 1998; Jouriles et al. 2008). However, both forms of family violence may increase risk for subsequent violence of either type, especially in the context of additional risk factors such as poverty, stress, and lack of social support (Jouriles et al. 2008; Margolin and Gordis 2003). Given these findings in this nationally representative sample, there is a need for future research to attend to specific mechanisms that foster the perpetration of CPA within families characterized by PV. Examples of such models include models investigating the functioning of aggressive individuals who perpetrate both violence types, and stress spill-over models that may explain the perpetration of CPA by both victims and perpetrators of PV.

Adolescent Context

To date, most studies examining these associations have primarily relied on cross-sectional designs, which preclude conclusions regarding directionality and/or have been conducted with young children (Bourassa 2007). Thus, a major contribution of the current study is the use of a nationally representative, longitudinal sample of adolescents. The focus on adolescents is important because, due to their age, they have had the opportunity to experience both PV and CPA for longer periods of time than younger children, which heightens the risk for long-term and potentially severe mental health, physical health, and social functioning problems. In line with the Cumulative Risk Model (Rutter 1981), an increase in the number of stressors (i.e., exposure to PV and CPA) likely corresponds to an increase in the number of behavioral and emotional difficulties. Additionally, although dual exposure to PV and CPA increases risk for behavior problems across age ranges, aggression and behavior problems more prominent during adolescence, including dating violence, physical assaults of peers, and other forms of delinquency, often have more serious consequences and further perpetuate intergenerational cycles of violence (Ehrensaft et al. 2003).

Need for Prevention/Intervention Efforts

Many studies have focused on the risk that PV confers for subsequent CPA during early stages of development, but this is the first investigation to identify PV as a predictor of first CPA occurring during the adolescent period. From a public health perspective, given that primary prevention targets modifiable risk factors (Klevens and Whitaker 2007), adolescents appear to be a particularly vulnerable, at risk group warranting prevention and intervention efforts. These findings support the importance of targeting early PV as a way of preventing future risk to the adolescent. Indeed, it has been noted that historically, co-occurrence of PV and CPA is often overlooked by child protective services (Kohl et al. 2005; Casanueva et al. 2009). This has led to several important assessment and practice recommendations including: improved identification of child maltreatment when parents are being screened/receiving services for PV, increased availability of preventive services for families already impacted by PV but not yet engaged in CPA, and improved awareness among clinicians and case workers about the risk of co-occurrence and the development and dissemination of assessment guidelines that encompass both CPA and PV (Casanueva et al. 2009).

Configurations of Family Violence

Another strength of this study was the ability to investigate the identity of the perpetrator of both the witnessed PV and the CPA, as this has not been routinely collected as part of broad survey investigations and is limited in studies with select samples (e.g., studies including only female victims of domestic violence; or studies including only women who have perpetrated child abuse). Although it is important to note that violence perpetrated by women is less likely to be associated with severe harm or child fatalities, (Kimmel 2002) current results supported prior findings indicating that women perpetrated violence in the home at nearly comparable rates to men. This male/female breakdown (i.e., approximately 45 % mothers and 55 % fathers) was similar for both single perpetrator homes (i.e., only one perpetrator of the PV and the CPA) and sequential perpetrator homes (i.e., one perpetrator of PV and the victim of PV perpetrated the CPA). Interestingly, when adolescents reported CPA by both of their parents, the PV was more likely to be perpetrated by their father or father figure (80 %), which likely reflects an overall more severe level of violence in the home.

Limitations

Several limitations are of note in the current study. First, as this study focused on relatively severe forms of PV and CPA, findings may not generalize to psychological or emotional abuse between parents (i.e., yelling, cursing, manipulation), or to less severe forms of interparental abuse (e.g., pushing/shoving; throwing objects without causing injury) that may be more frequent but not necessarily present higher risk for subsequent CPA. The same issues apply in our definitions of CPA – only severe forms were included; we did not include such behaviors as “hit with an object” or “spanking/slapped” when these behaviors did not leave a bruise, mark, or welt, as these items likely overinflate rates of CPA because of their widespread, and generally accepted use as discipline strategies in the U.S. (Appel and Holden 1998). Additionally, as a nationally representative sample was used, low base rates of new onset violence (PV and CPA) over a 1 year period likely contributed to wide confidence intervals for some estimates, particularly in the prediction of new onset PV.

As with any retrospective, self-report study, findings are subject to biased recall, minimization or over-exaggeration of difficulties. In addition, sole reliance on adolescent self-report means that the violence exposure events were not corroborated by additional sources, which likely underestimated true occurrence of parental violence. Further, the reliance on adolescent report means we were unable to account for violence that may have occurred very early in the adolescent’s development, and thereby not remembered, or violence between parents that the adolescent did not witness. No information was gathered regarding adolescent perpetration of violence against parents, and so family dysfunction configurations were not examined. An additional limitation relates to the fact that we were only able to collect incident details for up to three episodes of PV and CPA, and these incident details were used to evaluate configurations, which resulted in only one perpetrator and one victim per incident. Other studies, which have found higher rates of marital violence configurations as opposed to sole/sequential perpetrator configurations employed different methods, such as asking specifically whether each behavior has been displayed by both the mother and the father (Mahoney et al. 2003; Slep and O’Leary 2005). Neither of these methodologies comprehensively addresses the complex nature of family violence and self-defense—our methodology may have missed some instances of bidirectional violence between mothers and fathers (i.e., our conceptualization of perpetrator/victim within the incident may be limiting to complex scenarios), and other authors’ methodologies may attribute self-defense as additional violence (i.e., seeing one parent hit another in the context of self-defense). Additional studies will continue to shed light on the prevalence and differential correlates of configurations of family violence.

Additionally, only adolescents with household telephones were sampled which decreases generalizability of study findings. However, since data indicate that most households had adequate telephone coverage at the time of this study (Keeter et al. 2000), it is likely this sample is representative of the great majority of adolescents across the U.S. Relatedly, although steps were taken to ensure that adolescents felt comfortable answering questions (see Methods), the phone-based nature of the interview allowed for potential lack of privacy. For example, parents may have been present in the room, making participants uncomfortable to answer questions honestly, if PV or CPA had occurred. Since the completion of this interview, rates of internet access have continued to climb. As validation continues to build regarding the safe and confidential use of e-mail or internet-based interviews for nationally-representative research investigations, it will be important to consider such methods to further increase privacy. Finally, attrition could have influenced findings in the current study, as only 69.6 % of participants completed both waves of data collection. However, as noted above, small effect sizes indicated minimal differences between completers and non-completers on all relevant study variables.

Implications and Directions for Future Research

Despite these limitations, several notable strengths include the focus on adolescents to examine a longer time period for family violence exposure; a longitudinal design that allowed for the prospective examination of these associations over time; inclusion of perpetrator identity; and the use of behaviorally specific questions administered in an individual interview format to assess the targeted constructs. Regarding directions for future research, it will be important for studies to explore other factors that may increase risk for PV and CPA. Factors that deserve attention include parental variables (e.g., substance abuse, mental health status, legal history, employment and marital status) as well as other family and environmental characteristics (e.g., number and age of children in the home, biological relationship between parents and children, use/availability of a social support network). In addition, studies need to use clear definitions of abuse, with a focus on the more severe end of the spectrum because of the risk of greater harm, and further investigate other forms of partner violence associated with child abuse, such as psychological abuse. The present study also highlights the importance of considering violence perpetrated by both men and women and taking into consideration specific family contexts that may surround different configurations of family violence when designing and implementing targeted prevention and intervention programs. Qualitative approaches to better understand the lived experiences of individuals impacted by family violence are paramount in the development of sensitive and context-salient educational and intervention materials that take into account the nuances of victims’ experiences (LaPierre 2010). Relatedly, a better bridge between literature aimed to understand the romantic and social underpinnings of violent relationships could enhance our understanding of the co-occurrence, as social support and a two-parent family are protective factors for child abuse, whereas a violent partner is a risk factor for child abuse (Dixon et al. 2009). Finally, future research is needed to understand the mechanisms and family processes that underscore the link between PV and CPA, especially in the context of diverse configurations (i.e., differing mechanisms between single perpetrator and sequential perpetrator homes, as well as dual perpetrator homes). This will be particularly salient in the development of preventive and intervention strategies to serve families impacted by PV and/or CPA.