Are Generalist Batterers Different from Generally Extra-Family Violent Men? A Study among Imprisoned Male Violent Offenders

Intimate Partner Violence (IPV) against women is a widespread problem that affects all societies and cultures. In Europe, estimates are that 22% of women have suffered physical and/or sexual partner violence since the age of 15 (European Union Agency for Fundamental Rights - FRA, 2014) and most of these IPV cases occurred while being in an intimate relationship. More recently, the World Health Organization (WHO) (2016)has indicated that 30% of women worldwide have suffered physical and/or sexual IPV. As for psychological IPV, more than 30% of women have experienced this form of violence by her current or a previous partner and 43% of women suffered psychological behaviors like controlling, economic violence and blackmail (European Union Agency for Fundamental Rights - FRA, 2014).

While IPV against women is a rather complex topic of study that needs probably to be addressed from multiple views, the last decades have witnessed an increasing interest in the study of the characteristics of aggressors. The study of typologies of IPV against women is a good example of this approach (Capaldi & Kim 2007;Logan, Walker, & Leukefeld 2001). Thus, many studies have examined different types of male perpetrators of intimate partner violence against women (Cavanaugh & Gelles 2005;Fowler & Westen 2011;Holtzworth-Munroe & Stuart, 1994;Ross & Babcock, 2009;Walsh et al. 2010) being the typology of male batterers established byHoltzworth -Munroe and Stuart (1994; see also Holtzworth-Munroe & Meehan, 2004;Holtzworth-Munroe, Meehan, Herron, Rehman, & Stuart 2003) one of the most influential. These authors studied male batterers on three dimensions (severity of marital violence, generality of the violence, and psychopathology or personality disorder) and the typology that they proposed comprised three types of batterers: (a) family only batterer, (b) dysphoric or borderline batterer, and (c) generally violent or antisocial batterer (see alsoDutton, 1988;Saunders, 1992). More recently,Cavanough and Gelles (2005) andFowler and Westen (2011)proposed two typologies with similar characteristics to that ofHoltzworth-Munroe and Stuart (1994 ).Cavanough and Gelles (2005) used three dimensions - (a) severity and frequency of the violence,

(b) criminal history, and (c) psychopathology level - to identify three types of male batterers: (a) low risk offenders, (b) medium risk offenders, and (c) high risk offenders. In the same vein,Fowler and Westen (2011) differentiated three subtypes of batterers: (a) subtype 1 or psychopathic, (b) subtype 2 or hostile/controlling, and (c) subtype 3 or borderline/dependent.

Batterers in Prison

Overall, research on male batterers have suggested that in most cases of severe IPV against women male aggressors were also generally violent beyond the family realm (Cavanough & Gelles, 2005;Fowler & Westen, 2011;Gondolf, 1988;Hamberger, Lohn, Bonge, & Tonlin, 1996;Harris, Hilton, & Rice, 2011;Holtzworth-Munroe & Stuart, 1994). AsHerrero, Torres, Fernández-Suárez, & Rodríguez-Díaz (2016) have recently suggested, however, the fact that severe IPV against women is overrepresented in penitentiary-based samples of male batterers might lead to a lack of representativeness of the less violent batterer. To overcome this limitation, they proposed a typology that incorporates male batterers? criminal history and identified two types of convicted male batterers: generalist and specialist batterers. Generalist batterers had a long and varied criminal history that also d IPV against women among other offenses, while specialist batterers had exclusively offenses related with IPV against women. In their study,Herrero, Torres et al. (2016)linked their typology to the debate about the etiology of IPV and general violence.

On the one hand, several researchers (Dutton, 1988;Felson, 2006;Felson & Lane, 2010;Moffit, Krueger, Caspi, & Fagan, 2000) have suggested that IPV might be explained by general theories of violence and aggression. For instance, IPV perpetrators and other offenders with criminal behavior non-related to IPV seem to be similar (Felson & Lane, 2010;Herrero, Torres, Rodríguez, & Juarros- Basterretxea, 2017;Kiss, Schraider, Hossain, Watts, & Zimmerman, 2015). In fact, heterogeneous and long criminal history has been found in IPV batterers (Cavanough & Gelles, 2005;Fowler & Westen, 2011;Holtzworth-Munroe & Stuart, 1994;Piquero, Theobald, & Farrington, 2014) and both types of offense seem to be related to early antisocial behavior (seeHilton, Harris, & Rice, 2001). Typologies typically label this intimate partner batterer who commit other illegal acts as generally violent or antisocial batterer (Logan et al., 2001; see alsoHoltzworth-Munroe & Stuart, 1994;Shield, McCall, & Hanneke, 1988). These generally violent batterers or generalist batterers represent a sizable group of the total of imprisoned batterers (Herrero, Torres et al., 2016), in accordance with theHoltzworth-Munroe and Stuart?s (1994)suggestion that generally violent or antisocial batterers will be mainly present in prison.

On the other hand, there is also research empirically supporting the gender perspective on IPV against women.Thomas, Dichter, and Matejkowski (2011) observed differences between intimate partner murder offenders and non-intimate partner murder offenders. Specifically, intimate partner murder offenders were more socially integrated and used fatal violence to satisfy emotional aspects in contrast to non-intimate partner murderers whose violence seemed to be rather instrumental. Furthermore, they seemed to differ in marital status, history of severe mental illness, and motives (Thomas et al., 2011). Likewise,Swogger, Walsh, and Kosson (2007) compared antisocial batterers with other antisocial criminals that did not perpetrate IPV against women and found a different profile in affective experience, impulsivity, and irresponsibility.

The Current Research

Drawing both from the debate on the etiology of general violence and male IPV against women (seeBouffard & Zedaker, 2016;Felson, 2006;Felson & Lane 2010;Hilton & Eke 2016;Moffit et al., 2000) as well as from recent findings on batterers typologies in prison (Herrero, Torres et al., 2016), the present study sought to answer the following research question: Do male batterers with other criminal offenses (generalist batterers) differ from other violent offenders without a criminal record of IPV against women? We hypothesize that if the IPV perpetrated by GB and general violence have a common etiology, GB and GEVM will not differ in individual, familiar, and community variables.

Using data of convicted male offenders from the Penitentiary Center of Villabona (Asturias, Spain), the objectives of the present study were twofold: a) to explore the distribution of generalist batterers against women and common violent criminals without an IPV criminal record and b) to their individual, family of origin, and community characteristics, seeing violence as a multidimensional phenomenon with multiples causes at different levels (seeBronfenbrenner, 1977;WHO 2002).

As for the individual characteristics of offenders we d personality, sexist attitudes, and alcohol and drug dependence. Personality has been linked to both general criminal behavior (Davison & Janca, 2012;Ehrensaft, Cohen, & Johnson, 2006) and IPV against women (Capaldi & Kim, 2007;Mauricio, Tein, & López, 2007). As for the influence of sexist attitudes toward women, there is empirical evidence suggesting that they are an important risk factor to understand IPV against women (Ali & Naylor, 2013). Finally, different studies that substance and alcohol use and abuse are important factors related with general violence and IPV (Fagan, 1990;Feingold, Kerr, & Capaldi, 2008;Golinelli, Longshore, & Wenzel, 2009; Kretschmar & Flannery, 2007;Lisco, Parrot, & Tharp, 2012;Reingle, Jennings, Connel, Businelle, & Chartier, 2014;Rothman, McNaughton, Johnson, & LaValley, 2012;Smith, Homish, Leonard, & Cornelius, 2012).

Family of origin characteristics have also been regarded to both IPV against women (Ehrensaft et al., 2003) as well as to general violence (Fagan, 2005;Farrington, 2003;LeBlanc, 2005). Several studies of IPV focused their attention on family of origin of batterers, taking in ac parental skills, educational models, socioeconomic level, or presence of abuse from parents (Herrero, Torres et al., 2016). In the present study family of origin variables were family climate and family ing.

Empirical evidence suggests that contextual factors also influence IPV against women (Pinchevsky & Wright, 2012). Thus, disadvantaged communities may influence IPV (Pinchevsky & Wright, 2012;Van Wyk, Benson, Fox, & DeMaris, 2003) but also general violence (O´Brien & Sampson, 2015;Markowitz, Bellair, Liska, & Liu, 2001). In this sense, community social disorder, community participation and community integration were community context variables in the present study. These are characteristics of disadvantaged communities (Herrero, Torres et al., 2016) which might be related with IPV against women and violence in general.

To better control for potentially biased responses in sensitive topics such as IPV (Bell & Naugle, 2007;Caetano, Schafer, Field, & Nelson, 2002;Henning, Jones, & Holdford, 2005;Saunders, 1991;Sugarman & Hotaling, 1997;Van de Mortel, 2008) we d a measure of social desirability in the study.

Participants

Participants of the study were 153 men imprisoned in the Penitentiary of Villabona (Asturias, Spain). Using information from criminal and justice official records, two types of imprisoned men were identified. The first group consisted of generalist batterers (GB), defined as those men who commit IPV against women as well as other varied criminal history (seeHerrero, Torres et al., 2016). The second group consisted of those imprisoned violent offenders with a varied criminal history that not d IPV against women (Geneally Extra-family Violent Men - GEVM). GB were significantly older (M = 36.48, SD = 9.39) than GEVM (M = 32.21, SD = 9.35) (F = 7.586, p = .007, h²= .049). Marital status for both groups was mainly single (48.80% and 65.10% respectively) or divorced/separated (30.20% and 19% respectively). Minor differences were found in the marital status of the two groups, but it lacked statistical significance (χ²= 4.001, p = .135). Most participants perceived themselves of middle class (GB = 69.77%, GEVM = 65.67%) (ns), and had primary (GB = 50%, GEVM = 55.22%) or secondary (GB = 34.88%, GEVM = 25.37%) studies at the most (ns).

Procedure

The researchers approached the governmental and penitentiary authorities and explained the study objectives in order to obtain permission to evaluate inmates on a set of variables. After official access was ed, the researchers identified voluntary participants and obtained informed consent from them to take part in the study. Different sources of information were used for this study. Specifically, judicial and penitentiary reports were used to obtain information about the criminal history of participants. This d information about socio-demographic variables as well as the complete record of arrests and imprisonments for each participant. Also, different self-report measures were used to evaluate the personal, family, and community contexts of participants.

Variables

Outcome variable. Based on criminal trajectory (see below), participants were divided into two groups: generalist batterers (GB; 56.2%) and generally extra-familiar violent men (GEVM; 43.8%). GB were characterized by a versatile criminal history, which d IPV against women, while GEVM had a history of criminal violence which not d IPV against women.

Criminal trajectory variables. The most frequent offenses committed by GB as well as GEVM were crimes against property (81% and 73.3% respectively). Both groups showed similar rates of other different offenses such as homicide, injuries, crimes against liberty and against privacy, among others. Furthermore, information about age of first non-sanctioned offense, age of first arrest and age of first imprisonment was retrieved from interviews and official records. GB and GEVM participants showed a similar onset of criminal behavior as can be seen both by the age of first non-sanctioned offense (GB, M = 13.53, SD = 6.78; GEVM, M = 13.93, SD = 6.56; F = .134, p = .715), the age of first arrest (GB, M = 19.21, SD = 6.56; GEVM, M = 18.32, SD = 5.98; F = 545, p =.461) as well as the age of first imprisonment (GB, M = 27.70, SD = 9.70; GEVM, M = 25.10, SD = 8.09; F = 3.005, p =.085). Length of criminal trajectory was measured as the time lapse between age of first non-sanctioned offense and actual age (i.e., subtracting age of first offense from actual age), being GB?s criminal trajectory longest than GEVM?s (GB, M = 17.08, SD = 9.91; GEVM, M = 13.89, SD = 8.55; F = 4.201, p = .042).

Biased responses. As response bias consisting in the adoption of positive characteristics (faking good) to portray themselves in a favorable light to get prison benefits and/or parole must be suspected in prison evaluation setting (Arce, Fariña, Seijo, & Novo, 2015), a psychometric measurement of social desirability was obtained from participants to control its effects as a covariate. As for this, psychometric approach is valid and preferable than a clinical approach (Arce, Fariña, & Vilariño, 2015). The Social Desirability Scale (α = .80) of the Spanish adaptation of the Millon Clinical Multiaxial Inventory-III (MCMI-III) (Cardenal & Sánchez, 2007;Millon, 1997;Millon, Davis, & Millon, 2007) was used to control biased responses.

Intimate relationship variable. The Revised Conflict Tactics Scale?s (CTS-2) Psychological Aggression Subscale (α = .88) (Straus, Hamby, Boney-McCoy, & Sugarman, 1996) was used to measure levels of psychological IPV against women in participants.

Individual Variables

Personality. The Spanish adaptation of MCMI-III was used to detect personality disorders (Axis II) (Cardenal & Sánchez 2007;Millon, 1997;Millon et al., 2007). Specifically, Histrionic (17 items, α = .80), Narcissistic (24 items, α = .70), Antisocial (17 items, α = .76), and Borderline (16 items, α = .82) scales were used in this study. Means and standard deviations for each scale and study group are presented inTable 1.

Substances use and abuse. Alcohol (15 items, α = .71) and Substance Dependence (14 items, α = .80) scales from the Spanish adaptation of Millon Clinical Multiaxial Inventory-III (MCMI-III) were used. Means and standard deviations for each scale and group are presented inTable 1.

Sexism. The Ambivalent Sexism Inventory (ASI) (Glick & Fiske, 1996, 1999 ) was used to evaluate sexist attitudes. It s 22 items and measures both Hostile (α = .85) and Benevolent Sexism (α = .82). All of the items were rated on a five-point scales ranging from strongly disagree to strongly agree. Means and standard deviations for each scale and group are presented inTable 1.

Family of Origin Variables

Family climate. The Family Relationship Index (Moos & Moos, 1994) was used to measure the quality of social relationships in the family environment. It measures cohesion (α = .91), expressiveness (α = .81) and conflict (α = .83). Means and Standard Deviations for each group are presented inTable 2.

Table 1

Means, Standard Deviations, and Group Differences on Individual Variables

Family ing. The APGAR scale was used to assess participants? perception of family ing (Smilkstein, 1978). It is a five 3-point items scale of family ing (Adaptability, Partnership, Growth, Affection, and Resolve) with responses ranging from 1 (hardly ever) to 3 (almost always). In the present study, a summed-up scale score was used. Internal consistency was high (Cronbach?s α = .92). Means and standard deviations for the scale in each group of batterers are presented in Table 2.

Table 2

Means, Standard Deviations, and Group Differences on Family Variables

Community variables

Community social disorder. Three items about the frequency of crime (fight, weapons, robbery, etc.), presence of drug traffic, and nightlife in the community (see Gracia & Herrero, 2006; Herrero & Gracia, 2005) were used to measure community social disorder. Item responses raged on five-point scale from disagree strongly to agree strongly. Internal consistency was adequate (α = .85). Means and standard deviations in each group are presented in Table 3.

Community integration and participation. The Community Integration and Community Participation Scales (α = .82 and α = .81 respectively) of the Perceived Community Support Questionnaire (Herrero & Gracia, 2007) were used to measure sense of belonging and identification as well as participation in social activities in the Community. Category responses ranged from 1 - strongly agree - to 5 - strongly disagree. Both scales have shown adequate psychometric characteristics and predictive validity (Herrero & Gracia, 2007; Herrero, Torres et al., 2016). Means and standard deviations in each group are presented in Table 3.

Table 3

Means, Standard Deviations, and Group Differences on Community Variables

Data Analyses

We first conducted two-step cluster analysis of The Psychological Aggression Scale scores. This technique is appropriate to find the number of groups that better ac for the sample distribution of participants across the scores of Psychological Aggression. We also conducted multivariate analyses of variance (MANOVA) to test for differences in criminal history between the two groups: age of first nonsanctioned offense, age of first arrest, and age of first imprisonment.

Finally, we conducted separate multivariate analyses of covariance (MANCOVA) to test for differences across groups in individual, family, and community dependent variables. Social desirability was d as a covariate in all the models tested. Dependent variables were grouped in a theoretically meaningful way: personality (cluster B), substance abuse, sexism, family of origin, and community. Table 5 presents a summary of MANCOVA results.

Results of bi-step cluster analysis showed that according to their scores on Psychological Aggression participants belonged to any of two groups (see Table 4): low (M = 4.49, SD = 4.61) and high psychological aggressio (M = 24.47, SD = 7.69). As expected, most of participants in the High Psychological Aggression Group belonged to the group of GB (68.8%), suggesting that GB participants were involved in IPV against women to a greater extent than GEVM (31.1%). This relationship was statistically significant according to the chisquared test (c2 = 4.334, p = .037, j = -.17).

Table 4

Pearson?s Chi-square Test between Study Groups and Clusters

Results of MANOVAs carried out to the differences in the criminal trajectory of participants showed that GBs and GEVM did not differ

5 in age of first non-sanctioned crime, age of first arrest, nor age of first imprisonment (Wilk?s Lambdα = .975, F3, 142 =1.198, p = .313, h2 = .025).

Results from MANCOVAs did not statistical significance, suggesting that GB and GEVM scored similarly on the individual, family, and community dependent variables (see Table 5).

Table 5

Summary of MANCOVA?s Means on Individual, Family of Origin, and Community Characteristics

In the present study, a comparison between two groups of violent inmates was carried out. Participants were 153 imprisoned men from the Penitentiary Center of Villabona (Asturias, Spain) belonging to two groups: the first group was comprised by generalist batterers who had heterogeneous and long violent criminal history which d IPV against women (generalist batterers -GB) (see Herrero, Torres et al., 2016), and the second group comprised generally extrafamily violent men (GEVM), who had heterogeneous and long violent criminal history which did not IPV against women. GB had common characteristics and similarities with other types of batterers proposed in other typologies: generally violent or antisocial batterer (Holztworth-Munroe & Stuart, 1994), subtype I or psychopathic partner-violent men (Fowler & Westen, 2011), high-risk offenders (Cavanough & Gelles, 2005), or type II antisocial batterer (Gondolf, 1988) among others. Our research sought to both the distribution of batterers among generally violent inmate population and their potentially different individual, family, and community characteristics (see Moffit & Caspi, 1999).

Regarding to the first objective, GB and GEVM group sizes were similar (n = 86 vs. n = 67), indicating that more than a half of generally violent inmates of the study had a versatile criminal history which comprised IPV against women as well as other violent crimes (see also Bouffard & Zedaker, 2016). These results give support to the idea that general violence and IPV might have a common etiology (Felson, 2006; Felson & Lane, 2010; Herrero, Torres et al., 2016; Moffit et al., 2000).

Next, we explored the criminal history as well as sociodemographic, individual, family of origin, and community variables for each group. GB and GEVM participants showed a criminal trajectory of violence (the most frequent crimes in both groups were crimes against property, more than 73%). We could not find any statistical differences between the two groups in their criminal trajectory, including age of first offense, age of first arrest, and age of first imprisonment. Taking into ac the length of criminal trajectory, results showed that criminal trajectory of GB was longer than GEVM?s. In this sense, other research has suggested that for some generally violent men criminal trajectory starts victimizing non-intimates, but progressively s family member (like partner) among their victims (Shield et al., 1988), and it seems that long criminal trajectories increased the odds of criminal violence as well as IPV (Piquero et al., 2014).

As for the individual characteristics, GB and GEVM did not differ in personality characteristics, nor substance and alcohol dependence, nor ambivalent sexism. Antisocial, histrionic, borderline, and narcissist personality have been related with violence in general, IPV against women (Cavanaugh & Gelles, 2005; Fowler & Westen, 2011; Herrero, Torres et al., 2016; Holtzworth-Munroe & Stuart, 1994; Kiwi & Sadeghi, 2015; Peters, Derefinko, & Lynam, 2016; Ruíz-Hernández, García-Jiménez, Llor-Esteban, & Godoy-Fernández, 2015; Stefánsson & Hesse, 2008; Theobald, Farrington, Coid, & Piquero, 2015) and reoffending (Kiwi & Sadeghi, 2015). According to this, our results suggest that personality characteristics do not allow to differentiate between types of criminal offenders: GB and GEVM participants showed a similar personality profile.

Further, drug and/or alcohol abuse and dependence have been traditionally linked to general violence (Fagan, 1990; Kretschmar & Flannery, 2007) and IPV against women (see Gondolf, 1988; Herrero, Torres et al., 2016; Shield et al., 1988; Saunders, 1992), which is consistent with our finding that both GB and GEVM participants showed similar levels of alcohol and substance abuse and dependence.

We also found in our study that levels of sexist attitudes toward women were similar across groups of offenders. Although this finding might challenge, research showing that more sexist, misogynistic, traditional, and gender rigid attitudes are related to IPV against women (Allen, Swan, & Raghavan, 2009; Flood & Pease, 2009), empirical research on male batterers have consistently found that antisocial/generally violent batterers the highest levels of hostility toward women (Herrero, Torres et al., 2016; Holtwortz-Munroe & Stewart, 1994). Also, evidence from general population studies indicates that more violent individuals might also hold hostile attitudes toward women (Herrero, Rodríguez-Díaz, & Torres, 2016) as compared to less violent individuals.

The same tendency emerged when GB and GEVM participants were compared in their family characteristics: we could not find any significant statistical difference in family ing and family climate (cohesion, expressiveness, and conflict) across groups. These results are in line with empirical evidence showing that the family environment is related to general violence (Fagan, 2013; Flannery, Singer, Van Dulmen, Kretschmar, & Belliston, 2007; Láng & Birkás, 2014; Meldrum, Connolly, Flexon, & Guerette, 2015; Vanassche, Sodermans, Matthijs, & Swicegood, 2013). Our findings further suggest that dyss in family of origin might be unrelated to IPV against women beyond its influence on general violence. We cannot conclude, however, that family dyss are unrelated to IPV against women as there is research showing, for instance, that specialist batterers (i.e., only involved in IPV-related offenses) tend to greater levels of conflict in their family of origin than generally violent or generalists batterers (Herrero, Torres et al., 2016).

Finally, regarding community variables, GB and GEVM participants showed the same levels of community social disorder, and community integration and participation. Herrero, Torres et al. (2016) have recently showed in their study of generalist and specialist batterers that those with a more versatile criminal history refer highest levels of social disorder and lowest levels of integration and participation in their communities. In this sense, results from our study indicated that disadvantaged communities and community integration and participation are related to general violence but do not seem to distinguish among those violent offenders who also commit IPV against women.

Strengths and Limitations

The current study present strengths and potential limitations. Among the strengths are the characteristics of participants and the control of potential response bias. As for the participants, violent offenses and/or IPV against women were identified from information provided by official records, so we might be confident about the presence of general violence and IPV against women in each participant. This circumstance alleviates the problem of identification of both violent individuals and batterers through the use of self-report measures that might probably be affected of recall bias. As compared to self-reported information about criminal history, official records have the advantage of objectivity and lack of social desirability and dishonesty of inmates. When self-report measures were used in the study, however, additional statistical controls were undertaken to obtain more accurate estimates. Thus, the use of the social desirability measure in the statistical analyses might have helped to identified true statistical relationships in our study.

Participants in the study were not representative of the inmate population, so we should be cautious about generalization of results. For instance, the true distribution of GB and GEVM participants in the population is unknown, so we are unsure whether the existence of generally violent batterers is rather unusual among the inmate population or not. Given that the existence of this group of batterers is key to understand the relationship between general violence and IPV against women, future research with representative samples should be undertaken to ascertain if the size of this group is relevant in the population. Available empirical evidence, however, suggest that a sizable proportion of male batterers in prison committed also non-IPV violent offenses (Felson & Lane, 2010;Herrero, Torres et al., 2016), so the classification used in this study seems tenable.

Related to this, the classification of participants with regard to the types of committed offenses might have affected group stability. At first, categorization of inmates as GB or GEVM was based on official records, allowing that participants with not sanctioned IPV offenses were misclassified in the GEVM group. To overcome this potential problem, final categorization was done including self-reported IPV perpetrations disclosed during the interview. Despite the fact that few inmates admitted that they occasionally committed acts of violence that can be categorized as IPV, other inmates might have concealed similar behaviors to the interviewer. In this sense, the limitation of the existence of undetected IPV was mitigated, but not completely ruled out.