Generalists versus specialists: Toward a typology of batterers in prison

Batterer's criminal history has been traditionally linked to the existence of severe violence toward partner ( Bender & Roberts, 2007; Cavanaugh & Gelles, 2005; Holtzworth-Munroe & Stuart, 1994 ). The most severe type of violence in Holtzworth-Munroe and Stuart's (1994) typology belongs to the generally violent/antisocial batterer with a long criminal history, a profile that is also found by other researchers. In Gondolf's (1988) typology, Type I or sociopathic batterer is also violent outside the home and presents a longer criminal history than the antisocial batterer (Type II) and the typical batterer (Type III), who is similar to the family-only batterer in Holtzworth-Munroe and Stuart's typology. Conversely, in Hamberger, Lohr, Bonge, and Tonlin's (1996) study of 204 maritally violent men, when violence is directed exclusively toward their partners it used to be less frequent, less severe, and with no psychopathology associated (the non-pathological batterer). There is empirical evidence, however, that suggests that both criminal history and severe violence toward their partners, although related, might be relatively independent (see for instance Boyle, O’Leary, Rosenbaum, & Hasset-Walker, 2008 ). This is especially important in studies with batterers conducted in prison, where reports of more severe violence are expected. For instance, using the typology of Holtzworth-Munroe and Stuart, Walsh et al. (2010) found the same levels of physical violence for the generally violent/antisocial group (with longer criminal history) and the family-only violent group in a sample of civil-psychiatric patients. As it is discussed below, criminology literature on crime specialization provides theoretical arguments to anticipate that family-only batterers could be also involved in moderate to severe violence toward their partners.

Both Holtzworth-Munroe and Stuart's (1994) and Cavanaugh and Gelles’ (2005) classifications of batterers seem to suggest a direct relationship between psychopathology and partner violence: one would expect moderate to severe psychopathology in individuals with severe violence toward their partners. There is empirical evidence about the role of psychopathology in the etiology of partner violence. In a longitudinal study of 543 participants belonging to a community sample followed over 20 years, Ehrensaft, Cohen, and Johnson (2006) found that men most seriously abusive toward their female partners also showed both antisocial and dramatic, emotionally dysregulated personality features (see also, Holtzworth-Munroe, Meehan, Herron, Rehman, & Stuart, 2003; Moffitt, Robins, & Caspi, 2001 ) and that Cluster B symptoms (narcissistic, antisocial, histrionic, and borderline) were the only significant personality predictors of increased risk of injury to a partner. The Holtzworth-Munroe and Stuart's batterer typology, as well as others, emphasizes the role of psychopathology to differentiate between groups of batterers. In their follow up of batterers, however, Holtzworth-Munroe et al. (2003) found that the generally violent and dysphoric/borderline groups were almost indistinguishable (see also Delsol, Margolin, & John, 2003; Holtzworth-Munroe & Meehan, 2004 ) and that level of psychopathy in these groups were similar ( Holtzworth-Munroe, Meehan, Herron, Rehman, & Stuart, 2000; Huss, Covell, & Langhinrichsen-Rohling, 2006; Walsh et al., 2010 ). While the empirical evidence suggests that psychopathology could be on the onset of partner violence for some individuals, it might not allow to clearly distinguish between subgroups of batterers (dysphoric/borderline and generally violent/antisocial in Holtzworth-Munroe and Stuart's typology; moderate and high-risk batterer in Cavanaugh and Gelles’ typology).

In summary, the scientific literature on partner violence has provided empirical evidence about the heterogeneity of batterers and the most influential classifications of batterers distinguish between the less violent (family-only, low-risk offenders) and the more violent batterers (dysphoric/borderline and generally violent/antisocial; moderate and high-risk offenders). In these classifications, the batterer's severity of violence seems to be linked to the presence of an antisocial trajectory (i.e., criminal history) or severe psychopathology. In the case of imprisoned batterers, however, severity of violence tends to be present, leading to a lack of representativeness of the less violent batterer (family-only, low-risk offender). Also, as several researchers have pointed out, there seems to be an overlap between the more violent and psychologically distressed groups of batterers ( Delsol et al., 2003; Holtzworth-Munroe et al., 2000; Holtzworth-Munroe et al., 2003; Holtzworth-Munroe & Meehan, 2004; Huss, Covell, & Langhinrichsen-Rohling, 2006; Walsh et al., 2010 ). These two circumstances limit the potential utility of the classical typologies when applied to the study of imprisoned male batterers and suggest the need for a classification that takes into account the characteristics of this population.

In the last few years, scholars have debated whether intimate partner violence might be considered different from other types of crimes. This debate has two sides: those who maintain that partner violence is a unique type of crime and those who support the idea that partner violence could be empirically indistinct from general crime. The accumulated empirical evidence in this topic is not conclusive. For instance, in their study of 2,124 offenders from a nationally representative sample of inmates from state and federal facilities in the U.S., Felson and Lane (2010) did not find empirical evidence supporting the idea that offenders who attacked partners were different from other offenders and could be regarded as typical offenders. Moffitt, Krueger, Caspi, & Fagan (2000) showed how general crime and partner violence were two different, although correlated, conceptual constructs. Using data from a longitudinal follow up for more than 20 years of 800 young adults, they found that many batterers also engaged in violence against non-intimates but the etiology of both types of violence seemed to be different as indicated by the existence of different correlates for each type of violence. For instance, low self-control (Constraint) predicted crime but not partner violence. As Baker, Metcalfe, and Jennings (2013) have recently pointed out, the versatility/specialization debate is both theoretical and methodological, where theories of general tendencies of antisocial behavior ( Farrington, 2005; Gottfredson & Hirschi, 1990 ) predict greater versatility than theories that assume different delinquent trajectories ( Moffitt, 1993 ), or theories that underline the cultural ( Dobash & Dobash, 1992) or intrafamilial ( Gelles & Straus, 1979; Giles-Sims, 1983 ) origins of partner violence. The debate is also methodological in the sense that the new statistical methods tend to find greater levels of specialization (log-linear models, latent class analysis, quantile regression, etc.) than more traditional techniques such as factor analyses or simple comparisons of crime involvement. Also, the offender population under study and the type of source data used (official records vs. self-reports) might affect the degree of versatility/specialization found in the studies ( Bouffard, Wright, Mufti?, & Bouffard, 2008 ).

A generalist offender would commit different offenses on various occasions, with no inclination to pursue a specific criminal act or pattern of criminal acts whereas a specialist offender shows a greater tendency to repeat the same crime or offense over time ( Baker et al., 2013 ). In the literature, specialization is seen as a type of consistency through a criminal career, which, indeed, might be relatively versatile. For the present study, a specialist male batterer is an offender whose only type of offenses are related to violence against his partner whereas a generalist male batterer is an offender with a criminal history of various types of offenses, including violence against his partner. This characterization presents an important difference with previous studies focusing on the degree of specialization among batterers (see Moffitt et al., 2000 ) and links the specialist batterer with the classical family-only violent batterer in Holtzworth-Munroe and Stuart's (1994) typology. Unlike typologies such as Holtzworth-Munroe and Stuart's, and Cavanaugh and Gelles’ (2005) , where the family-only or low-risk offender are characterized by a low profile of violence (less severe and less frequent), in the present study the specialist batterer is not expected to use different levels of violence from the generalist batterer's. This would be in line with types of batterer such as the intimate terrorist in Johnson's (1995) typology (severe violence in the family whether unidirectional of mutual) or the Type II antisocial batterer in Gondolf's (1988) typology, where moderate to severe violence is expected inside the family but not outside the family.

The objective of the present study is to explore if there are substantial differences between these two types of imprisoned male batterers. To do so, the study followed Moffitt et al.’s (2000) suggestion when searching for differences among types of batterers. These authors suggested answering to two research questions: are these batterers the same or different people? Do they share the same or different correlates? The distribution of batterers across groups might help to answer the first question. For instance, if there were not a sizable number of members of any one group, the evidence for the existence of these two types of batterers would be weak. The analysis of the most common correlates of crime in general and partner violence in particular would help to answer the second question. If few or no statistically significant differences were found between the groups in a set of correlates, the evidence of true differences between the groups would be weak. To answer this second question the present study analyzes correlates in the individual, family, and community contexts that have been consistently linked to general crime and partner violence.

As individual characteristics, the study analyzes personality, sexist attitudes, and alcohol and substance dependence. There is compelling evidence that personality is linked to the onset of both criminal behavior ( Blonigen & Krueger, 2007) and partner violence ( Capaldi & Kim, 2007; Cavanaugh & Gelles, 2005 ), with a special incidence of Cluster B personality profile, which includes the narcissistic, antisocial, histrionic, and borderline personality ( Ehrensaft et al., 2006 ). As for the role of attitudes toward women in partner violence, the literature is consistent in pointing at sexist attitudes as an important correlate of partner violence. This influence of sexist attitudes on partner violence is shared by several theoretical approaches, from theories that underline the importance of gender power disparity in society (feminist theories) to theories that emphasize the influence of cultural and family values (ecological) on the justification of abusive behavior toward female partners (see Ali & Naylor, 2013, for a review of studies).

The presence of alcohol and substance dependence has been consistently reported as an important correlate of both general aggression and partner violence ( Wilkinson & Hamerschlag, 2005 , for a review of studies) both at the time of a violent incident or as a distal correlate. In their meta-analysis of studies on the link between alcohol abuse and partner violence, Foran and O’Leary (2008) found a small to moderate effect size for the association between men alcohol abuse and partner violence. This relationship was greater in clinical samples and when severe alcohol problems were measured.

Family of origin has been proposed as a source of poor parenting skills, antisocial modeling, socioeconomic deprivation, and low attachment between the child and the parents that might be linked to violence in general ( Farrington, 2003; LeBlanc, 2005; Thornberry, 2005 ) and partner violence in particular ( Ehrensaft et al., 2003; Lussier, Farrington, & Moffitt, 2009 ). In the present study, family of origin climate and functioning represent the family context.

As for the community context, there are sound theoretical arguments as well as a vast array of empirical evidence on the ecology of partner violence ( Lauritsen & Schaum, 2004; Pinchevsky & Wright, 2012; Van Wyk, Benson, Fox, & De Maris, 2003 ). According to Pinchevsky and Wright (2012) , disadvantaged communities might influence partner violence in several ways. For instance, these communities may facilitate alienation and foster social isolation among citizens, which in turn influences the transmission of mainstream values that disapprove violence within couples. They also may hinder the formation and maintenance of social ties, leading to an increased vulnerability of residents. Also, they may intensify stress among couples, thus increasing the likelihood of partner violence. Thus, communities characterized by social disorder, and low levels of community integration and participation are expected to foster partner violence among its citizens. Community social disorder, community integration, and participation are variables representing the community context in the present study.

Finally, research has shown that offenders do not distribute homogeneously across sociodemographic characteristics. Thus, age, education, social class, and marital status as well as the delinquent trajectory of the offenders (age of onset of criminal behavior and age of first entry in the prison system) have been related to delinquent behavior and partner violence ( Mazerolle & Maahs, 2000; Sabina, 2013; Van Wyk et al., 2003 ). In order to explore the profile of generalist and specialist offenders, we also included their sociodemographic characteristics.

Results show that all of the MANOVAs conducted were statistically significant, suggesting that generalists and specialists batterers scored differently on the individual, family and community variables (see Table 4 ). Looking at the partial effect as depicted by η 2 , the effects were greater in alcohol and substance dependence, community and family and Cluster B personality variables.

In the present study a typology of imprisoned male batterer is proposed: the generalist vs. the specialist batterer. Drawing from both the literature on crime specialization and the typologies of male batterers, the criminal history of 110 imprisoned male batterers was used to distinguish between the generalist and specialist batterer. The group of generalist batterers was formed by inmates with a criminal history of various types of offenses, including violence against his partner. The group of specialist batterers was formed by those inmates whose offenses were always related to violence against his partner. Although both types of batterers have characteristics in common with other types of batterers described in the literature, they also present important differences. The generalist batterer is closer to the generally violent/antisocial batterer in Holtzworth-Munroe and Stuart's (1994) typology and the high-risk offender in Cavanaugh and Gelles’ (2005) classification. The specialist batterer has not a long criminal history, in line with the family-only and low-risk batterers of Holtzworth-Munroe and Stuart's and Cavanaugh and Gelles’ classifications. They do not necessarily present, however, the low-violence profile of these typologies. In this sense, they would be closer to the intimate terrorist in Johnson's (1995) typology or the Type II antisocial batterer in Gondolf's (1988) typology where moderate to severe violence seems to be limited to the family environment.

Two research questions regarding these groups guided the present research: 1) are they the same people? 2) do they share the same correlates? As for the first question, the distribution of batterers across groups suggests that although most offenders were in the generalist group ( n = 86, 75%) there were also a sizable number of batterers with no other criminal record or with a criminal history consistently related to violence toward his partner ( n = 24, 25%). This finding seems to give support to the idea that although crime generalization is the most typical profile of offenders, there is also a certain degree of specialists among them ( Baker et al., 2013; DeLisi et al., 2011; Felson & Lane, 2010 ).

Next, we explored the sociodemographic, individual, family, and community characteristics for each group. The generalist batterers not only present a longer and more varied criminal history but also an earlier onset in his criminal activity. Most generalist batterers had committed offenses related to drug use and drug traffic (crimes against public health, 74%) and robbery (74%), but there were also murderers (17%) in this group. Specialist batterers presented a shorter criminal history and specifically related to partner violence, mainly for violation of no-contact orders. These results suggest that, indeed, there is a different criminal trajectory in each group: the generalist batterer is younger and with an earlier onset in his criminal activities, which by definition is more varied. This finding would be consistent with research showing that the age of delinquency onset is inversely related to the number of offenses ( Mazerolle, Burton, Cullen, Evans, & Payne, 2000; McGloin, Sullivan, Piquero, & Pratt, 2007; Piquero, Paternoster, Mazerolle, Brame, & Dean, 1999 ) and that younger and persistent offenders tend to have a more varied criminal history ( Arce, Fariña, & Vázquez, 2011 ), including partner violence ( Mazerolle & Maahs, 2000).

As for the differences in their individual characteristics, the specialist batterer seems to present a lower profile of psychopathology than the generalist batterer, specifically in Cluster B personality aspects such as antisocial and borderline personality. This would be in line with the family-only batterer ( Holtzworth-Munroe & Stuart, 1994) or the low-risk offender (Cavanaugh & Gelles, 2005 ) proposed in the literature. Unlike these types of batterers, however, the specialist batterer also presents moderate to severe (even lethal) violence. In fact, there were two murderers among this group of specialist batterers, which challenges the classical notion that the family-only batterer usually presents a low level of partner violence. Specialist and generalist batterers showed no significant differences in other Cluster B characteristics such as histrionic or narcissistic personality that have been related to partner violence ( Ehrensaft et al., 2006; García-Jiménez, Godoy-Fernández, Llor-Esteban, & Ruiz-Hernández, 2014; Torres, Lemos-Giráldez, & Herrero, 2013 ), and batterer treatment efficacy ( Novo, Fariña, Seijo, & Arce, 2012 ).

Generalist batterers showed greater levels of substance dependence than the specialist batterers and similar levels of alcohol dependence. The fact that most of generalist batterers of the study had been convicted for crimes related to drug traffic (around 75% of generalist batterers), which tend to be linked to drug use and abuse, might explain this finding. Also, alcohol and substance dependence has been regarded to partner violence ( Wilkinson & Hamerschlag, 2005 ). Both alcohol and substance dependence may exert their influence at the social, economic, and relational level, thereby increasing the stress in the relationship and the likelihood of partner violence.

Regarding their sexist attitudes, the generalist batterer showed greater levels of hostile sexism but similar levels of benevolent sexism when compared to the specialist batterers. Research on ambivalent sexism has shown that those who are high in hostile sexism are more tolerant of intimate partner violence ( Glick, Sakalli-Ugurlu, Ferreira, & Souza, 2002 ) while benevolent sexism have been related to victim-blaming attitudes ( Viki & Abrams, 2002 ) toward women. These attitudes serve as an anchorage that guides the information interpretation, supporting preconceptions against women ( Fariña, Arce, & Novo, 2002 ), which contributes to sustain offending ( Maruna, 2004).

As for their family of origin correlates, family functioning was different in each group. Specialist batterers scored significantly higher in conflict in family of origin, indicating that they portrayed their family of origin as a context were family members more openly expressed anger and conflict than in the case of generalist batterers. This finding would be consistent both with theories that emphasize the intrafamilial origins of partner violence ( Gelles, 2007 ) and with the empirical evidence linking exposure to family violence and partner violence in adult life. Longitudinal research has found that exposure to violence between parents is a consistent predictor of partner violence in adult life ( Ehrensaft et al., 2003; Lavoie et al., 2002; Simons, Lin, & Gordon, 1998 ). In this sense, Lussier et al. (2009) have found evidence supporting that family environment increases the risk of partner violence, mainly by fostering the development of antisocial behavior and neuropsychological deficits (see also Capaldi & Clark, 1998 ). This would be in line with the antisocial/generally violent batterer in Holtzworth-Munroe and Stuart's (1994) and the high-risk offender in Cavanaugh and Gelles’ (2005) typologies. Alternatively, the existence of moderate to severe violence toward female partner in the group of specialist batterers would illustrate the role that conflictive family environments might have on the development of patterns of aggression in intimate relationships, with no further need of development of antisocial behavior. This would explain the existence of types such as Johnson's (1995) intimate terrorist or Gondolf's (1999) Type II antisocial batterer, where moderate to severe violence is expected inside the family but not outside the family.

Besides this debate about the direct or indirect influence of family functioning on partner violence (through antisocial behavior and neuropsychological deficits), our findings indicate that conflicts in the family of origin might be a key influence on partner violence in the case of the specialist batterer. No differences were found regarding the more positive aspects of family functioning such as cohesion, expressiveness, adaptability, partnership, growth, affection, and resolve. The fact that these types of batterers shared most of the family correlates is compatible with the idea shared by many scholars that both the more general antisocial trajectory and the specialization in partner violence are related to the existence of dysfunctional families of origin ( Ehrensaft et al., 2003; Farrington, 2003; Gelles, 2007; LeBlanc, 2005; Lussier et al., 2009; Thornberry, 2005 ).

Finally, it seemed that these two groups came from different communities or residential areas: the generalist batterer described his community as more socially disordered and showed lower levels of integration and participation toward it. As Pinchevsky and Wright (2012) concluded in their extensive review of studies about the impact of neighborhoods on intimate partner violence, disordered neighborhoods have been consistently linked to higher risk of partner violence while those characterized by the existence of support ties are more protected from such violence (see also Gracia & Herrero, 2007; Gracia, Herrero, Lila & Fuente, 2009; Herrero & Gracia, 2005 ). Our findings seem to support this claim, especially in the case of the generalist batterer with a more varied criminal history. The lower levels of participation and integration in a community and the higher levels of community social disorder exemplify a well-studied path between neighborhood conditions and both general and partner violence ( Lauritsen & Schaum, 2004; Markowitz, Bellair, Liska, & Liu, 2001; Van Wyk et al., 2003 ). While generalist batterers in our study seem to fit well into this explanation, our findings suggest that community correlates were not so important in the case of the specialist batterer. Thus, the specialist batterer lived under community conditions of lower social disorder and higher integration and participation with apparently less influence on partner violence. Other research has found levels of partner violence under low disordered neighborhood conditions. For instance, Van Wyk et al. (2003) found that intimate partner violence may increase its likelihood even at levels of low social disorganization if certain circumstances are present (i.e., lack of contacts for women).

Overall, the study findings help to portray a generalist batterer, which shows higher levels of psychopathology, substance dependence, and sexist attitudes. Also, their communities seem to be more socially disordered and their levels of community participation and integration are lower. They also present a longer and more varied criminal history with an earlier onset, both in terms of the age of first arrest as well as the age of first imprisonment. The specialist batterer, who represents a minority of the batterers analyzed in the present research, presents a profile of lower psychopathology, substance dependence, sexism, and community social disorder and a profile of higher levels of conflict in the family of origin along with higher levels of community integration and participation.

The study presents several potential limitations, however. First, most information regarding the criminal history of inmates was collected through official records. Previous research has found that specialization of offenders is more evident when using official records instead of self-reported information, although with some exceptions (see Bouffard et al., 2008 for a review of studies). If this were the case, it would be possible that some of the specialist batterers in our study had committed other than partner violence-related offenses that were not officially reported. The significant differences found in the criminal history of specialist and generalist batterers seem to suggest, however, that there are true differences among them (i.e., later onset of criminal activity in specialist batterers). With the present data we cannot rule completely out other alternative explanations based on the existence of undetected criminal activity in participants of the study. Further research using both self-reported as well as officially reported information should clarify this point.

Second, participants of the study might not be representative of the convicted batterer population, so generalization of results is not warranted. Again, the clear differences in both the onset of criminal activity and the age of first imprisonment seems to suggest that these two groups exist in the population of imprisoned male batterers although we should be cautious about the distribution of batterers across groups. In our study, three out of every four batterers belonged to the generalist group, in contrast with Holtzworth-Munroe and Stuart (1994) suggestion that both generalist/antisocial and dysphoric/borderline batterers would account for almost 50% of the population of batterers. Further research with representative samples should provide more accurate estimates of the true distribution of generalist and specialist batterers.

And last, but not least, the definition of specialist batterers used in this study might be too restricted, potentially leading to low group stability across time. In this sense, the specialist batterer could belong to the generalist group if a non-partner violence-related offense is committed. It seems clear, however, that for most specialist offenders their criminal history is limited to this type of offense, as seen by their late onset of criminal activity and age at first imprisonment. Future research focusing on the trajectory of batterers could add relevant information about the formation of these two groups of batterers.

The authors of this article declare no conflict of interest.

Support for this research was provided by grant from the Spanish Ministry of Economy and Competitiveness (FEM2012-30659).