Introduction

Violence against women is a major public health problem in the United States, with approximately one quarter of American women reporting they have been a victim of intimate partner violence (IPV) at some point during their lifetime (Tjaden and Thoennes 2000). While awareness of the magnitude of this problem has grown, treatment programs for men who engage in partner violence have not kept pace, and researchers conducting reviews of the treatment outcome literature have concluded that these programs have limited effectiveness (Babcock et al. 2004; Davis and Taylor 1999). This reality has led to increased efforts to determine whether distinct subgroups of partner violent men could be identified in order to inform the development of more effective interventions designed to address the unique characteristics of subgroups of partner violent men.

Interest in the development of typologies of partner violent men was heightened by an influential review on the topic by Holtzworth-Munroe and Stuart (1994). However, efforts to replicate their proposed typology have met with mixed success (Hamberger et al. 1996; Holtzworth-Munroe et al. 2000, 2003; Waltz et al. 2000). In addition, some researchers have found that complex typologies utilizing personality disorder profiles derived from standardized personality testing are difficult for clinicians to employ reliably (Langhinrichsen-Rohling et al. 2000; cf., Lohr et al. 2005). The present paper examines a more easily applied method of distinguishing between subgroups of partner violent men, and is based on a theoretically important behavioral distinction (i.e., the generality of violence engaged in). The generality or specificity of aggression engaged in by men was one of the criteria in the Holtzworth-Munroe and Stuart (1994) typology, but it also had two other criteria (i.e., severity of partner violence and psychopathology). Categorizing batterers according to the specificity/generality criterion is quite easy to do, and could be heuristically and practically valuable if it proves to be useful in distinguishing between groups.

Researchers from diverse disciplines and perspectives have for the most part studied IPV as a phenomenon distinct from general violence (Moffitt et al. 2000). This distinction is supported by etiological theories of IPV, which suggest that there are different determinants of partner violence and general violence. Sociological (e.g., Gelles 1993) and feminist (e.g., Dobash and Dobash 1979; Yllo 1993) theorists argue that the unique configuration of families and patriarchal society make IPV a distinct phenomenon from violence against non-intimate victims. Similarly, cognitive-behavioral researchers have studied IPV as distinct from other violent behavior. For example, they have examined the impact of unique partner-specific and dyadic processes on IPV, such as partner-specific anger and hostility (Boyle and Vivian 1996), spouse-specific assertiveness (O’Leary and Curley 1986; Rosenbaum and O’Leary 1981), and relationship satisfaction (Rosenbaum and O’Leary 1981; Vivian and Langhinrichsen-Rohling 1994).

Furthermore, recent longitudinal research conducted by Moffitt and colleagues (2000) suggests that generally and partner-only violent (PO) men have distinct personality profiles. These researchers found that individuals who engaged in acts of general violent criminal conduct were characterized by the personality trait of weak constraint (a construct similar to that of impulsivity and behavioral disinhibition), while PO individuals were not so characterized (Moffitt et al. 2000). Criminologists have also found differences between IPV and more general violence (e.g., Avakame 1998; Parker 1989; Parker and Toth 1990; Stout 1992). From a societal and legal perspective, too, IPV and violence against third parties have been regarded as distinct, and have received differential treatment from the criminal justice system (Englander 1997; Micklow 1988). Thus, there may be distinct characteristics of generally and PO men, as the context and consequences of each behavior are different.

The Present Study

The present investigation was designed to build on prior work examining the distinctions between generally (GV) (i.e., violent both with intimate and non-intimate victims) and PO (i.e., violent only with intimate partners) violent men (Cadsky and Crawford 1988; Shields et al. 1988). We endeavored to fill gaps in prior research by: (1) using multiple measures of constructs of lifetime violence and antisocial behavior; (2) including measures of behavioral disinhibition and impulsivity; and (3) using the same response format to measure general and partner specific violence. This last point addresses a critique by Moffitt and colleagues (2000) of prior research on the distinction between partner violence and general violence in which researchers used different types of measurements of the two behaviors (e.g., using official crime reports to measure general violence, while using self-report to measure partner violence).

We hypothesized that GV and PO violent men would be distinguished from each other based on: (1) lifetime antisocial and violent behavior, (2) exposure to violence in their families of origin, (3) impulsivity and behavioral disinhibition, and (4) severity of psychological and physical abuse of their female partners.

Researchers have found that violent and antisocial behaviors are highly stable across the lifespan (Eron 1997; Huesmann et al. 1984; Lahey et al. 2005; Olweus 1979). Given these findings, we hypothesized the GV men would have both engaged in more violence and antisocial behavior during their lifetime, and been exposed to more violence in their families of origin, than PO violent men (i.e., the first two research hypotheses). Meta-analytic research (Stith et al. 2000) and longitudinal research with a community sample (Ehrensaft et al. 2003) have found that witnessing IPV in the family of origin, as well as being the victim of abusive discipline, are associated with later engaging in IPV. These experiences have important treatment implications from a social-cognitive perspective since the longer the duration and range of aggressive behavior engaged in, the less likely a given individual will be able to change cognitive schema and aggressive behavior patterns (Eron 1997).

Thirdly, we hypothesized that GV men would report greater behavioral disinhibition and impulsivity than PO violent men. This hypothesis is supported by a longitudinal study in which Moffitt and colleagues (2000) found that weak constraint (e.g., impulsivity and behavioral disinhibition) was uniquely related to general violence but not IPV. Weak constraint is one of three personality factors (the other two factors are positive and negative affectivity, respectively) identified by Tellegen (1982).

Similarly, Clark (2005) identifies three broad, innate temperament dimensions (i.e., disinhibition, positive affectivity, and negative affectivity) in her model of personality and psychopathology. Temperament is typically seen as a more biologically based set of predispositions that contribute to personality development (Nigg 2000). In Clark’s conceptualization, disinhibition is a broader temperament that includes the trait of impulsivity. Disinhibition refers to the tendency to behave in an inadequately controlled manner. A disinhibited individual pursues stimulating experiences and acts on current thoughts or feelings with little regard for the consequences of his/her actions (Clark 1993, 2005).

Disinhibition and impulsivity are two closely related constructs. Impulsive individuals have been described as overly sensitive to rewards, deficient in their ability to inhibit responses despite potential punishment, and likely to engage in novelty and sensation-seeking experiences (Moffitt 1993). Impulsivity has been associated with both general violence (Lane and Cherek 2000; Plutchik and Van Praag 1995) and with IPV (Stuart 1998; Stuart and Holtzworth-Munroe 2005). Finally, Huss and Langhinrichsen-Rohling (2000) have suggested that a distinction between psychopathic batterers and others be made, with impulsivity being a defining characteristic of the psychopath.

Fourth and finally, we hypothesized that GV men would have engaged in more psychological and physical abuse of their intimate partners. That is, we believed that GV men would not only be violent toward a wider range of targets, but that they would also be more psychologically and physically abusive toward their partners. We based this belief on the previously stated hypotheses, that is, that GV men would have been exposed to more violence, have more extensive histories of violent and antisocial behavior, and that they would be more behaviorally disinhibited and impulsive than PO violent men.

Method

Participants

A sample of 73 men entering a University-based treatment center for partner violence served as participants for this study. All of the men approached (N = 100) agreed to take the packet and to participate. Twenty-seven ultimately did not return/complete the packet, which left 73 study participants. This project had IRB approval from the two Universities that were involved in this study. Informed consent was obtained in advance from all participants.

Participants were classified into one of two groups based on each man’s self-report of violence during the past year on the General Violent Tactics Scale (GVTS; Boyle, General violent tactics scale, unpublished instrument), an adapted version of the Revised Conflict Tactics Scale (RCTS; Straus, Hamby, Boney-McCoy and Sugarman 1996). The GVTS is described further in the Measures section, later in this paper. An individual was classified as GV (n = 46) if he reported engaging in any 1 of 12 violent acts (frequency of acts range from once to greater than 20 times) against someone other than an intimate partner in the past year. Alternatively, individuals were classified as PO (n = 27) if they reported no violence against individuals other than intimate partners in the past year.

Chi-square analyses were conducted to examine differences on education and race/ethnicity. As is shown in Table 1, no significant differences were found between the GV and PO subgroups in race/ethnicity or educational attainment.

Table 1 Sample demographic characteristics
Full size table

Next, separate t-tests with group as the independent variable were used to examine group differences on age and income; no significant differences were found in either. Mean age was 30.0 years (SD: 7.2) for GV men; and 32.0 years (SD: 8.9) for PO violent men. Mean income was $20,930 (SD: $14,853) for GV men; and $23,437 (SD: $15,768) for PO violent men.

Procedure

During his initial intake appointment at the Men’s Violence Treatment Program, we offered each individual the option of participating in the study in exchange for a $30 incentive payment. Participants were required to mail the completed questionnaire to the investigator (who was affiliated with another university) before the first session of their treatment program in order to be included in the study.

Measures

Revised Conflict Tactics Scale (Straus et al. 1996)

The Revised Conflict Tactics Scale (RCTS) is a revised and expanded version of the Conflict Tactics Scale (Straus 1979).It consists of 78-items, and five subscales designed to assess conflict in intimate relationships. Four of the subscales measure the frequency (on a seven-point scale: 0 = never, 1 = once, 2 = twice, 3 = 3–5 times, 4 = 6–10 times, 5 = 11–20 times, and 6 = more than 20 times) with which an individual has engaged in negotiation, psychological aggression, sexual coercion, and physical violence during relationship conflict. For the present study, participants were asked to indicate how often they engaged in such behaviors over the course of their adult life in the context of intimate relationships. Participants were not asked about their partners’ behavior.

In the present study, indices of lifetime psychological aggression and physical violence were derived by summing the endorsed frequency for each subscale. In keeping with prior research (e.g., Straus 1990), responses indicating a range of frequencies were recoded to the midpoint (i.e., 3–5 = 4 times, 6–10 = 8 times, 11–20 = 15 times, and 20 or more = 25 times). Straus (1990) reported good internal consistency of the RCTS scales, ranging from 0.79 to 0.95. In the present study, alpha for the lifetime reports on the overall RCTS was 0.91.

General Violent Tactics Scale (Boyle, General violent tactics scale, unpublished instrument)

The General Violent Tactics Scale (GVTS) is an adapted version of the RCTS. It consists of physical violence items from the RCTS. However, the wording was changed to inquire about use of these tactics during conflict with people other than intimate partners (e.g., strangers, acquaintances, friends, co-workers) over the past year. As described earlier in the Participants section, responses to the GVTS were used to form the GV and PO subgroups (i.e., participants reporting any act of general violence in the past year were classified as GV and the remainder were classified as PO). In the present study, alpha for the GVTS was 0.91.

Lifetime History of Aggression (Coccaro et al. 1997)

The Lifetime History of Aggression (LHA) is a ten-item scale with three subscales measuring aggression (five items, e.g., verbal aggression, physical aggression against other people and objects), antisocial behavior (four items, e.g., illegal activity, problems with supervisors), and self-directed aggression (one item, i.e., attempts to physically hurt oneself). Respondents were asked to indicate on a 6-point Likert scale how often they engaged in these behaviors over the course of their lifetimes, ranging from “never happened” to “happened so many times I couldn’t give a number.” Alpha for the overall scale was 0.88. Alphas for the aggression and antisocial behavior subscales were 0.87, and 0.74, respectively. We present data on both the overall scale as well as the antisocial behavior subscale, since we were interested in separating out purely antisocial behavior from aggressive behavior. Additionally, because we formed the groups based on the general aggression engaged in during the past year, we were concerned that comparing the groups based on the LHA aggression subscale might be tautological.

Adapted Family Violence Questionnaire (Malone et al. 1989)

The Adapted Family Violence Questionnaire (AFVQ) assesses violence in the respondent’s family of origin. Respondents were asked to rate on a 5-point Likert scale (1 = Never to 5 = Very often) how often various acts of verbal or physical abuse occurred when they were growing up. The AFVQ consists of 32 items, including violence by mother and father toward respondent; violence between parents; violence by respondent toward parents and toward siblings; as well as violence by siblings toward respondent. Each subscale consists of five items with the exception of violence between parents, which consists of seven. In the present study, alpha for the AFVQ was 0.88. Alpha for the individual subscales ranged from 0.82 to 0.92. We present the overall scale, and for exploratory purposes the individual subscales.

Conduct Disorder and Delinquency Scale (Boyle, Conduct disorder and delinquency scale, unpublished instrument)

The Conduct Disorder and Delinquency Scale (CDDS) consists of 23 items, and two subscales (i.e., Conduct Disorder and Delinquency). The conduct disorder subscale is modeled on the criteria outlined in the Diagnostic and Statistical Manual of Mental Disorders (4th ed.; American Psychiatric Association 1994). To that end, the conduct disorder subscale (15 items) assesses aggression (e.g., bullied and intimidated others), destruction of property (e.g., deliberately destroyed the property of others), theft (e.g., broke into someone else’s house or car), and rule violation (e.g., skipped school). The delinquency subscale (8 items) assesses delinquent behaviors (e.g., suspended or expelled from school; smoked marijuana; hit a school teacher) measured in prior research (Malamuth et al. 1995). Respondents were asked to indicate how often they engaged in these behaviors during childhood and as a teenager, on a 5-point Likert scale (0 = Never to 4 = Very often). Alpha for the overall scale was 0.90. Alphas for the conduct disorder and delinquency subscales were 0.86 and 0.80, respectively.

Barratt Impulsiveness Scale (Barratt 1994)

The Barratt Impulsiveness Scale (BIS-11) contains three subscales, that is Motor (e.g., I do things without thinking); Attention (e.g., I have “racing” thoughts); and Non-Planning Impulsiveness (e.g., I plan tasks carefully; which is reverse coded). Each subscale consists of 10 items. Respondents were asked to rate on a 4-point Likert scale, ranging from rarely/never to almost always/always, the degree to which the statements describe their behavior and thoughts. Higher scores are an indication of greater impulsivity. In the present study, alpha for the BIS-11 was 0.82. The BIS-11, as the name suggests, is the 11th version of the original BIS which first appeared in published research in 1959. Over the course of decades of research, the developer has tested it with a wide range of respondents, ranging from college students to individuals incarcerated for violent crimes (Barratt 1994).

Schedule for Nonadaptive and Adaptive Personality, Disinhibition Subscale (Clark 1993)

The Schedule for Nonadaptive and Adaptive Personality (SNAP) is a factor analytically developed self-report instrument designed to assess trait dimensions in the domain of personality disorders. The SNAP contains 12 trait scales and three temperament scales; as well as 13 diagnostic scales assessing personality disorders. The 16 item Disinhibition Temperament scale [SNAP, Disinhibition Subscale (DIS)] was employed in the present study. Clark’s model of personality has three broad, innate temperament dimensions (i.e., disinhibition, positive affectivity, and negative affectivity) (Clark 2005). Disinhibited individuals tend to behave in an inadequately controlled manner. They pursue stimulating experiences, and act on current thoughts or feelings, with little regard for the consequences of their actions (Clark 1993; 2005). Respondents were asked to indicate whether the statements (e.g., when I am having a good time, I don’t worry about the consequences; I get a kick out of really scaring people) were true/mostly true or false/mostly false for them. In the present study, alpha for the SNAP-DIS was 0.91. Higher scores reflect greater behavioral disinhibition.

Results

Initially, we were interested in examining the pattern of intercorrelations among the independent and dependent measures. As displayed in Table 2, there were significant correlations among IPV (RCTS), general aggressiveness (LHA), and conduct disorder/delinquency (CDDS), with general aggressiveness and conduct disorder/delinquency being the most strongly correlated. In addition, as would be expected, measures of similar constructs – that is, impulsivity (BIS-11) and behavioral disinhibition (SNAP-DIS) – were strongly related.

Table 2 Correlations among dependent and independent variables (N = 73)
Full size table

Next, we proceeded to examine group differences on the overall scales of the various dependent variables (i.e., LHA, AFVQ, CDDS, SNAP-DIS, and the BIS-11) using MANOVA, and found significant differences between GV and PO violent men (Wilks Lambda = 0.82, p = 0.02, \( \eta ^{2}_{p} = .18 \)). Since the overall MANOVA was significant, we followed it up with separate two-tailed t-testsFootnote 1 for the individual research hypotheses.

We tested the first research hypothesis (i.e., GV men would have engaged in more lifetime antisocial and violent behavior than PO violent men) with group comparisons on the CDDS, and the LHA total scale and antisocial subscale. As hypothesized (see Table 3), the GV men reported engaging in more behaviors associated with conduct disorder and delinquency on the CDDS. GV men also reported more lifetime violent and antisocial behaviors on the LHA Total Scale and Antisocial Behavior subscale.

Table 3 Means, standard deviations, and group differences on life-course antisocial behavior
Full size table

The results in Table 4 lend support to the second research hypothesis. That is, GV men reported greater exposure to family of origin violence than PO violent men on the AFVQ total scale. Since the overall AFVQ revealed significant differences between the two groups, we conducted a series of exploratory t-tests on the subscales that compose the overall scale. When the individual subscales were examined, GV men reported more violence directed by them toward their parents and siblings; and more violence directed toward them by their mothers.

Table 4 Means, standard deviations, and group differences on violence in the family of origin (AFVQ sum and subscales)
Full size table

Contrary to expectations, however, there were no differences between GV and PO violent men in terms of violence directed toward them by their fathers, or in terms of inter-parental violence witnessed. One possible explanation is that some participants may have been raised in single parent households. In fact, 35% of respondents reported that their fathers (or step-fathers) were rarely or very rarely present during their childhood. Nevertheless, father’s presence, as measured on a five-point Likert scale, was not related to reports on the AFVQ of fathers’ abuse, or inter-parental violence.

As seen in Table 5, the GV men had a higher mean Disinhibition score (SNAP-DIS) than the PO violent men. However, the two groups did not differ significantly on mean impulsivity scores (BIS-11), although the differences approached significance (p ≤ 0.09). Had a one-tailed test been used, these differences would have reached the p ≤ 0.05 significance level. These findings lend some support to the third research hypothesis (i.e., that GV men would report more disinhibition than PO violent men).

Table 5 Means, standard deviations, and group differences on impulsivity and behavioral disinhibition
Full size table

Finally, we found some support for the fourth research hypothesis. That is, GV men reported engaging in more psychologically abusive behaviors toward their intimate partners than PO violent men as measured by the RCTS subscale (GV: mean 80.6, SD 43.8; PO: mean 52.2, SD 31.6; T = 3.2, p = 0.00). The groups did not, however, differ on the RCTS physical violence subscale (GV: mean 30.8, SD 34.1; PO: mean 22.9, SD 29.2; T = 1.0, p = 0.32).

Discussion

Overall, it appears that GV men (when compared to the PO group) are characterized by a lifelong pattern of violence against others and antisocial behavior. This pattern is seen in greater violence directed by GV men against others in their families of origin, as well as more conduct disordered and delinquent behavior in late childhood and adolescence. This pattern is continued during adulthood, with greater antisocial behavior, general violence, and partner psychological abuse (though not more self-reported partner violence). GV men were also characterized by greater behavioral disinhibition, though group differences on impulsivity only approached significance.

This study represents an improvement over past studies by using multiple measures of lifetime violence and anti-social behavior, including behavior during childhood, adolescence, and adulthood. The study also adds to the literature regarding the relationship among behavioral disinhibition, impulsivity, and violence. Finally, employing the same response format for generally and partner-specific violence addresses a critique by Moffitt and colleagues (2000) of prior partner-violence research.

Our findings regarding early childhood experiences of witnessing and being the victim of family violence, engaging in conduct disordered and delinquent behaviors, and later perpetrating IPV are consistent with prior research. For example, Ehrensaft and colleagues (2003) found that conduct disordered behavior during adolescence mediated the relationship between early experiences in the home of witnessing IPV and being the victim of child abuse, and later perpetrating IPV as an adult. This finding is consistent with our result of subgroup differences between GV and PO men, since the GV subgroup reported not only more conduct disorder and delinquency, but also more violence in their families of origin.

The GV men were also more likely to be behaviorally disinhibited, while group differences on impulsivity approached statistical significance. This may suggest that interventions tailored to the underlying temperament of GV men should be incorporated with traditional treatment. Some researchers (Huss and Langhinrichsen-Rohling 2000) have suggested that a distinction between psychopathic batterers and others be made, with impulsivity being a primary defining characteristic of the psychopath. The present study lends some support to the idea that there is a subgroup of batterers with psychopathic tendencies (i.e., greater impulsivity/behavioral disinhibition, and more antisocial behavior). This subgroup difference may have emerged more strongly if a higher threshold for inclusion in the GV subgroup was employed. That is, perhaps if severity of general violence engaged in was factored into the decision on subgroup classification, more distinct subgroups might have been derived. This is a topic that deserves additional future research.

Regarding the general composition of the sample in terms of the proportion of participants who were GV, it is difficult to discern how our sample compares to samples in prior, published studies on the criterion of general versus partner specific violence. Often, other researchers’ descriptions of how their groups were derived are not complete. Also, some studies have formed multiple subgroups, and drawn from a different population (e.g., general population rather than men referred to batterer programs). Nevertheless, some comparisons are possible.

Shields and colleagues (1988) formed three subgroups of men (referred from social service agencies), two of which are equivalent to the PO and GV subgroups; a third subgroup of men were violent only with nonintimate victims. The relative proportion of GV to PO men in that sample was roughly the same as in the present study (i.e., two-thirds GV to one third PO). Some additional prior research has distinguished among batterers referred for treatment. Cadsky and Crawford (1988) reported that roughly 40% of their batterer sample was GV, while six out of ten were PO. However, this sample is not directly comparable because the authors excluded from the study men who reported assaulting others more frequently than their female partners. Unfortunately, Cadsky and Crawford do not state how many men were excluded on this basis. In other words, the relative percentages of GV and PO violent men will vary according to the criteria employed to define the subgroups, and the population from which they are drawn.

Limitations

The main limitation of the present study is that it employs retrospective reports. Thus, it is possible that individuals justified their own present-day behavior by reporting more negative early life experiences. However, some of the findings indicate that this may not have been the case. For example, GV men might have been expected to justify their IPV by reporting greater harm inflicted on them by their fathers, or greater inter-parental violence; however, this was not seen with the present sample.

A second limitation is that we only have men’s reports. No reports were gathered from the female victims. Researchers have consistently found that men tend to underreport IPV (O’Leary and Murphy 1992). The tendency to report in a socially desirable direction (e.g., report less IPV against their partners) may explain why our two subgroups did not differ on the measure of IPV.

Prevention and Treatment Implications

Our findings regarding family-of-origin violence, conduct disorder, and delinquent behavior suggest the importance of prevention programs for at-risk populations. Specifically, secondary prevention programs with at-risk children (e.g., children referred to Child Protective Services; children of women seeking assistance at domestic violence shelters; families where police have responded to complaints of domestic violence) should be employed. The findings also support the importance of interventions for adolescents with conduct disorder, and for those who engage in delinquent acts, as there is an association between such behaviors and later engaging in both IPV and general violence. Interventions targeted toward child and adolescent populations may help to prevent future violence against partners and others.

The results of the present study also suggest treatment implications for batterer programs. Since GV men have violent tendencies that are long-standing, they may be less optimal candidates for a rehabilitative treatment option. That is, given the range and duration of their violent behavior, as well as greater behavioral disinhibition, GV men’s violent behavior patterns may be most resistant to change. This finding may also provide support for the use of more traditional criminal justice sanctions with GV men.

This subgroup comparison employed a fairly simple technique of dividing men into groups based on their own reports of violence over the past year, thereby producing subgroups that differed on a number of important characteristics that may have implications for treatment. An advantage of this technique is that it would be relatively easy for other treatment programs to apply. In contrast, some proposed typologies (e.g., Holtzworth-Munroe and Stuart 1994) have proven difficult to employ reliably. The method described in the present study is simpler, more reliable, and can be used by clinicians without formal training in standardized personality testing.

In conclusion, despite the limitations, the present study contributes to the literature by filling existing gaps, including the use of multiple measures of lifetime violence constructs. The findings suggest that PO violent men (slightly more than one-third of the sample) may be better candidates for traditional IPV programs. The fact that roughly two thirds of this sample were GV, and had long histories of violent and antisocial behavior, may help to explain why researchers evaluating outcomes of IPV treatment programs have found these programs to have limited effectiveness.