Prevalence of Family Violence in the United States

Fortunately, there has been a recent steep decline in intimate partner violence. Even so, there are approximately 700,000 violent crimes, including about 1,700 murders, committed by intimate partners in the United States each year (U.S. Department of Justice 1996, 2000a). Domestic violence is the leading cause of injury to women aged 15 to 44 (U.S. Department of Justice 1996 ), and it accounts for about one-third of the total number of murders of women in this country (U.S. Department of Justice 2000a). Approximately half of all victimized women live in a home with children under the age of 12 (U.S. Department of Justice 2000a), and an estimated 3.3 million children are exposed to violence against their mothers or female caretakers each year (American Psychological Association [APA] 1996; Jaffe et al. 1990).

About half of all men who abuse women also abuse their children or other children who live in their homes (APA 1996; Appel and Holden 1998; Straus and Gelles 1990). Children who live in homes where domestic violence occurs are 1,500 times more likely to be abused than those who live in homes without violence (U.S. Department of Justice 1993). The physical abuse of children is a component of family violence that is all too common in the United States: It is estimated that somewhere between a little less than 1% (U.S. Department of Justice 1998) and just over 5% of all children in this country experience physical abuse (Gallup 1995).

Outcomes Associated with Family Violence

A variety of short- and long-term negative outcomes have been associated with experiencing physical abuse as a child. In general, abused children seem to have behavioral, emotional, and social problems (see review by Malinosky-Rummell and Hansen 1993). Researchers commonly find that, as adults, abused children mature and display violence toward nonfamilial others (e.g., McCord 1983; Rosenbaum and Bennett 1986) as well as toward their children and intimate partners (e.g., Kempe et al. 1962; Straus et al.1980; Widom 1989).

Like direct experiencing of physical abuse as a child, witnessing of interparental (or a parent and his/her intimate partner) abuse as a child or adolescent has been linked to a number of negative outcomes, including aggressive and delinquent behaviors (Fantuzzo et al. 1991; Graham-Bermann and Levendosky 1998; Hershorn and Rosenbaum 1985), developmental and academic deficits (Pfouts et al. 1982), depression, anxiety, lower self-esteem, and somatic symptoms (Fantuzzo et al. 1991; Graham-Bermann and Levendosky 1998; Spaccarelli et al. 1994). The consequences of witnessing also appear to continue into adulthood, and long-term effects include depression, trauma, antisocial behaviors, substance use, general violence, and partner violence (Downs et al. 1996; Ehrensaft et al. 2003; Henning et al. 1997; Widom 1989).

Theory of Intergenerational Transmission of Violence

Many researchers have reported a link between violent childhood experiences (including witnessing domestic violence and/or being physically abused) and violent adult offenses, and this phenomenon is frequently called the intergenerational transmission of violence or aggression (e.g., Dutton et al. 1995; Jankowski et al. 1999; Sugarman and Hotaling 1989). One often-hypothesized mechanism of such transmission is observational learning (e.g., Grych and Fincham 1990; Holtzworth-Munroe and Stuart 1994; Jaffe et al. 1990), most commonly described as learning from modeling with a social learning perspective (Bandura 1973, 1977). Social learning theory suggests that a child learns not only how to commit violence but also learns positive attitudes about violence when he (or she) sees it rewarded (Dutton and Holtzworth-Munroe 1997; Kalmuss 1984). This suggests that children who have witnessed violence, or have been abused, learn destructive conflict resolution and communication patterns. Sternberg et al. (1997) suggest that Bandura’s social learning theory would predict that both observers and victims can be affected, with children from more violent environments being more likely to acquire aggressive modes of behavior.

There is a large body of evidence suggesting that social learning theory can account for violent behavior in general (e.g., Bandura et al. 1963; Ellis and Sekyra 1972; Plomin et al. 1981). Further, a few more recent tests of the social learning model have found evidence that family violence is learned, particularly through modeling (Kwong et al. 2003; Gwartney-Gibbs et al. 1987; Lystad 1986; Mihalic and Elliot 1997). Kalmuss (1984) proposed that the intergenerational transmission of family aggression involves both generalized and specific modeling: Generalized modeling refers to the acceptance of aggression within families, while specific modeling refers to the perpetration of particular types of aggression the individual was exposed to within the family of origin. In theory, families with high levels of aggression produce both generalized and specific forms of modeling. Children from homes where multiple forms of violence or severe violence occur are exposed to more modeling, which increases the probability that violence is learned and perpetrated (Kalmuss 1984). However, evidence of the future effects associated with specific modeling is mixed (Kwong et al. 2003; Stith et al. 2000).

There is some evidence that different types of childhood traumas are related to different patterns of abusive behavior in violent men. Dutton and Hart (1992) found that offenders who had been physically abused as children were more likely to commit crimes of physical aggression than sexual crimes. These researchers also found that men who commit family violence are more likely to report violence in their family of origin than men who commit nonviolent crimes and men who commit violent crimes against strangers (Dutton and Hart 1992). Numerous studies have found that adults who abuse their children are more likely to have been abused than the general population (e.g., Silver et al. 1969; Straus et al. 1980). It may be the case that individuals who were abused are more likely to abuse their children than individuals who witnessed violence, but were not themselves victims (Kalmuss 1984). There is some evidence that adults who witnessed violence, but were not themselves abused, are more likely to perpetrate domestic violence than those who were abused but did not witness violence as children. However, there is also evidence that adults are most likely to perpetrate domestic violence if they were both physically abused and witnessed domestic violence as children (Downs et al. 1996; Holtzworth-Munroe et al. 1997; Kalmuss 1984; Widom 1989), and some researchers have found no role-specific patterns of violence (Kwong et al. 2003). Family violence appears to be learned, although the roles of generalized and specific modeling are unclear. Given that most perpetrators are male, it is important to study how learned violence affects men.

Differences in Forms of Violence and Perpetrators

Previous studies looking at male perpetrators of domestic violence have found they are not alike with respect to severity and frequency of violence or levels of psychopathology (e.g., Hamberger and Hastings 1986; Holtzworth-Munroe and Stuart 1994). The most violent perpetrators are most likely to have been physically abused and witnessed domestic violence as children and also be violent outside of the home. These men report moderate levels of anger and often have antisocial personality traits. Less violent batterers are less likely to have experienced violence (either as a witness or victim) as a child and do not tend to show evidence of high levels of general violence or psychopathology (Holtzworth-Munroe and Stuart 1994; Waltz et al. 2000).

Domestic violence offenders do not often evidence severe mental disorders (Saunders 1999). However, they do often meet criteria for personality disorders, most commonly Antisocial, Borderline, Dependent, Depressed, and Narcissistic (Hamberger and Hastings 1986; Hamberger et al. 1996; Waltz et al. 2000). Generally, domestic violence perpetrators evidence more mood and psychotic disorders than nonviolent men (Hamberger and Hastings 1988), but there are differing degrees and patterns of psychopathology and offenses committed within the battering population.

Many previous studies (e.g., Hamberger and Hastings 1986; Holtzworth-Munroe and Stuart 1994; Tweed and Dutton 1998) have examined differences among domestic violence offenders. Yet, to date there is no research that clearly distinguishes offenders who were witnesses of violence from those who were abused as children. It appears from the social learning literature and previous work on family violence that differences between the groups exist, but the extent to which exposure relates to offense and psychopathology is still unclear. This study examined whether there are differences between perpetrators of domestic violence who, during childhood, witnessed domestic violence, were physically abused, neither witnessed nor were abused, or both witnessed and were abused.

Method

Participants

A sample of 1,099 male batterers (85% African-American, 14% Caucasian, and 1% another race or unreported), who ranged in age from 18 to 65 (with a mean age of 32), participated in this study. The sample represented the population of males arrested for battering in the area and is comparable in terms of age (national mean age is 31) but not race (national racial mix is more evenly balanced) of the perpetrator to national statistics. All participants had been court ordered between 1998 and 2002 for assessment at a domestic violence center in a Southern metropolitan city (see Table 1 for a summary of participant characteristics).

Table 1 Demographic characteristics of groups
Full size table

Measures

Measures used in this study were selected to assess three areas of theoretical interest: (a) generality of violence and nonviolent criminal behavior, (b) frequency and severity of domestic violence, and (c) psychopathology as evidenced by personality attributes. Specific items from the assessment protocol that were relevant to each of these areas were included as variables for analyses.

Generality of Violence and Nonviolent Criminal Behavior

The extent to which violent behavior generalized from intimate partner violence to other forms of criminal violence and nonviolent criminal behavior was assessed in two ways: (a) via two ratings based on police reports and (b) via the abuse scale score of the Child Abuse Potential Inventory (CAP; Milner 1986). Ratings were assigned by the domestic violence center interviewer on a three-point scale, with 0 indicating no prior arrest, 1 indicating one minor prior arrest, and 2 indicating one severe or two or more prior arrests, for non-intimate partner violence and for nonviolent charges, respectively.

The CAP score was used to assess generalization of intimate partner violence to the tendency to physically abuse children.

Frequency and Severity of Domestic Violence

Frequency and severity of domestic violence were assessed via selected items from a questionnaire designed for use at the domestic violence center. It is composed of items from the physical abuse (violence) section of the Conflict Tactics Scales (Straus 1979).

To measure physical spouse abuse, the following items’ ratings were summed: 1) threw something, 2) pushed, grabbed, or shoved, 3) slapped, 4) kicked, bit, or hit, 5) hit or tried to hit with something, 6) beat up, 7) choked, 8) threatened with a gun or knife, and 9) used a gun or knife. To assess severity, as previously done in the CTS literature (Straus and Gelles 1990), the ratings for items 5 through 9 were summed.

Psychopathology of the Offender

Five personality scales of the Millon (MCMI-III; Millon 1994) were used to assess psychopathology/personality attributes of theoretical interest. The individual scales were chosen for two reasons: (a) the MCMI-III, which is the most commonly used measure of psychopathology in the domestic violence literature, has no measure of overall distress, and (b) previous research has shown batterers differ with respect to the personality attributes measured by these scales (e.g., Holtzworth-Munroe and Atuart 1994). The five subscales are: Antisocial, Borderline, Dependent, Depressive, and Narcissistic.

Procedures

Each participant completed an assessment battery, which contained several questionnaires designed specifically for the center in order to assess information about the offender, the victim, and the recent offense. All forms were completed in a group format and were supplemented by an individual interview. The interviews typically lasted about 45 min and were conducted by advanced graduate students or Masters- or Doctoral-level professionals.

Participants were assigned to one of four groups (neither, witnessed only, abused only, both) on the independent variable, which represents status of exposure to violence as a child. This placement was determined by the participant’s answers to items about his exposure prior to the age of sixteen. Specifically, assignment of status was based on yes–no responses to items about having witnessed either or both parents being aggressive towards the other and yes–no items about having been abused as a child.

If a participant answered yes to any of the relevant items, he met criteria for inclusion in the category of reference. If he did not answer yes to any of these items, the participant was placed in the “neither” group (n = 517; 47%). If he answered yes to witnessed items, but endorsed no abused items, he was placed in the “witnessed only” group (n = 84; 8%). If he answered yes to abused items, but did not endorse any witnessed items, then he was placed in the “abused only” group (n = 300; 27%). If he answered yes to both witnessed and abused items, he was placed in the “both” group (n = 198; 18%).

Analyses

Bivariate correlational (Pearson product-moment) analyses were conducted to assess relationships among the offender characteristics. This was done in order to examine differences in demographic characteristics of offenders in each group, to determine whether any covariates were necessary in further analyses.

Analyses of variance were conducted to assess the three areas of interest. Childhood trauma status served as the independent variable; the four levels of the variable were: neither, witnessed only, abused only, and both. One-way Analyses of Variances (ANOVAs) were performed to assess differences in generality of violence, nonviolent criminal behavior, and frequency and severity of domestic violence. A Multivariate Analysis of Variance (MANOVA) was performed to assess differences in psychopathology, defined in terms of personality disorder.

For each ANOVA and the MANOVA performed, the assumptions of independence, homogeneity of variance, and normality of distribution were examined. Analyses of variances were deemed appropriate for each assessment. Due to concerns about unequal sample sizes and distributions increasing Type I error risk, alpha levels were set at .01 or less. Post-hoc comparison tests, specifically Dunnett’s T3 tests that do not assume normal distribution, were done as necessary to look for specific differences between groups.

Generality of Violence and Nonviolent Criminal Behavior

The dependent variables were: assigned 0 to 2 rating of number of prior assaults and violent offenses against a non-intimate partner, assigned 0 to 2 rating of number of non-violent offenses, and total CAP abuse scale score.

Frequency and Severity of Domestic Violence

The dependent variables were: the frequency (total sum of physical abuse items) score from the modified CTS items and the severity score (sum of items 5 through 9) from the modified CTS items.

Psychopathology of the Offender

The dependent variables were the Millon scores on the scales measuring Antisocial, Borderline, Dependent, Depressed, and Narcissistic personality disorders.

Results

Correlational Analyses

No significant relationships between childhood trauma status and the participant demographic variables of age, race, and level of education were found, and it was deemed unnecessary to use any covariates in further analyses.

Generality of Violence and Nonviolent Criminal Behavior

A one-way ANOVA was used to assess the impact of childhood trauma status on ratings of violent offenses against someone other than an intimate partner. There was a significant difference in non-intimate violence between groups, F (3, 1095) = 5.83,p = .001, with non-intimate violence increasing with level of exposure to violence as a child. The eta-squared was .016, which means that a little less than 2% of the variance in violence could be accounted for by group membership.

Post-hoc comparison of means tests (Dunnett T3s) revealed significant differences between those who had neither witnessed domestic violence nor were abused as children (“neither” group) and those who had both witnessed domestic violence and been abused as children (“both” group). Significant differences were also found between those who witnessed domestic violence only (“witnessed only” group) and those who had both witnessed and been abused (“both” group), with the participants in the “both” group having committed more non-intimate violence than any other group (see Fig. 1).

Fig. 1
figure 1

Mean ratings (0 to 2 scale) per group of violent offenses committed against someone other than an intimate partner

Full size image

A one-way ANOVA was also used to assess the impact of childhood trauma status on ratings of non-violent criminal offenses. It was expected that results would significantly mirror non-intimate violent behavior. This notion was not confirmed.

The differences between groups with respect to the potential for physically abusing children as measured by the CAP were also assessed with an ANOVA. There was a significant difference in potential for child abuse between the groups, as measured by this scale; F (3, 278) = 8.72, p < .001. The eta-squared was .071. Significant differences between the “neither” group and “abused only” group were revealed by post-hoc comparisons using Dunnett T3s. Significant differences were also found between the “neither” group and the “both” group. Participants in the group that experienced neither form of violence had lower CAP scores than any of the other groups.

Frequency and Severity of Domestic Violence

Differences in frequency of domestic violence offenses between groups were assessed using an ANOVA with the frequency score (sum of the modified physical CTS items) as the dependent variable. Significant differences were found in frequency scores between groups, F (3, 1094) = 26.90, p < .001; eta-squared was .069. Post-hoc analyses using Dunnett T3s revealed significant differences between those in the “neither” group and all of the others. Significant differences were also found between those in the “abused only” group and those in the “both” group. Comparison of means revealed that those who had neither witnessed domestic violence nor been abused as children committed less domestic violence, and witnessing domestic violence as a child added to the likelihood of committing domestic violence as an adult.

Differences in severity of domestic violence offenses between groups were assessed using an ANOVA with the severity score (derived from the subtotal of modified CTS items) as the dependent variable. There was a significant difference between groups, F (3, 1095) = 14.95, p < .001, with those with more violence exposure committing more severe violence. Eta-squared was .039. Post-hoc analyses revealed significant differences in severity of domestic violence offenses between those in the “neither” group and all other groups, with the “neither” group exhibiting the lowest severity scores. (For a summary of frequency and severity of domestic violence offenses, see Fig. 2.)

Fig. 2
figure 2

Mean severity of domestic violence score per group—sum of five CTS items rated on a 0–4 scale

Full size image

Psychopathology

Psychopathology and personality characteristics were assessed via five personality scales of the MCMI (Antisocial, Borderline, Dependent, Depressive, and Narcissistic). A MANOVA using the five aforementioned scales of the MCMI as dependent variables was conducted. There was a significant effect for status, F (3, 851) = 18.89, p < .001. Psychopathology increased as level of exposure to violence increased. One-way ANOVAs were then conducted with each MCMI scale as the dependent variable to specify differences (See Fig. 3).

Fig. 3
figure 3

Mean scores for each status group on the depressive, dependent, narcissistic, antisocial, and borderline scales of the Millon clinical multiaxial inventory, third edition

Full size image

Analysis of the Narcissistic scale analysis revealed no significant differences, but all other univariate analyses showed significant or near significant* differences in level of psychopathology between groups. The Depressive scale results were F (3, 852) = 11.84, p < .001 (eta-squared was .053). The Antisocial scale results were F (3, 852) = 21.00, p < .001 (eta-squared was .079). The Borderline scale results were F (3, 852) = 17.60, p < .001 (eta-squared was .070). The Dependent* scale results were F (3, 852) = 3.14, p < .05* (eta-squared was .019).

Post-hoc tests for the Depressive scale revealed significant differences between the “neither” and “witnessed only” groups. Significant differences were also found between the “neither” and “both” groups, and between those in the “abused” and “both” groups. Comparison of means indicated that the most depressed were those with the most exposure to violence as children and the least depressed were those with the least exposure to violence. It seems to be the case that witnessing added to the likelihood of being depressed, as it does for the likelihood of frequent and severe domestic violence.

Post-hoc tests for the Antisocial scale revealed significant differences between the “neither” and “witnessed” groups, the “neither” and “abused” groups, the “neither” and “both” groups, and the “abused” and “both” groups. This followed the same pattern as the other MCMI scales, with means being ordered from highest to lowest for “both”, “witnessed”, “abused”, and “neither”.

Post-hoc tests for the Borderline scale revealed significant differences between the “neither” and “witnessed” groups, the “neither” and “abused” groups, and the “neither” and “both” groups. Again, the same pattern was evidenced in the means, and the hypothesis about psychopathology was confirmed.

Post-hoc tests for the Dependent scale revealed significant differences between the “neither” and “both” groups. Interestingly, there was a non-significant difference between those in the “witnessed only” group and those in the “both” group, with the participants who witnessed only being more dependent than those who were exposed to both forms of violence.

Discussion

The purpose of this study was to investigate the relationship between childhood exposure to violence and characteristics of adult male domestic violence offenders, who were placed in groups according to their reports of whether or not they witnessed domestic violence or were physically abused as children. Participants completed measures to assess generality, frequency, and severity of their violent offenses, occurrence of other criminal behavior, and level of psychopathology. Although there is a plethora of research on batterers, no prior study had directly examined differences in men who had witnessed, been abused, neither had witnessed nor been abused, or both had witnessed and been abused.

The likelihood of committing violence against someone other than an intimate partner (general violence) increased as the participants’ exposure to violence as a child increased. Batterers who were abused as children were more likely to abuse children than those who were not abused. These results are consistent with previous findings that children who witness violence (e.g., Downs et al. 1996; Henning et al. 1997) or are physically abused (e.g., McCord 1983; Rosenbaum and Bennett 1986) often become aggressive adults. With respect to previous findings about male batterers, these results are in harmony with the research that states that the most generally violent men often report being exposed to violence as a child (Holtzworth-Munroe et al. 2000; Waltz et al. 2000). Results are also consistent with the previous finding that abused children are likely to become child abusers (e.g., Kempe et al. 1962; Widom 1989). And it adds support to the notion that adults who were abused as children may be more likely to abuse children than those who witnessed violence but were not themselves abused (Kalmuss 1984).

Nonviolent criminal behavior did not increase with exposure to violence as a child. This result seems contradictory to the research on the consequences of witnessing domestic violence and being physically abused as a child, which repeatedly reports that witnessing and being abused are associated with a greater number of legal problems and arrests (e.g., Graham-Bermann and Levendosky 1998; Widom and White 1997).

This non-significant difference between groups may be accounted for by the fact that any childhood exposure to violence is associated with criminal activity. It is the case that domestic violence witnesses and abused children are similar with respect to negative outcome (Jaffe et al. 1986). Perhaps, the men in the “neither” group experienced some other form of violence (e.g., neighborhood, media) and are affected in the same manner as men who experienced familial violence. Another potential reason for the lack of findings is that the arrest rate, which is what ratings were based upon, may have been inflated due to the racial mix of the population. This sample was largely (85%) African-American, and African-American males have a higher likelihood than the general population of being imprisoned or jailed. Approximately 5% of the general population will be in jail or prison during their lifetimes; this number jumps to 28% for African-American males (U.S. Department of Justice 2000b). This unfortunate statistic may have masked differences between groups given the high likelihood of arrest for nonviolent crime in the entire sample.

It may also be the case that battering men commit more nonviolent crime than the general population, making it a behavior that is prevalent for all the groups but not distinguishing between them. Base rates specifically for nonviolent crime are not readily available, but statistics on criminal corrections may help understand the lack of significance. Approximately .01% of the general population was on probation or parole in 1997 (U.S. Department of Justice 1996), but 40% of men arrested for domestic violence had criminal justice status (probation, parole, or restraining order) prior to arrest (U.S. Department of Justice 2000b). The failure to separate types of crime is not uncommon. In fact, most prior studies looking at criminal and legal activity with respect to childhood history and adult offense have not separated violent from nonviolent crimes, so it is unclear whether witnesses and abused children actually commit more nonviolent offenses as adults. This finding and explanation are consistent with a review by Malinosky-Rummell and Hansen (1993) that found no relationship between physical abuse and criminal behavior.

Another possibility for non-significant findings is that the ratings of nonviolent crime were open-ended. Police and legal records were obtained by the domestic violence center and the number of offenses was coded exactly for zero and one offense, but a rating of 2 was assigned for two or more offenses. More precision of the variable may have revealed differences between groups.

Frequency of domestic violence offenses committed increased as exposure to violence as a child increased. This finding is consistent with previous reports of the most frequent offenses being committed by men with the highest level of exposure to violence in childhood (Holtzworth-Munroe et al. 2000; Waltz et al. 2000). The finding that men who had witnessed domestic violence committed more frequent domestic violence than men who had not is consistent with previous research about the likelihood of perpetration of domestic violence (Downs et al. 1996; Kalmuss 1984). This adds support to the modeling theory, given that men who witnessed domestic violence committed that offense, which they had seen as children, more frequently than men who were abused only or had no exposure to either form of violence.

Severity of domestic violence offenses committed also increased as exposure to violence as a child increased. This is consistent with the finding that men who were exposed to the most violence as children commit the most severe domestic violence (for a review see Holtzworth-Munroe et al. 1997). Men who had both witnessed domestic violence and been abused committed the most severe offenses, which is also consistent with previous findings (Downs et al. 1996; Kalmuss 1984). There was a trend toward witnesses having committed more severe offenses than those who were abused only, but this difference was not significant.

Level of psychopathology increased as exposure to violence as a child increased. This is consistent with psychological difficulties reported by adults who were witnesses of domestic violence (e.g., Jaffe et al. 1986) or were abused as children (e.g., Kinard 1980). Previous research has shown childhood exposure to violence is related to personality disorders in adult male batterers (Hamberger and Hastings 1986; Waltz et al. 2000). However, most studies have not found significant differences between batterers that differed with respect to childhood history or other characteristics (Holtzworth-Munroe et al. 2000; Waltz et al. 2000). In contrast, this study found significant differences for three (Antisocial, Borderline, and Depressive) of the five personality disorders assessed.

Significant results may have been found due to the fact that this study, which categorized batterers according to their childhood exposure to violence, allowed for more detail to be uncovered than most prior research. One explanation for the lack of significant findings on the MCMI scale measuring narcissistic characteristics is that narcissism is a likely characteristic of all batterers. In fact, the mean score of 67, which is just below the threshold for clinical concern, on the Narcissistic scale was considerably higher than any other mean score.

Limitations

One limitation of this study was that there was no nonviolent comparison group. It would have been informative to have men who were exposed to violence as children but did not become violent adults. This would have been a more stringent test of modeling and could have revealed specific differences in characteristics between those who became violent and those who did not.

Another, somewhat related, limitation is that all of the participants had been arrested and court-ordered for assessment. It may be the case that the subset of battering men who “get caught” are different on important dimensions from those whose violence goes without punishment. Generalizability may also be questioned due to the uneven racial mix of the sample.

A third limitation of this study was that the independent variable, childhood exposure to violence, was assessed solely through retrospective self-report. It may have been difficult for adult men to accurately remember their levels of exposure to violence as children, and whether or not they answered questions truthfully could be questioned. It may be that some men, eager to blame their histories rather than accept responsibility for their violent behavior, over-reported childhood exposure to violence. Witnessing and being abused may also have been under-reported due to social stigma. This issue of self-report also affects the dependent measures. Victim report of frequency and severity of domestic violence offenses was available only for about one-third of the sample.

Some of the measures had weaknesses. For example, the ratings of prior nonviolent offenses were open-ended. Ratings of violence against someone other than an intimate partner might have been slightly contaminated as well as being open-ended and involving some subjectivity. Although the measure intended to exclude domestic violence offenses, data collected from the police for the first few months did not provide victim information, and therefore intimate partners may not have been removed from the database. This is unlikely, given that prior to DVAC involvement, very few arrests were made for domestic violence charges, but nonetheless it should be taken into consideration. Ratings were basically assigned based on number of offenses, but a 2 was assigned for two or more offenses or for one severe offense, and although DVAC staff members were trained and reliability was checked, there may have been some unintended variability in this measure. The CAP is not a very good assessment tool. It is the only instrument that is widely accepted as a measure of child abuse potential, but it is often not interpretable due to participants’ lying in response to its face valid items. Furthermore, the psychometric properties have been rarely studied by any one other than the test creator.

Recommendations for Future Research

Improvements can be made with respect to study population. A sample accurately reflecting national racial mix would increase generalizability. This could be done by sampling from a variety of areas or by using a stratified sample. Two comparison groups, one of nonviolent men who had exposure to violence as children and one of non-court-ordered domestic violence offenders, should be included in future tests of the relationship between childhood exposure to violence and characteristics of domestic violence offenders. A retrospective study following children who had exposure to violence would be ideal. If this is not possible, and adults are assessed, a solution would be to obtain medical or legal records, or parental reports, that confirm reported childhood exposure to violence or lack thereof.

More consistent measures from sources other than the participant to assess generality, frequency, and severity of violence and psychopathology would be beneficial. Generality of violence, particularly child abuse, should be assessed objectively. This could be done by using medical records or police records where the child is clearly stated as the victim. This study did not include victim reports and ratings made by the center staff based on police and court records, and those and other external sources should be used more extensively. Careful separation of nonviolent and violent offenses and separation of intimate versus non-intimate offenses are recommended.

To date, few studies have examined all of the dimensions on which batterers differ. Continued assessment of frequency, severity, and psychopathology is important, as more data are needed. This was the first study that looked at specific differences based on offender’s childhood exposure to violence, and significant findings both replicated and added to previous work. This distinction may be the first step toward identifying how batterers develop differentially, which has been identified as missing from the literature (Holtzworth-Munroe and Stuart 1994), and is clearly important for prevention and intervention efforts. Therefore, it is recommended that this distinction be made in future projects.

Results from this and future studies may clarify the sequelae of exposure to violence as a child, and this clarification should be used to work with children in order to stop the intergenerational transmission of violence. When the unfortunate need arises to develop intervention, treatment should be tailored to the history of the batterer. Perhaps the identification of the importance of childhood exposure to violence is key in the successful treatment of batterers, and thereby would slow the cycle of violence.