Jump to Body Text

Characteristics of Batterers in a Multi-site Evaluation of Batterer Intervention Systems

Edward W. Gondolf, EdD, MPH
Mid-Atlantic Addiction Training Institute, 1098 Oakland Avenue, Indiana University of Pennsylvania, Indiana, PA 15705,Phone: 412-357-4749: Fax: 412-357-3944

Preliminary Report

Acknowledgments

Many individuals and agencies have contributed to this first phase of our multi-site evaluation of batterer intervention. The program directors at the four research sites offered essential assistance in developing and implementing the data collection: Robert Foster, Toby Myers, Sherry Lundberg, and Robert Gallup. Several research assistants; especially Jennifer Daly, Jody Scruggs, Don Bozich, and Paul Burchfield; helped to administer questionnaires and test materials at the sites, along with assistance from staff at the respective sites. Jeff Coben, Ron Laporte, and Ed Ricci of the University of Pittsburgh have served as co-investigators. Jewel Lee Doherty, project director, Crystal Deemer, administrative assistant, Neil Fulton, data manager, and several data entry workers, graduate assistants, and student staff of the Mid-Atlantic Addiction Training Institute provided a variety of support, supervision, and service. The Texas Council on Family Violence, the Pennsylvania Coalition Against Domestic Violence, the Mid-Atlantic Addiction Training Institute, Graduate School of Indiana University of Pennsylvania, and Center for Injury Research and Control at the University of Pittsburgh Medical Center, all contributed to the development of the research proposal and coordinating its implementation. The research was made possible through a grant from the Centers for Disease Control and Prevention, U.S. Department of Health and Human Services (Grant number R49/CCR310525-02).

Executive Summary

Th characteristics of batterers in court-referred programs have significant implications for program development and evaluation research. Background and test data, including information from the MAST and MCMI-III, were systematically collected from four geographically distributed batterer programs as part of our multi-site evaluation of batterer intervention. Cross-tabulations using chi square were used to describe the characteristics of the program enrollees overall and across the sites.

The men demographically appeared similar to previous portrayals of court-referred batterers, except that our sample had a greater portion of African American and Latino men than previous studies. According to formalized screening tests, over half of the men were apparently "alcoholic," and over a quarter had severe mental problems (major clinical syndromes [Axis I disorders] or severe personality pathology [Axis II disorders]). No distinct personality profile, however, emerged. Self-reports for heavy drinking and for depression produce similar prevalence rates as formalized testing instruments; however, they do not successfully identify alcoholism or depressive disorders in individual cases.

Several dichotomies appear among the men that might influence intervention, such as differences in living arrangements (over half are not living with their partners and nearly half are), education (a quarter without a high school diploma and a third with some college), employment (a third under employed versus two-thirds fully employed), and drinking (nearly a third seldom used alcohol versus a half who reported heavy drinking). Over half of the men in our sample have been previously arrested.

As practitioners assert, the men appear to grossly under report their abusive behavior with only about 40% acknowledging their recent assaults, and less than a fifth admitting to using severe tactics in the last three months. Over 40% of the partners responded by contacting the criminal justice system sometime in the past, but only about an eighth of the women had previously contacted a women's shelter or a counseling service about their being abused.

The four sites reflected regional differences in demographics, but had relatively similar portions of men with mental and drinking problems despite these differences. One site did have significantly more men reporting alcohol-related fights and drug use (20%) and another had more men involved in alcohol or mental health treatment. The men at one site were, moreover, significantly more likely to report recent assaults, but less likely to report assaults in the past. This difference may reflect differences in levels of disclosure related to a minor difference in the data collection procedures. The men's report of injury (a fifth of the women sought medical help in the past) and the women's response to the abuse were similar for all four sites.

Additional research will examine possible types or categories of batterers, discrepancies between the men's report of abuse and the women's reports, and the effect of batterer characteristics on program outcome.

Introduction

Issues:The characteristics of men in batterer programs are important to the domestic violence field for three main reasons. One, many of the speculations about appropriate intervention and treatment for batterers rests on assumptions about their characteristics and behavior ( Tolman & Bennett, 1993). Two, the characteristics of men in a program contribute to program outcome and therefore need to be considered in evaluating program effectiveness. Three, batterer characteristics of a particularly program assists in determining how representative a sample of men are of other programs. One program may appear more effective than another simply because it has different "kinds" of men in it.

Several generalizations have emerged about batterers' characteristics. On one hand, there are profiles that describe the typical or predominant batterer in programs. These point to younger men in their late twenties to early thirties, who are under employed laborers, often with drinking problems, prior arrest records, and personality problems ( Eisikovits & Edleson, 1989; Tolman & Bennett, 1990). On the other hand, there is an increasing interest in typologies that propose a few major categories of batterers ( Gondolf, 1988; Saunders, 1992). A recent summary of typology research suggests three categories: one of men who are highly dominating, and narcissistic; another of men who tend to be impulsive, dependent on their partners and who are highly jealous, controlling and reactive; and a third category of men who are generally antisocial, negativistic, and defiant and who are generally violent and sadistic ( Holtzworth-Munroe & Stuart, 1994).

These findings are drawn primarily from studies of single program sites that often employ different measures, information, and data collection procedures. In order to verify and elaborate these previous findings, a more systematic study of batterer characteristics is needed. The ideal would be data collected from a variety of program sites with geographical representation and using consistent measures and procedures.

Research Questions:The attention to batterer characteristics poses several fundamental questions for researchers and practitioners alike to examine:

  1. what are the prevailing demographic, personality, and behavioral characteristics of batterers in programs?

  2. what types, categories, or diversities among the men emerge from these characteristics?

  3. how do the characteristics vary among programs from different regions, court systems, and intake procedures?

We might speculate, as well, on the implications that the answers to these questions have for program development and future research.

We explored these questions using data derived from our multi-site evaluation of batterer interventions systems funded by the Centers for Disease Control (CDC). The data is drawn from men enrolling in court-referred batterer programs in four different cities nationwide. The data collection measures and procedures, as described below, were consistent across the four sites enabling us to approach the ideal for the representation and comparison of several programs. We first summarize the batterer characteristics in our total sample combining all four sites (n=840). The demographics, alcohol use and mental health, and abusive behavior (controls, threats, and assaults) are described using the indicators derived from a background questionnaire, personality test, and alcohol screening test. The significant differences among the four program sites are noted in a second section. A concluding section attempts to highlight the major findings, suggest some program implications, and discuss possibilities for further analyses using our data base.

Method

Multi-Site Study

Research Sites:The database from our evaluation study offers the most systematic and consistent data on batterers to date and is particularly suited for addressing the prevailing questions about the characteristics of men referred to batterer programs. The first phase of the study included recruiting and testing 200-215 subjects at four different urban sites for the Northeast, Central South, and West of the United States. The sites represent model programs ranging in their program duration (3 to 9 months) and extent of services (basic batterer education to victim assistance). Site 1 is the shortest in duration and extent of services, and Site 2 is the longest and most comprehensive, with Sites 3 and 4 in between the two in terms of duration and extent. By model programs, we mean that they comply with their respective state standards from batterer programs and implement curriculum that approximates the cognitive-behavioral approach registered in the prevailing published manuals in the field ( Kivel, 1992; Pence & Paymar, 1993; Russell, 1995; Stordeur Stille, 1989). In terms of their linkage to the courts, association with victim services, and provision of additional services, the programs, however, represent varying intervention systems. A research assistant administered the research materials to a monthly quota of men who appeared at the program site for enrollment. Only 5% of the men refused to complete and submit the research materials, suggesting that the respondents are highly representative of the men contacting the program during the subject recruitment period of 1995.

Sample: The men in the sample are predominately men that were arrested for domestic abuse and referred or mandated by the court to a batterer program. Men who came to the programs voluntarily or were referred by another source were also included in the sample. Men referred to the program were excluded if the victim was not an intimate female partner. Our sample does not account for men who might have been ordered or referred to the programs but did not comply. The predominantly urban sample also does necessarily represent suburban or rural programs.

Nearly a fifth of the men (18%) were voluntary program participants (13% of these men were referred by a mental health counselor). The "voluntary" men were likely to be better educated (53% had some college education versus 33% of the court referrals; X2=20.22[3]; p=.001), and white collar workers (52% versus 32%; X2=16.66[1]; p=.001) than the court-referred men, as with previous comparisons of voluntary and court-referred men ( Saunders & Parker, 1989). They also were about 10% more likely to have been fully employed, married, received mental health treatment, and obtained other counseling (these differences were not statistically significant, however). The voluntary and court mandated men were similar, however, in terms of racial composition, heavy drinking, and living arrangements. Nearly all of the men at Site 1 [94%] were court referred, primarily because the program at the time referred "voluntary" participants to independent counselors.

Referral

An additional structural difference emerged in our examination of the background data. The amount of time from court date to program intake varied considerably across the sites (see Table 1). While the average referral time was overall two months (M=8.4 weeks; sd=9.6), the delay ranged from two and a half weeks (M=2.6 weeks.; sd=1.9) at Site 1 to four months (M=16.0; sd=12.3) at Site 3 (F=32.24 [3, 839]; p=.001). This translates into as few as 4% of the men at Site 1 waiting 3 months before intake at a batterers program, as opposed to nearly a third (30%) of the men at Site 2, a half (46%) at Site 4, and over 3/4 (78%) at Site 3. The variation, at face value, appears to reflect the different referral systems used at the various sites, ranging from judicial referral directly to the program at a preliminary hearing, to referral from individual probation officers following a formal conviction. The linkage to the courts varies as well. Site 1 employs a court liaison to receive men at their hearing, and other sites rely on program staff that serve in the court as liaisons to probation officers to refer men who are part of their caseload.

The actual times from arrest to the court date and program intake to group enrollment are likely to vary as well, but we did not systematically collect data on these steps of the process. We know from observation that the time from arrest to court appearance at Site 1 is on average under a week, but tends to be substantially longer at that other sites due to different court systems and linkages. Also, the time from program intake to enrollment in a group session tends to vary across the sites depending on assessment procedures and program openings. Site 1 again appears to have the shortest time from intake to group enrollment structured into their system. Therefore, differences in the time from arrest to beginning the program are likely to be even more dramatic than those reflected in time from court date to research intake, which is reported above.

The variation in referral time may contribute to some differences in reported characteristics and abuse with certain types of men more likely to dropout or disappear during the referral delays ( Gondolf & Foster, 1991). It may also influence outcome in terms of providing time for "cooling off," additional interventions, or eventual separation from one's partner, all of which may improve program outcome. Deterrence studies suggest, on the other hand, that the closer the intervention is to the crime, the more effective it tends to be ( Fagan, 1989). We can test all these speculations further across and within sites in the outcome study in progress.

Instruments

The research materials included a background questionnaire that asked about the men's demographics, living situation, parent's behavior, mental health problems, alcohol use, prior treatment and counseling, abusive behavior, previous arrests, partner's response, and partner's helpseeking. Several of the questions represent self-report indicators for alcohol abuse and mental disorders, such as depression. The Michigan Alcoholism Screening Test (MAST; Selzer, 1971) was also administered. This test is comprised of 25 weighted items about drinking behavior and drinking-related problems. The scoring system is designed to detect potential alcoholism despite the tendencies of alcoholics to deny their drinking problems (a total score of 5 or more indicates the potential for alcoholism). The men also completed the Million Clinical Multiaxial Inventory (version III; MCMI) which includes 175 true-false items used to detect personality disorders and major mental disorders through 24 subscales ( Million, 1994). (Only 1.4% of the MCMI tests completed by our subjects were identified as invalid according to the validity measure incorporated in the test.) A computerized scoring system calculates the base rate scores (BR scores) on each subscale ranging from 0 to 115; a score of 75 or over suggests that symptoms for a particular disorder are above the norm, or "present," and a score of 85 or over suggest the "prominence" of a possible disorder. It is important to note that the MCMI is designed primarily as a clinical tool to assist in assessing and diagnosing individuals. The subscale scores are generally combined as a "profile" that is interpreted drawing on clinical experience, observation, and other sources. The MCMI scores by themselves may over predict personality disorders in batterers, according to comparisons with other personality tests ( Hart, Dutton, & Newlove, 1993).

Analysis

The following analyses were conducted to provide an overview of the batterer characteristics. The ordinal and inventory responses on the background questionnaire were collapsed primarily into dichotomous variables to facilitate interpretation of a large set of variables and comparisons among the sites. The collapse variables also tend to represent "markers" frequently used by clinicians to describe clients (e.g., employed versus under-employed, married versus not married, drug use versus no drug use). To help summarize inventories (e.g., the Conflict Tactics Scale [ Straus, 1979] for assaultive behaviors, or list of possible alcohol treatments), variables for having answered "yes" to any of the categories were developed as well as for the "total" number of categories answered in a particular inventory. The total scores for the MAST were computed using the weights prescribed for each item, and the batterers scoring above the MAST cut score (5 or greater) were identified as possibly "alcoholic." We also examined various items on the MAST (e.g., drunk drinking, alcohol-related fights) to further describe the alcohol-related behavior of the men. Finally, the BR scores on the MCMI were categorized as 75 or greater to represent those individuals with evidence of a possible "disorder." We refer to these individuals as "scoring positive" on a particular subscale (BR less than 75).

Cross-tabulations of the background variables and sites were used to examine differences across the sites, and cross-tabulations of potentially related variables were also used to test for associations. Chi square at the .01 significance level was used as the test statistic. The relatively conservative significance level compensates for the large sample size (n=840) which raises the prospect for significant results. While statistically significant differences for some variables exist among the sites, the clinical significance or impact of such differences may be minimal. The analyses, moreover, must be considered exploratory and tentative at this stage. More complex multivariate analyses using more detailed responses may detect further relationships and patterns among the variables.

Qualifications

There are two other areas that warrant caution. The selection and representation of characteristics suggests a deficit approach to batterers. That is, we tend to focus on characteristics that represent a lack of some positive or preferred aspect, or the presence of something negative or undesirable. This focus does help highlight areas that might warrant special attention during intervention or treatment; however, the strengths, changes, and possibilities of men ultimately need to be recognized and heightened as well. Furthermore, it must be emphasized that the data is based on the reports of batterers only. There are likely to be substantial discrepancies between the men's self-reports and the reports of their partners, especially given the tendency of batterers to minimize and deny their abuse. The men's self-reports are still of interest because they represent the primary, or at least most immediate, information available to batterer program staff and the courts. Verifying reports have been obtained from the men's partners and will be compared to the men's reports in subsequent studies.

Total Sample

Demographics

The characteristics of the men in our sample suggest the prevailing generalization of a younger, working class man with a high school education, who may or may not be married being the most likely to appear in programs. A closer look at our descriptive statistics, however, reveals a diversity of men marked by some important dichotomies in terms of race, age, education, employment, and marital status (see Table 1).

Demographics:The batterers in our four sites were on average in their early thirties (M=32; sd=8.8) with 2/3 of the men being between 23 and 41 years of age. A substantial portion (22%) of the men were under 25, nearly a third (31%) were over 35. Slightly over one half (45%) were men of color (31% African American, 18% Latino, and 6% other races). Almost one quarter of the men (24%) did not have a high school education, but over a third (36%) had some college education. A third (36%) were under-employed (working only part-time [20%] or unemployed [16%]), and 64% were working full-time. (An additional 7% were either disabled, students, or retired.) Employment status was, however, not significantly related to education level (X2=2.18[2]; p=.336); nor was this relationship significantly influenced by racial background (employment by education by race). Men of color were, however, about 10% more likely to be unemployed or part-time employed than other men (41% versus 31%; X2=8.93[1]; p=.003). Nearly 2/3 of the men could be classified as blue collar workers (skilled and semi-skilled laborers), while 17% indicated they were professionals, administrators, or managers.

Living Arrangements:Over half were not married and one half were not living with a partner. As one might expect, the married men were more likely to be living with their partners (60% of married men were currently living with their partners versus 41% of non-married men; X2=30.02[1]; p=.001). Half of the men (51%) did not have children living with them, but this was also influenced by whether they were living with their partner or not. The men were over twice as likely to have children living with them if they were living with their partner (67% versus 24%, X2=161.57[1]; p=.001). African American men were the least likely to not be married (59%) as compared to whites (24%), Latinos (44%), and other races (52%) (X2=8.38[3]; p=.039). Studies of battered women in shelters suggest a similar tendency with regard to martial status and associate it with differences in economic opportunities and cultural traditions ( Gondolf, Fisher, & McFerron, 1991).

Summary:The demographics of the batterers in our overall sample suggest that men of very different socio-economic and racial background are likely to be sitting next to one another in the group sessions of batterer programs. A substantial portion of men do not have a high school diploma and probably need assistance with written materials, while a similar amount have some college education. A significant group is under-employed and probably coping with financial problems, while a notable portion of men are in substantial white collar jobs. The programs' curriculums focus on one's relationship with his or her partner, yet at least half of the men were not married and half were not living with a partner.

Parents' Behavior

Alcohol and Violence:As in previous research on batterers (see Tolman & Bennett, 1990), a substantial portion of the men in our sample acknowledged being raised in troubled families--where their parents were physically abusive or had an alcohol/drug problem (see Table 1). Over a third of the men (36%) identified a parent as having a drug and/or alcohol "problem. Nearly a third (33%) acknowledged that their parents hit one or the other, and a quarter (26%) claimed to be physically "harmed" by their parents while they were growing up. Parents were twice as likely to have hit one another (60% versus 18%; X2=142.86[1]; p=001), and almost three times as likely to have harmed the man (43% versus 15%; X2=74.43[1]; p=001), if they were identified as having a drinking or drug problem. Only 8% of the parents without a drinking problem and who did not hit one another had harmed their sons (X2=69.05[1]; p=001). The men with problem drinking parents were predictably more likely to report being heavy drinkers themselves (drinking on a weekly or daily basis), or prone to drunkenness (drunk two times or more in the past three months)--but they were only 10-15% more likely to be heavy drinkers or drunk than men without problem parents (heavy drinkers=43% versus 35%; X2=4.72[1]; p=.030; and drunk=67% versus 51%; X2=21.61[1]; p=.001).

Summary:A substantial portion of the batterers admitted to violence and alcohol being problems in their family of origin. The parents' drinking problems appear to contribute to the violence in the home and to the man's eventual drinking, as previous research suggests ( Gondolf & Foster, 1991). It is significant that reporting of parental problems is as high as it is, given that these sorts of "private" issues tend to be under reported or denied. The men's recognizing and acknowledging such problems suggest that these particular men are especially conscious of their past and its impacts, and/or that the actual rate of parental alcohol problems and physical abuse is extremely high.

Mental Health

Self-reported Problems:The mental health of the men in our sample was assessed through self-reports on a conventional inventory of psychological problems (7 items) and through scores on the MCMI subscales. Few batterers reported problems on the inventory which might be associated with major mental disorders, but a substantial portion acknowledged emotional problems and even serious depression (see Table 2). Specifically, a quarter to a third of the men reported serious emotional states during the past three months in the form of angry outbursts (35%), serious anxiety (27%), or mood swings (24%). Nearly one fifth (18%) admitted suffering serious depression in the last 3 months. Two-fifths (41%) of the men did not report any problems on the self-report inventory, while almost a third (29%) reported two or more problems (outbursts, anxiety, mood swings, serious depression, suicide, thoughts of killing, hearing voices).

The relationship between past suicide and current depression implies that the current depression of at least a third of the men may be long-term rather than a reaction to the men's immediate situation. Of those reporting recent depression, nearly a third (31%) reported previous suicide threats or attempts, as opposed to only 7% of those not depressed (X2= 55.71(1); p=.001). Twelve percent (12%) of the men overall reported having threatened or attempted suicide sometime in the past. Only small percentage of the men acknowledged more serious symptomatology in the form of recent suicide ideation (7%), homicide ideation (3%), or auditory hallucinations (1%) occurring in the past three months. Over one fifth (22%) of the men had received some form of mental treatment sometime in their past, either in the form of prescribed medication (8%), mental health counseling (16%), or psychiatric hospitalization (6%). A small percentage (6%) were currently on medication.

MCMI Results:The results of the MCMI suggest that as much as a fifth of the men may have what are considered to be major mental disorders (DSM-IV: Axis I disorders excluding anxiety disorder, and alcohol and drug dependence) (see Table 5). At least half of these men (or 11% of the total sample) acknowledged symptoms of major depression. In addition, nearly 40% of the men also appeared to have an anxiety disorder. This surprisingly high level of anxiety may reflect the men's response to being arrested and ordered to the batterer programs.

Similar to men tested with the MCMI at a Midwestern program ( Hamberger & Hastings, 1991), nearly all of the men (90% in our 4-site sample versus 88% in the midwest) had scored positive (BR score is less than 75) on at least one of the subscales for personality disorders (Axis II disorders). Nearly half of the men in our four-site sample achieved a score (BR is less than 85) on one of the personality subscales (Axis II), suggesting a personality pattern was "prominent" as opposed to "present." However, only 16% of the men showed at least minimal evidence of what are considered severe personality pathologies (BR is less than 75 for schizotypal, borderline, or paranoid disorders).

Previous MCMI Studies:As suggested in previous research on batterers using the MCMI ( Beasley & Stoltenberg, 1992; Hart, Dutton, & Newlove, 1993; Hamberger & Hastings, 1991), there was not a prominent or typical set of personality problems among the batterers. Over a third (38%) of the men scored high (BR is less than 75) on at least 4 or more of the 14 personality subscales, suggesting a complex pattern of personality problems. Moreover, the evidence for some of the notable generalizations about batterer personalities was lacking. For instance, a low portion of men in our sample scored positive on the borderline (6%) and compulsive (10%) personality subscales ( Hart, Dutton, & Newlove, 1993; Dutton & Starzomski, 1993). Also, very few scored positive on post-traumatic stress syndrome (4%), contrary to a previous study with a smaller clinical sample from a Canadian program ( Dutton, 1995). The men were most likely to score positive on the narcissistic (25%), passive-aggressive (24%), antisocial (19%), and depressive (19%) subscales. Interestingly, the prevalence rate for a depressive personality disorder (19%) is equivalent to the rate for those reporting recent "serious depression" (18%).

A factor analysis of the basic personality scores (excluding the three subscales for clinical "personality pathology") produced two primary factors (Eigenvalue >1) approximating an anti-social/passive-aggressive factor (51% of the variance) and a schizoidal/borderline factor (14% of the variance), as compared to a three factor structure of schizoidal/ borderline (44% of the variance), narcissistic/ antisocial (25%), and passive dependent/compulsive (11%) from the midwestern program study ( Hamberger & Hastings, 1991). (The difference in factors may in part be due to the use of the MCMI-III with our sample, as opposed to the MCMI-II with the other.) The factors, which do not of themselves indicate "types," were used to identify 8 different personality types in the midwestern program.

Summary:At least a third of the batterers did report agitated emotional states that are likely to correspond to their recent abuse and arrests. Most notably, almost a fifth reported being seriously depressed, and at least a third of these had a history of suicide threats or attempts. A higher portion of the men than expected have received some mental treatment, but few admitted to ever being hospitalized. There is little evidence of major mental disorders on a conventional inventory; however, a personality test (MCMI) suggests that over a quarter (29%) of the men may have major disorders (Axis I, excluding anxiety, alcohol, and drug disorders) or a severe personality pathology (Axis II schizotypal, borderline, or paranoid personality disorder).

Self-reported psychological problems do not successfully identify corresponding disorders registered on the MCMI, even though prevalence rates for the self-report of recent depression and anxiety are equivalent to the rates for depressive personality disorder and anxiety disorder derived from formalized testing. (Reported depression identifies only 40-45% of those scoring positive on the MCMI for depression-related disorders.) Reliance on self-reports to screen for mental disorders may not, consequently, be a sufficient. It remains unclear what disorders affect program compliance or outcome, and need to be systematically identified and addressed. Ideally, our multi-site outcome will offer some indications in this regard.

Alcohol and Drug Use

Self-reported Alcohol Use:According to both self-reports and the MAST, the portion of men with drinking problems is at the high end of the range generally ascribed to arrested batterers (see Tolman & Bennett, 1990). Approximately a third of them in our sample, moreover, have alcohol-related behavioral problems, such as drunk-driving arrests and fights. Nearly 38% of the men actually admitted to what might be considered heavy drinking (drinking at least weekly) or to being a recovering alcoholic (9%) when asked about their frequency of their current drinking (see Table 2). A quarter of the men admitted to drinking at least few times per week to everyday. Over a half (57%) indicated they were drunk at least twice in the last three months with a fifth (21%) drunk once a week or more. By contrast, almost a third of the men (31%) reported that they did not drink during the past year. (This amount includes the 9% of men who identified themselves as being in "recovery".) The men claimed that nearly as many of their partners were heavy drinkers (32%) or were as frequently drunk (48%) as they were. A fifth (20%) of the men admitted to using marijuana in the past year, and 10% had used other drugs during the previous year.

In response to a series of questions about treatment, a quarter (26%) of the men responded that they had been in alcohol or drug treatment sometime in the past: either detox (5%), self-help groups including AA or NA (15%), or inpatient or outpatient treatment (16%). Ten percent of the men had received alcohol treatment and mental health treatment sometime in their past, and nearly half of the men (48%) had received some other form of counseling.

MAST Results:The screening test for alcohol abuse (Michigan Alcoholism Screening Test [MAST]) indicated that over half (56%) of the men in our sample may be considered "alcoholic" (see Table 6). (The more liberal cut score [4 points or more] often used with clinical populations suggests that as much as 64% of the sample may be alcoholic.) Specific items on the screening test illustrate the nature of the drinking and related behavior (see Table 6). Forty-one percent (41%) of the men admitted they are not "normal drinkers" on the screening test, a third (34%) admitted they have difficulty stopping after two drinks, and over a quarter (28%) acknowledged attending Alcoholics Anonymous in the past. About a third of the men have severe behavioral problems associated with their drinking (31% scored 10 points or more on the MAST) in the form of alcohol-related fights (31%), drunk driving (24%),or other alcohol related arrests (20%). (The amount of drunk driving may be lowered by the fact that many of the inner city men do not have cars or walk to and from local bars.) Half of the men (49%) admitted to being previously arrested for any offense other than domestic violence, irrespective of drinking.

There were some encouraging similarities in the prevalence rates of self-reported alcohol abuse and the test results for alcoholism and dependence. The rate of men reporting heavy drinking or alcohol recovery approximates the rate scoring as "alcoholic" on the MAST, and the rate reporting frequent drunkenness (at least once a week) approximates the rate testing as "alcohol dependent" on the MCMI. (The MCMI alcohol subscale is sensitive to addiction with a psychological basis, rather than heavy drinking and problem behavior as in the MAST.) However, the "heavy drinking" correctly corresponds to the MAST classifications (cut score 5) in 54% of the cases, and "frequent drunkenness" corresponds to the MCMI alcohol dependency subscale (BR 75) in 64% of the cases.

Summary:Our findings regarding alcohol use substantiate the general impression that a disproportionately high percentage of men in batterers programs have alcohol problems. Well over half of the batterers in our sample screen positive for alcoholism and nearly as many men admitted to heaving drinking or drunkenness. At least a third have serious behavioral problems associated with their drinking, as well. As suggested with regard to other characteristics, the batterers in our sample pose, however, a dichotomy with regard to drinking. While a substantial portion have serious drinking problems, almost a third currently do not drink or seldom drink. While self-reports of drinking frequency and an alcohol screening test indicate similar prevalence levels of alcoholism, the self-report and test results do not substantially correspond for individuals. As with mental disorders, programs may need to use assessment tools for alcohol abuse.

Abusive Behavior

Not surprisingly, the men in our sample admitted to a relatively low level of abuse in the form of control, threats, and assaults (see Table 3). The abuse that they did report is often classified as the less severe forms ( Straus, 1979), yet a substantial portion of the men did tell of their partner's seeking medical attention for injuries. The vast majority did not think that they would be violent in the near future.

Control:Very few of the men (less than 10%) admitted to controlling behaviors often associated with abusive relationships (e.g., stopping a woman from going someplace, ordering her off the phone, keeping her from friends, limiting her access to family finances, following her against her will) during the past three months. A fifth of the men (19%) had committed at least one of these controlling behaviors. Nearly half of the men (46%), however, indicated that they had swore, screamed, or insulted their partner in the past three months, and about a fifth of the men had accused their partner of being with another man (18%), or threw, smashed, or banged something (23%).

Threats:Similarly, over a quarter of the men (29%) acknowledged making threats of some sort in the previous three months. Very few (less than 5%) conceded making what battered women rate as the most serious kinds of threats (threatening to kill the partner, take or ham the children, kill or hurt others, kill or hurt oneself), as opposed to threatening to throw something (17%) or threatening to hit, harm, or attack one's partner (13%).

Assaults:The obvious discrepancy appears with the men's reports of assaults on the Conflict Tactics Scale ( Straus, 1979). Less than half of the men (42%) reported some form of assault during the last three months, even though this 3-month period includes the arrest incident that brought the men to the program reporting-period for "recent" abuse. A portion of the men (36%) were, however, arrested prior to the 3-month. Over half (58%) of these men reported no assaults; this accounts for a fifth (21%) of the total cases reporting no assaults because the arrest incident was more than three months ago. Conversely, nearly 40% of the recently arrested men denied committing an assault.

The overall severity of abuse appears to be minimized as well. Only a small portion of the men (less than 10%) reported severe forms of violence for the previous three months (as defined by the Conflict Tactics Scale: hit with fist, kick, bit; hit with something, beat up, burned or scald, choked, threatened with a weapon, used a weapon, forced sex). Less than a fifth (18%) of the men indicate using two tactics, or more during that period or committing a serve assault (17%). Only 1% indicated using weapons or forcing sex.

Past Assaults:Some hint of severity emerges in the reports about past abuse and injury. Approximately 60% (59%) of the men did concede to assaulting a partner sometime in the past with a third (32%) reporting the use of a so-called "severe" tactic against their partners. Half (52%) of the sample acknowledged causing bruises sometime in the past, and a fifth (19%) of the men reported their partner sought medical help because of injuries they caused. Only 5% of the men admitted to "physically striking" their children in the past three months. This low percentage may be influenced in part by the fact that over half (56%) of the men currently have no children living with them, and that the respondents were notified that the researchers may have to report child abuse.

Half of the men (51%) insisted that the first assault occurred within the past year, and nearly a third (31%) admitted that the first incident was three years or more ago. However, at the initial program contact, only 15% of the men conceded that they are likely to be violent again in the next three months (somewhat likely, very likely, or uncertain)--that is, the vast majority of men (85%) think they will not be violent in the near future.

Summary:The low level of abuse reported by the batterers, even though the report period encompassed their arrest for battering, confirms the common concern about men's tendency to underreport or deny abuse. The men also appeared overconfident about stopping their abuse with the vast majority predicting they would not be violent in the near future. These apparent tendencies support the need for programs to address the minimization of abuse and to seek corroborating information about batterers' domestic violence arrests and their abusive behavior in general. It tends to support, moreover, the component of many court-referred programs devoted to defining abuse and prompting men to admit their abuse.

Partner Response

The battered women, according to the men, responded to men's abuse primarily through personal strategies (in striking back, threatening separation, going to a friend's house) and the criminal justice system (police calls, protection orders, pressing charges). Only a small portion of the women had previously contacted a shelter or social service agency. (See Table 4).

Partner Aggression:From the men's point of view, the women were nearly as aggressive as they were, but much more likely to need medical help for their injuries. Approximately 40% of the men (39%) claimed that their partners assaulted them during the past three months--nearly the same proportion of men (42%) who reported assaulting the partner during the same time period. In the vast majority of cases (80%), the men who reported assaulting their partners were the ones who reported their partner acting violent toward them. A quarter of the men (26%) claimed their partners committed an act that might be considered "severe" (according to the Conflict Tactics Scale), as compared to a fifth of the men (17%) admitting severe tactics toward their partners. This pattern holds for violence reported happening ever in the past: 59% of the men admitted violence toward their partners, and 52% claim being assaulted by their partners in the past. There was a more distinct difference in terms of past injury, with a third of the men (37%) reporting being bruised or injured and only 5% reporting ever seeking medical help for injuries, as opposed to over half (52%) of the women being bruised and a fifth (19%) of them seeking medical help.

Helpseeking:Over two thirds of the women (68%) responded by also seeking help sometime in the past. According to the men, the women tended to turn to what might be considered "personal" or "informal" help sources in response to the men's abuse. Approximately one third (38%) threatened divorce or separation, and a quarter (26%) stayed overnight a friend's or relative's residence. However, only a small percentage (14%) sought counseling (10%) or contacted a shelter (7%) ever in the past. These battered women instead appeared to rely on law enforcement: 40% of the women had contacted law enforcement either through a call to the police (32% prior to the current incident that brought the man to the program), obtaining a protection order (15%), seeking legal assistance (5%), or pressing criminal charges (9%).

Summary:The men's report of women's aggressive response to their abuse raises several issues. It verifies many men's tendency to accuse their partners' of being abusive and their perception of abuse as a "two-way" street. Many practitioners in the field argue that this tendency is a means to deflect responsibility for one's abuse by blaming one's partner and justify one's own violent behavior ( Pence & Paymar, 1993). The men's reports also corroborate the concern that a behavioral checklist of abuse, such as the Conflict Tactics Scale, may grossly distort the dynamics and nature of battering. In subsequent analysis using our multi-site database, we will be able to compare the women's narrative accounts of battering and descriptions of the battering on police reports with the men's reports on the tactics checklist. Finally, the men's reports imply that the battered women of court-referred batterers rely on themselves to fend off their partner's abuse or cope with it. In fact, the women tended to use personal strategies, such as threatening divorce or staying at a friend's or relative's house, rather than more formal help-seeking sources.

These battered women, furthermore, appear more likely to use the criminal justice system to deal with the abuse, rather than social services including battered women's programs. Their helpseeking appears to move directly to the most intensive help source (the criminal justice system), instead of progressing through various levels of help sources (e.g., personal strategies, to informal sources, to formal help sources, to criminal justice interventions). The women partners of men in batterer programs are not likely to have been involved in shelters--a finding that is consistent across our research sites. The low percentages of women contacting shelters or counseling services may reflect the men not being as aware of the women seeking shelter or counseling as they are of them contacting the police. On the other hand, some women may simply find calling 911 more convenient, or may not consider the local women's shelter as readily accessible (e.g., some shelters do not accept male teenage children; crisis counselors are sometimes overloaded).

Site Differences

Demographics

Race:The research sites, while remarkably similar in age distribution, substantially differ in racial composition and corresponding in socio-economic status (see Table 1). The program differences in race across the sites appear to reflect, at least in part, respective differences in regional urban populations. Over 2/3 (68%) of the participants at Site 3 are men of color, as opposed to over one third (38%) of Site 2 being of composed of men of color (X2=45.63[3]; p=.001). While the men of color at Site 1 are nearly all African American (47% of the men in Site 1 are African American), Site 3 has at least a third of the men (34%) identifying themselves as Hispanic or Latino (X2=166.30[9]; p=.001).

The racial composition of the batterer programs reflects the proportions of their respective city populations, except for Site 1. This site has over 20% more African Americans than the rate for the city itself. This may reflect: 1) the court jurisdiction being confined largely to inner city neighborhoods at Site 1, 2) African American women being the most likely to call police in response to domestic violence and Latino women the least likely, according to recent research on police utilization ( Hutchison, Hischel, & Pesackis, 1994), and 3) the legal system's discrimination against African American males in arrests, convictions, and sentencing.

Other Characteristics:Site 1 appears to have more lower-income, less educated, under employed men, and unmarried men than the other sites, and Site 2 tends to have more men of higher socio-economic status. More specifically, Site 1 has fewer men with some college education as compared to Site 2 having almost twice as men with college education (27% versus 45%; X2=24.49[6]; p=.001). Similarly Site 1 has nearly twice the unemployed or part-time men (58%) as the other three sites (30% versus 30% versus 27%; X2=57.03[3]; p=.001), and has the vast majority of men (78%) identifying themselves as blue collar workers as opposed to white collar along with Site 3 (72%) (X2=37.66[3]; p=.001). These differences may reflect the socio-economic situations of the respective sites, with Site 1 having the least employment opportunities among the sites.

The men at Site 1 were also more likely not to be married to they women they had abused (70% versus approximately 50% across the other sites; X2=38.26[3]; p=.001), even though there was no significant difference in the proportions not currently living with their partners or not having children living with them. The marital status of the men at the various sites appears to reflect marital tendency among the racial groups. Lower-income African American men were the least likely to be married, and Latino men were the most likely to be married, reflecting in part their respective social circumstances and cultures ( Gondolf, Fisher, & McFerron, 1991).

Summary:Site 1 is distinguished by more African American men and men with less education and less employment opportunity, and Site 2 tends to have more white men and men with higher education and employment status. (Race does not of itself significantly predict the educational and employment levels of the respective sites.) The other two sites are more comparable in their racial composition, educational levels, and employment status. These differences most likely reflect regional differences, but may also be influenced by the court systems that the respective programs serve.

Mental Health Problems

MCMI Results:The portion of men with mental disorders was relatively similar across the sites overall, according to the MCMI (see Table 5). Although not statistically significant, the men at Site 4 were 10% more likely to score positive for a major disorder (Axis I) (60% versus 54%, 50%, 51%; X2=4.51(3); p=.211) and for personality disorders (Axis II) (95% versus 89%, 89%, 86%; X2=10.41(3); .015), but this relationship is not statistically significant. Men at Site 1 were also more likely to score positive on a combination of personality disorders (less than 7; 16%, versus 9%, 8%, 6%; X2=31.65(6); p=.001). These tendencies reflect significant scoring differences in a few areas. Site 4 was almost twice as likely to have men scoring positive for the presence of passive aggressive tendencies than Sites 3 and 4 (33% versus 19%, 19%; X2=15.98(3); p=.001); and the men at Site 4 were also significantly more likely than the other sites to have men appearing to have a paranoid personality pathology (15% versus 12%, 6%, 9%,15%; X2=11.49(3); p=11.49; p=.009) and an anxiety disorder (49% versus 33%,36%, 40%; X2=11.92; p=.008).

Self-report:Despite the differences in demographic characteristics, the differences in the men's report about parental problem behavior and their recent psychological problems did not vary significantly across the sites (see the bottom of Table 1 and top of Table 2). However, nearly twice as many men at Site 2 had received some mental health treatment than at the other sites (36% versus 16%, 17%, 20%; X2=31.12[3]; p=.001). Specifically, the men in Site 2 were more likely to have received medication (13% versus 4%, 9%, 7%; X2=12.65[3]; p=.005) or mental health counseling than men at the other three sites (25% versus 11%, 15%, 15%; X2=16.54[3]; p=.001). This treatment difference may reflect the fact that the men at Site 2 are more likely to be of higher income and economic status and, therefore, more able to seek and obtain these kind of services. Also, the number of therapists and mental health programming in the city of Site 2 is substantially greater than at the other sites.

Summary:While the prevalence of mental disorders and reported psychological problems was similar across the sites, Site 4 appeared to have a greater portion of severe disorders. This difference corresponds with the greater likelihood of men at Site 4 to report recent verbal abuse and assaults. The men at Site 4 were also more likely to score high on the disclosure scale of the MCMI (BR score is less than 75, 23% versus 14%, 11%, 15%; X2=12.34(3); p=.006) suggesting greater openness in acknowledging symptoms. The Site 4 difference may, therefore, in part be the result of the disclosure especially concerning anxiety and tension over the men's situation with the courts. Men with high disclosure (BR score is bigger than 75) were over twice as likely to score positive on the anxiety disorder subscale at Site 4 (83% versus 39%; X2= 29.11[1]; p=.001) and in general (85% versus 31%; X2= 133.42[1] p=.001). These differences in disclosure may be attributed to minor differences in data collection procedures. The research materials were administered at Site 4 in a more personable way than at the other sites. The MCMI was read aloud by the research assistant rather than through an audio tape, and was administered to smaller groups, after the men had introduced themselves and briefly discussed their relationship.

Alcohol and Drug Abuse

MAST Results:The MAST scores were equivalent at all four sites, suggesting the proportion of alcoholics was similar across the sites, as is the proportion of men with mental disorders (see Table 6). The alcohol screening, however, identifies men likely to have alcoholic tendencies, but does not necessarily indicate the men with the most severe drinking problems. Some MAST items suggest behavioral differences among sites that may be reflected in variations of the men's report about their drinking. For instance, the men at Site 1 were twice as likely to report being involved in alcohol-related fights than the men at Site 3 (39% versus 20 %; X2=20.92[3]; p=.001). The men at Site 2, on the other hand, were the most likely to have been involved in drunk driving (33% versus 22%, 18%, 25%; X2=20.92[3]; p=.002), but they are also more likely to have cars given their higher economic status and geographical dispersion of their city.

Self-Reports:The sites did significantly differ in the proportion of men reporting heavy drinking and drunkenness (see Table 2). Site 1 had the highest levels of drinking. This may in part reflect the lower socio-economic status of the men at this site, since an association exists between the demographics and reported drinking in the sample of batterers as a whole. Over half of the men (53%) at Site 1 reported drinking at least weekly as opposed to one-third (34%) at Site 2, a quarter of the men (25%) at Site 3, and 41% of the men at Site 4 (X2=37.78[3]; p=.001). Similarly, nearly three-fourths of the men (71%) at Site 1 reported drunkenness at least twice every three months as opposed to about half of the men at the other sites (X2=27.66[3]; p=.001). The men at Site 1 and Site 2 were also more likely to report smoking marijuana in the last year (29% and 24 % versus 12% and 17%; X2=23.62[3]; p=.001) or using other drugs (14% and 12% versus 6% and 7%; X2=12.95[3]; p=.005) than the men at the other two sites. Furthermore, the men at Site 1 were significantly more likely to report their partner being drunk (62% versus 48%, 37%, 44%; X2=28.65[3]; p=.001).

The proportion of men receiving previous alcohol or drug treatment did not, however, vary across the sites, but the proportion receiving inpatient or outpatient rehabilitation was significantly higher at Site 2 (24%) than at, especially, Site 3 (9%) and the other two sites (18% and 14%) (X2=20.43[3]; p=.001). The higher level of drug use and the available treatment options in the city where Site 2 is located may contribute to the higher proportion of men receiving alcohol or drug treatment at Site 2.

Summary:The proportions of men reporting parental problems, mental health symptoms, and alcohol and drug abuse were amazingly similar across sites despite the demographic differences. However, men at Site 1 were more likely to report heavy alcohol and drug use, even though the portion of men screened as "alcoholic" on the MAST test was similar across sites. The heavier drinking and drug use at Site 1 appears to be influenced in part by the greater portion of men with lower socio-economic status at that site.

Abusive Behavior

A greater percentage of men at Site 4 reported abusive behaviors in terms of recent verbal abuse and assaults (see Table 3). Yet men at Site 2 were especially more likely to report being assaultive in the past. The reporting levels of controlling behaviors, serious threats, and severe assaults are so low that it is difficult to detect more specific differences for abuse tactics. The apparent overall differences do not appear to be explained by background characteristics and may be related to system or testing differences.

Control & Threats:The percentage of men acknowledging various forms of controlling behavior in the past three months did not vary significantly across the sites, except that men at Site 4 (58%) were more likely, especially than men at Site 2 (36%), to have reported verbal abuse in the form of swearing, screaming, or insulting (X2=25.20[3]; p=.001). Men at Site 4 (23%) were also more likely to report threatening to throw something than the men at Site 1 (12%) and Site 2 (14%) (X2=10.21[3]; p=.002). In fact a greater portion of men at Site 4 reported one of six kinds of threats than at other sites (44% versus 24%, 23%, 27%; X2=26.28[3]; p=.001).

Assaults:The men's reports of assault, although admittedly low overall, were greatest at one site for the past three months, but higher at other sites for ever in the past. Site 4 had the highest percentage of men (64%) admitting to an assault in the past three months -- over twice the percentage of Site 2 (27%) -- and a greater percentage of men reporting severe assault (22%) than Site 2 (10%). These differences appear even though the delay between court appearance and program intake is similar at the two sites, and education level, employment status, and marital status are similar at the these sites. However, a greater percentage of men at Site 2 (60%) and Site 1 (74%) report assaults ever in the past than at the other sites (55% and 46%) (X2=38.05[3]; p=.001). The same pattern appears for severe violence reported in the past. The amount of men reporting injuring their partners is similar across the sites, however.

Summary:More recent abuse was reported at especially Site 4, and more abuse in the past was reported at Site 2. There are several possible reasons for this. The men at Site 4 may have committed more serious violence to get caught and referred to batterer programs. The laws and court at Site 2 may facilitate the apprehension and referral of less severe events, even though the abuse may have been going on longer and more severely in the past. Also, the questionnaire time frames for assault may have been explained differently at the sites, or the men may have been prompted more at Site 4 to disclose their most recent violence, as the higher scores on the disclosure subscale of the MCMI may suggest. (Our monitoring of the data collection did not, however, reveal these latter sorts of differences.)

Partner's Response

Variations in the men's reports of their partner's physical response reflected their reports of assaultive behavior, but the help-seeking of the partners was similar across sites regardless of the variation in demographics and reported abuse (see Table 4). The men at Site 4 (54%) were again twice as likely to report their partner assaulting them during the past three months as those at Site 2 (27%) (X2=32.39[3]; p=.001). However, men at Site 2 (60%), as compared especially to Site 4 (44%), were more likely to report their partners assaulting them ever in the past (X2=13.33[3]; p=.004).

There is no significant differences among the sites in terms of the prior help-seeking of the battered women, however. They have similar levels of shelter contact and police and court assistance as well. These findings are striking given the differences in demographic characteristics at the sites and different levels of reported recent versus past abuse. The court systems, resources, and available services may be very similar across the sites accounting for a constant level of help, or the woman may be responding to what amounts to similar levels of abuse over time. An additional possibility is that the limited help available to battered women and the tendency of batterers to blame and restrict battered women may contribute to the relatively low level of social service help (shelter and counseling) and the women's tendency to revert to the police in desperation.

Conclusion

Overall Summary

Generalizations:Our sample of batterers from four different cities confirms prevailing stereotypes of batterers in court-referred batterer programs. They tend to be men of color in their early thirties and of lower socio-economic status. Nearly half of our sample was not married and a half was not living with their partners. Over a third of the men acknowledged their troubled family backgrounds (violence and/or substance abuse) and at least a third admitted to their own heavy alcohol use. In fact, over half of the batterers may have alcoholic tendencies, according to a widely used alcohol screening test (MAST). Similarly, a substantial portion appear to have serious drinking problems, yet nearly a third report not drinking at all during the last year. Only a quarter report using other drugs in the previous year. Most of the men (68%) appear to have other problem behaviors beyond domestic violence in the form of either fights, drunk driving, or previous arrests.

Psychological Problems:A personality test (MCMI-III) suggests that over a quarter of the men may be diagnosed as having a severe personality disorder (shizotypal, borderline, or paranoid) or major mental disorder (excluding anxiety disorder and addiction). The vast majority score positive on at least one personality disorder subscale, but typifying or predominant profiles are evident. In terms of self-reported mental and emotional problems, the men do not readily appear seriously disturbed or deviant as a group. A third or so report recent emotional problems (angry outbursts, serious anxiety, mood swings) that may be associated with their abuse, but very few report symptoms associated with major mental disorders other than depression. As in previous studies, nearly a fifth of the men report being seriously depressed in the recent past. Approximately a third have received either previous alcohol and drug or mental health treatment of some kind.

Abuse:As court-referred cases, nearly all the men in our sample were arrested for a recent domestic assault. However, less than half admitted to committing an assault in the previous three months, and less than a fifth of the sample acknowledged severe abuse. Over half of the men admitted to at least bruising their partners sometime in the past, and a quarter admitted to their partners seeking medical attention because of their injuries. Interestingly, the men portray their partners as being as assaultive as they were, but few of the men (5%) sought medical attention for any injuries. These reports, of course, ignore the dynamics of the apparent conflicts and do not account for the women's account of the incidents. According to the men, the women tended to rely on the criminal justice system for assistance (40% sought a protection order, legal assistance, criminal charges or police response in the past) rather than social services, including battered women's shelters (14%).

Dichotomies:While there is some association between educational indicators and other case characteristics, the socio-economic status of the men does not significantly account for the behavior of the men. The batterers in fact might be more appropriately characterized as a diverse group of men marked by several dichotomies that have implications for intervention. While nearly a quarter of the men did not have a high school education and may have reading problems, another third have some college education. While as much as a third was under-employed, a fifth are professionals, administrators, or managers. Half were married, half not; half were living with their partner, and half not.

Site Differences

The men at the four research sites differed in demographics, yet were relatively similar in terms of alcohol and behavioral indicators. The sites significantly differed in their racial composition, largely as a result of their regional populations. Site 1 had a greater percentage of men from lower socio-economic status. The percentage of men screened as "alcoholic" (on the MAST) and having severe mental disorders was equivalent across all four sites. One site (Site 4) had a greater proportion of men reporting recent abuse than the other sites, but a lower amount reporting abuse in the past. Interestingly, the help seeking efforts of the women, both in terms of human service contact and criminal justice assistance, did not significantly vary across the sites.

Some areas where the sites differed are as follows. A higher portion of men at Site 1 reported drunkenness, alcohol-related fights, and use of other drugs. The lower socio-economic status of the men at this site (reflecting in part the greater percentage of court, as opposed to "voluntary", referrals) appears to account for some of this behavioral difference. Site 4, with the greater percentage reporting recent abuse, also had a slightly higher portion of men with anxiety and paranoid disorders, according to the MCMI. These differences may, in part, reflect a more personable administration of the research materials at Site 4, which might have facilitated more self-disclosure. Even though Site 2 had the smallest portion of serious disorders, it had a substantially higher portion of men who had sought treatment. The high number of therapists in the region of Site 2, and higher socio-economic status of the men there, may account for the high rate of previous treatment.

Program Implications

Compounding Problems:The characteristics of the batterers in our sample illustrate the additional problems which may, in fact, affect men's response to programs as well as compound their abuse. The low socio-economic status, abusive and alcoholic families of origin, mental health and alcohol problems, and previous arrests may all warrant special attention. Not surprisingly, previous research suggests these sorts of problems increase the likelihood of program dropout ( Grusznski, Carrillo, 1988; Hamberger & Hastings, 1989). The field is currently weighing at least two sets of options in this regard. Does case management, additional programming, or referral improve the outcomes for problematic cases? Or, does focusing on abusive behavior significantly interrupt and reduce reoffense regardless of the men's characteristics, much as the proponents of managed care and brief therapy might suggest.

Group Dichotomies:A second issue implied in our findings is that apparent dichotomies among the batterers warrant revisions for curriculum-based batterer programs. For instance, do men with such different levels of education (and reading levels) need different kinds of materials and instruction? Do men of different racial backgrounds need separate groups or culturally-sensitive curriculum? Do men living on their own need a different focus than men who are living with their partners? On the other hand, the behavioral commonalties of the men may transcend their demographic differences. The similarities in the nature of their abuse, their minimization and/or justifications about abuse, their experience with the legal system, and difficulty with their relationships, all offer some unifying issues and themes to address in batterer programs.

Screening Extremes:One could also argue that, while a portion of the men do have complicating problems, the majority of the men appear relatively "normal." Only a small percentage could be considered psychopathic or distinctly deviant. The diversity of the men, in fact, is fairly similar to that in many public substance abuse programs which focus on the men's substance abuse regardless of background. The mission of batterer programs may, therefore, be to deal as effectively as possible with the fairly "normal" majority, refer others screened as having severe mental or alcohol problems to additional services, and count on the courts to address the portion who drop out or do not comply. Our findings suggest that formalized testing or screening instruments, instead of self-reported problems, may be necessary to identify such cases.

Research Implications

Batterer Types:Our preliminary summary of batterer characteristics implies, as previous research suggests ( Holtzworth-Munroe & Stuart, 1994), groupings of different "types" of batterers. Educational levels, for instance, appear to be associated with drinking and abuse, as are parental problems. However, the initial efforts to identify prominent types amidst the diversity and dichotomies of batterers were unsuccessful. Cross-tabulations among demographic, family of origin, and behavioral indicators were either not significant or account for little variance. Moreover, associations that did emerge may not be pronounced enough to be clinically significant--that is, make a difference to program staff. The question remains whether a distinct typology of characteristics can be established with a more sophisticated multivariate analysis using ordinal variables as opposed to our collapsed response categories. One of the subsequent tasks, therefore, is to systematically explore the possibility of types and to determine if those types predict outcomes.

The personality problems of batterers have received increased attention in response to the interest in typologies and proposals to differentiate treatment. A preliminary look at the MCMI results with our sample suggests a diversified picture that warrants further examination. The MCMI is designed for a case-by-case interpretation based on the composite of elevated scores. The profiles with clinical samples are often complex and difficult to interpret without additional information. As with previous MCMI research ( Hamberger & Hastings, 1991), our combination of scores could be further analyzed using factor analysis and factor scores. We might also review the individual profiles of "unsuccessful" cases to determine if some disorders are predominant among men who drop out or reoffend. Both these analyses need to be controlled with socio-economic status, alcohol and drug use, and additional interventions to better assess the context of the outcomes. Profiles may be an extension of an individuals social circumstances, a manifestation of his abusive tendencies, merely consequential to a man's violence, or a primary factor in a man's violence and abuse.

Reporting Discrepancies:It is of course important to emphasize that the characteristics summarized above are based on the reports of batterers themselves. This population has been shown to especially minimize their reports of abuse at program intake ( Edleson & Brygger, 1986). Batterers' reports on other behaviors may be distorted or incorrect, as well. A subsequent analysis drawing on reports from the men's partners will identity the nature and extent of discrepancies between the batterers and their female partners. Do men who underreport violence, distort other aspects of their behavior, such as drinking? Do they misrepresent themselves consistently or selectively? Are there certain types of men that under report or distort their behavior? We will examine not only the discrepancies at intake, but also during the follow-up period of the multi-site evaluation. Do the discrepancies between men's and women's reports decrease over time and after intervention, as some previous research suggests ( Edleson & Brygger, 1986)? Finally, we will consider the validity of the self-reports in general compared to police and hospital records, and the intake testing (MAST and MCMI).

Benchmark Comparisons:Our sample of four sites, using the same measures, offers a benchmark to assess changes in batterer characteristics over time and across other cities and regions. The characteristics of our sample address questions about how much we can generalize from previous outcome studies and practice debates. The court-ordered batterers in our sample are, for instance, much more diverse racially than the samples used in notable previous outcome studies conducted in the Midwest (Minneapolis [ Edleson & Syers, 1990; Edleson, & Grusznski, 1989]; Indianapolis [ Roberts, 1987], and Kenosha, Wisconsin [ Hamberger & Hastings, 1988]), and Canada (Vancouver [ Dutton, 1986]). A greater percentage of men in our sample are also not married, more likely to be employed, and a few years older on average than men in some other research studies, but are comparable to others in terms of educational level, parental problems, and previous treatment even though the other studies occurred as many as 8 years ago ( Roberts, 1987).

Moreover, the demographics (except for race which was not indicated) and prior arrests were nearly identical to a 5-year-old experimental study of batterer counseling versus probation in Baltimore county ( Harrell, 1991). We must infer from the limited data on couples counseling with batterers and their partners that our sample of predominately court-referred batterers is dramatically different in characteristics and behavior ( Brown & O'Leary, 1995). The samples from couples counseling appear to be predominately white, from higher socio-economic status, and with substantially less alcoholism, fewer mental disorders, and fewer prior arrests. Except for couples counseling, the men in court-referred batterer programs appear surprising similar across sites and over recent years, even with differences in program structure, court systems, police practices, and victim services.

A Summary of Preliminary Findings

Research Summary

Four batterer intervention systems are being evaluated using a naturalistic, comparative research design, as opposed to a clinical trial or experimental design. The four research sites were selected to represent a continuum of intensive and extensive intervention. The sites include 1) a pre-trial, 3-month, didactic program with court liaisons (Pittsburgh), 2) a post-conviction, 3-month, process program with women's services (Dallas), 3) a post-conviction, 5-month, didactic program with legal advocacy (Houston), and 4) a post-conviction, 9-month, process program with complementary services (including substance abuse treatment, individual mental health counseling, and women?s services coordinators) (Denver). A background questionnaire, an alcohol test (MAST), and a personality test (MCMI) were administered to 210 men at each of the four sites for a total sample of 840. Their partners were also interviewed by phone at the time of intake.

The men and their battered partners (and 170 identified new partners) were interviewed by phone every 3-months for 15 months. Police arrest reports, program records and participation reports, rearrest records during follow-up, and women's medical records are also being reviewed and analyzed to help verify self-reports. Site-visits, which included staff interviews and observations of group sessions, were also conducted.

The preliminary findings listed below are based on completion of the 12-month follow-up. A response rate of 77% of the women was achieved over the 12-month period. The respondent sample is slightly biased toward program completers and Anglo subjects. The outcome findings are based on these women's reports.

Our initial hypothesis reflected a situational model of change: those cases with the least mental and substance use problems, attending the most counseling sessions, and receiving additional services would have the lowest reassault rates and better quality of life for the women. The preliminary findings suggest, however, that all the programs are associated with a short-term cessation of assault and improvements overall in the women's quality of life, irrespective of counseling duration and additional services. Unfortunately, the complexity of the respective intervention systems and the confounding of a quasi-control group (program dropouts) makes it difficult to determine the specific contribution of the program counseling to the outcome.

These preliminary findings remain tentative. The outcomes need to be controlled for a number of contingencies, the largely self-report data need to be verified with documentation, and the descriptive statistics are to be subjected to multivariate analyses. The preliminary data also need to be weighed against qualitative information and interviewee narratives, and more elaborate case studies, drawing on a variety sources, need to be constructed.

Program Structure

As part of the design implementation, the principal investigator conducted two visits to each site during the subject recruitment period. He attended staff meetings, conducted staff interviews, visited related agencies, reviewed program documentation and budgets, and observed group sessions. An inventory of program organization, structure, and approach was completed based on these sources of information (see "Program Structure Survey" in the Appendix). The objective of the inventory was to substantiate that the programs demonstrated the selection criteria, to identify any variations in program implementation over the course of the subject recruitment, and to describe any additional factors or issues that might influence program outcome. A major research finding emerged from this procedure: The programs varied much more than anticipated and more than our selection criteria specified (see Table 3). Our initial continuum of program length and extent of services was confounded by court linkages, responses to non-compliance, and community context. Batterer systems appear essentially unique to their community environments, social histories, organizational structures, available staffing, and available services. Replicating model programs may, therefore, be more complicated than some training approaches might suggest. Moreover, several factors emerged that help to further qualify and interpret program outcomes. These are summarized below:

  1. Significant variation exists among the intervention systems, before and after intake, which is generally not addressed by program standards or evaluation research. For instance, the pretrial, shorter-term program had an average of 2.5 weeks from court hearing to intake with no delay from intake to program participation. On the other hand, the post-conviction, longer-term program had as much as a 3-month delay on average from hearing to intake. In addition, time from arrest to court-hearing was only 2 weeks on average at the pre-trial, 3-month program, and weeks longer at the other sites.

  2. The dropout rate at 3 months was only 30% for the pre-trial, 3-month didactic program in Pittsburgh, but 40-45% for the 9-month and 5-month post-conviction programs. The majority of the dropout occurred within the first two months of the program; only an additional 5% dropped out of the longer programs after 3 months of program participation. (The pre-trial, 3-month program has a mandatory court review of compliance; at the other sites, non-compliance is addressed by probation offices.)

  3. Nearly a fifth (18%) of the program participants were attending voluntarily (not court-mandated). These men were more likely to have some college education, be white-collar workers, employed, and married. They were also likely to have symptoms associated with depressive and anxiety disorders, and less likely to have narcissistic or compulsive tendencies. Only 1/3 (34%) of the voluntary men showed tendencies toward "socially desirable" answers on the MCMI, as opposed to over 1/2 (55%) of the court-ordered men. The voluntary participants were twice as likely to drop out after intake or by three months, and were 10% more likely to reassault their partners.

Men's Characteristics

The demographics of the batterers vary across the four sites and reflect the compositions of their respective communities. However, within each program there are dichotomies in social circumstances (race, education, living with the partner, etc.) that are often not specifically addressed in program curriculum. For instance, a quarter of the men had not completed high school and may have difficulty reading or grasping some vocabulary, while 1/3 of the men had attended college. Surprisingly, the four programs had equivalent portions of men who had parental abuse while growing up, alcoholic tendencies, psychological problems, prior arrests, committed "severe" women battering. Despite the differences in court referral procedures and demographics, the distribution of these case characteristics was similar across the four sites.

Specific details about the demographics and characteristics of the batterers are as follows:

  1. Over half (55%) of the men were either African American (31%), Latino (18%), or from another racial minority (6%). The men of color were 10-15% more likely to be under 25, to not have a high school diploma, to be under-employed, or to be a blue-color worker. African American men were similar to the white men in severity of previous abuse, previous arrest, alcohol abuse, and most personality traits. They were, however, twice as likely to show evidence of severe Axis II disorders (25% vs. 11% for whites) and narcissistic tendencies (35% vs. 19% for whites). The African American men were 13% more likely than the white men to dropout by 3-months (48% vs. 35%), but reassaulted at the same rate as the white men.

  2. Over half (51%) of the men were not living with their partners at the time of intake, and that amount increased about 10% during the follow-up. There was no significant difference in the drop-out or reassault rates for those living apart or together at the time of intake.

  3. Over half (56%) of the men appear to have alcoholic tendencies according to the MAST, and about 25% show evidence of a severe mental disorder (15% on Axis I and 16% on Axis II), according to the MCMI. There is, however, little agreement between self-reports and validated tests for alcohol use and mental health, which raises questions about the optimal way to obtain such information.

  4. The personality test (MCMI) suggests a very diverse and complex picture of psychological problems that does not readily conform to previous typologies and profiles (e.g., "the abusive personality") and is less pathological than previous studies suggest. The most common personality traits of 11 categories were narcissistic (25% of the sample), passive-aggressive (24%), antisocial (19%), and depressive (19%). A typology based on factor-scores from the MCMI produced four groupings: "little pathology", "narcissistic/ antisocial", "avoidant/dependent" and "severe pathology". These batterer "types" did not predict reassault, although the severe pathology types were more likely to dropout.

  5. The only predictor for severity of the assault at the initial arrest was the severity of previous violence, irrespective of demographic and personality characteristics. A clustering of personality and demographic characteristics was also not predictive of the severity of violence at the initial arrest.

Women's Characteristics

A substantial portion of the initial partners suffered severe assaults and injury prior to the batterers' program intake. Over half had previously contacted the criminal justice system in response to abuse, but only about a quarter had received any counseling for domestic violence and less than 10% had previously visited a battered women's shelter. The women's perceptions of their batterers were overly optimistic, despite the severe abuse and information received from batterer programs. The partners of court-mandated batterers differ from battered women in shelters in terms of demographics, victimization and help-seeking, and appear to warrant special attention and research. The findings raise issues for what additionally might be done to better serve battered women of court-ordered batterers. The major findings about the women are summarized below:

  1. Three-fourths of the women reported being physically assaulted as part of the incident that led to their partner's initial arrest or program referral. Over 1/2 (53%) identified severe tactics being used against them (as defined by the Conflict Tactics Scale). Over 90% reported being assaulted sometime in the past; 70% had been bruised, 42% injured, 32% received some medical help, and 4% had been hospitalized as a result of past domestic violence.

  2. The women had relatively little shelter or counseling contact both before and after the initial arrest (less than 1/3 had shelter or counseling contact sometime in the past). Yet the majority (55%) of the women had used the criminal justice system sometime prior to the arrest leading to the batterers' court referral to the batterer program. A small portion of the women also had little contact with domestic violence services during the follow-up. This contact was likely to be in response to a reassault rather than a preventative addition to the batterer program intervention.

  3. A quarter of the women reported being heavy drinkers (at least weekly), and a fifth (18%) were frequently drunk (at least monthly) at the time of the men's program intake. This rate decreased after arrest and the number of women involved in alcohol treatment increased. Approximately 1/4 of the women were on welfare at the time of program intake, and this proportion remained relatively constant during the follow-up.

  4. The majority (66%) of women reported being physically aggressive toward their partners prior to the initial arrest, and 15% of the women had been arrested for domestic violence at the time of the man's initial arrest. During the 12-month follow-up, the vast majority of the women who were reassaulted were physically aggressive toward their batterers. Sixty percent of these women indicated that the man hit them first.

  5. The women appeared to be very optimistic about the prospects of batterer counseling. Two-thirds (64%) of the women said that their batterer had "admitted his problem," and 94% believed that he would complete the program despite substantial dropout rates. Almost half of the women reported feeling "very safe" (59%) at the time of program intake, and that it was "very unlikely" (42%) that they would be hit again in the near future.

Program Outcome

The "model" batterer programs appear to contribute to the cessation of assault at least in the short-term. The majority of women indicate their "quality of life" has improved and that they feel "very safe." A portion of batterers, however, appear to be resistant and unresponsive to intervention. They reassault soon after program intake, repeatedly reassault, and/or cause significant distress and fear in their partners. These findings lend support to the approach and linkages represented by the programs in our study, and to the emerging efforts to establish more intensive intervention for resistant batterers. However, we did not find the longer and more extensive programs to have lower reassault rates. The equivalent reassault rates over the four programs may be in part the result of other program features, court procedures, and community factors, discussed above, that offset the expected benefits of more comprehensive programs. The major findings about the program outcome are as follows:

  1. According to initial victims and new partners, 29% of the men reassaulted during the 12-month follow-up. The reassault rate varied only a few percent when adjustments were made for new partners, no partner contact, additional reports from men, and arrest records. Over half of the men (59%) who reassaulted a partner (17% of the total with responding partners) physically bruised or injured their partner, but only 12% of the men's partners (4% of the total respondents) sought medical attention for their injuries. Only a small portion "beat up" their partners (6%) or used a gun or knife (3%) in their reassault. Ten percent of the men at the Pittsburgh and Colorado sites were rearrested for domestic violence.

  2. Over half of the reassaults occurred within 3 months after program intake. Over a third (36%) of the men reassaulted a partner during more than one of the 3-month follow-up intervals, and 60% of the men (17% of the total with responding partners) committed more than one reassault during any one follow-up interval. Only 15% of the batterers' new partners (n=114) reported an assault during the follow-up, but the relationships with many of these women were still in their early stages.

  3. Nearly half (42%) of the men used controlling behaviors; 68% were verbally abusive; over one-third (40%) of the men threatened their partners, and 14% stalked them. The portion of women reporting these behaviors progressively decreased at each of the 3-month follow-up intervals.

  4. Nearly 3/4 of the women reported that their overall quality of life was "betterB since program intake, but 10% indicated their life was "worse off." The area of their life with the most problems was financial. Two-thirds reported feeling "very safe" over the course of the 12-month follow-up.

  5. The reassault rate at twelve months after intake was equivalent across the four sites. That is, the pre-trial, shorter-term, didactic program had the same outcome as the post-conviction, longer-term, process program. The program site, moreover, had no significant effect when demographic and case characteristics were controlled and interactions were introduced in a logistic regression.

  6. Program dropouts at intake were at most 5% more likely to reassault--a difference that is not statistically significant; but the dropouts by 3-months were approximately 15% more likely to reassault their partners (38% vs. 24%). (Dropouts at intake were likely to be jailed, in another kind of program, or not seeing their partner, and many of those living apart continued frequent contact with their partners.) Also, dropouts were twice as likely to be rearrested for domestic violence than those who had completed 3 months or more of a program. Program dropout remains a significant predictor for reassault when controlling for background variables, suggesting a "program effect" on the outcome.

  7. A severe previous assault and being drunk once a month are the only items at intake that significantly "predict" reassault. Other background information at program intake, including MAST and MCMI results, are not substantial predictors. Also, the batterer living with or apart from his partner does not appear to influence the reassault rates. Those men who were drunk, at least monthly during the follow-up, were, however, three times more likely to reassault during the 12-month follow-up.

  8. The reassault rates for the four "types" of batterers based on the MCMI (i.e., little pathology, narcissistic/antisocial, avoidant/dependent, and severe pathology) were equivalent, and did not significantly vary when controlling for didactic versus process-oriented programs. In other words, the narcissistic/antisocial type of batterers did not do better in didactic programs and the avoidant/dependent types did not do significantly better in the process programs, despite preliminary indications in previous research to the contrary.

Completed Papers from: A Multi-Site Evaluation of Batterer Intervention Systems, June 1997 Funded by the Centers for Disease Control Edward W. Gondolf, Principal Investigator

Research Design

Gondolf, E. (1997). Batterer programs: What we know and need to know. Journal of Interpersonal Violence, 12,3-98.

Gondolf, E. (in press). Expanding batterer program evaluations. In G. K. Kaufman & J. Jasinski (Eds.), Out of darkness: Contemporary research perspectives on family violence.Thousand Oaks, CA; Sage.

Gondolf, E., Yllo, K., & Campbell, J. (in press). Collaboration between researchers and advocates. In G. K. Kaufman & J. Jasinski (Eds.), Out of darkness: Contemporary research perspectives on family violence.Thousand Oaks, CA; Sage.

Heckert, A. & Gondolf, E. (1997). Agreement of Assault Among Batterer Program Participants and Their Partners.Paper presented at the 5th International Family Violence Conference, University of New Hampshire, Durham, NH, June 29-July 2.

Batterer Characteristics

Gondolf, E. (1996). Characteristics of batterers in a multi-site evaluation of batterer intervention systems.Report submitted to Centers for Disease Control, U.S. Dept. of Health and Human Services.

Gondolf, E. (under review). Characteristics of court-mandated batterers in four cities: Diversity and dichotomies.Criminal Justice and Behavior.

Gondolf, E. (under review). MCMI results for Batterer Program Participants in Four Cities: Less "pathological" than expected. Journal of Personality Disorders.

Gondolf, E. (under review). Batterer types based on the MCMI: A less than promising picture. Journal of Consulting and Clinical Psychology.

Victim Characteristics

Gondolf, E. (under review). Victims of court-mandated batterers: A different kind of battered woman? Violence Against Women.

McFarlane, J., & Gondolf, E. (under review). Characteristics of abuse and resource use: A comparison of partners of court-mandated batterers and pregnant abused women (Brief Report). Journal of Epidemiology and Community Health.

Coben, J., Forjuoh, S., & Gondolf, E. (under review). Injuries and health care use by women with partners in batterer intervention programs. Academic Emergency Medicine.

Forjuoh, S., Coben, J., & Gondolf, E. (under review). Risk factors for injury in battered women: A case-control analysis. Journal of the American Medical Association.

Program Outcomes

Daley, J. (1996). Predicting compliance among men who batter: The contributions of demographics, violence-related factors, and psychopathology.Dissertation submitted to Department of Psychology, University of Houston, Houston, Texas.

Gondolf, E. (under review). Patterns of reassault in batterer programs: Who, what, when, and why. Violence and Victims.

Gondolf, E. (under review). A comparison of reassault rates in four batterer programs: Do court referral, program length and services matter? Journal of Interpersonal Violence

Gondolf, E. (under review). The failure to predict reassault from batterer program intake. Journal of Family Violence.

Snow, A. & Gondolf, E. (1997). Post-program predictors of Re-assault for Batterer Program Participants.Paper presented at the 5th International Family Violence Conference, University of New Hampshire, Durham, NH, June 29-July 2.

Papers in Progress on Multi-site Evaluation of Batterer Intervention Systems funded by Centers for Disease Control

There are several other papers with more complex analyses in progress. These should be completed by the end of the summer 1997.

  • Robert White, a clinical psychologist at Indiana University of Pennsylvania, is conducting a case-based assessment of men in our sample using MCMI profiles and other background and follow-up information.

  • Alex Heckert, a family sociologist at Indiana University of Pennsylvania, is examining the reliability of men and women's reports of reassault during the follow-up, and also comparing the women's accounts of arrest incidents with corresponding police reports.

  • Alison Snow Jones, a health economist at Johns Hopkins Medical Center, is computing a time series analysis of reassaults using a variety of situational and helpseeking variables from the follow-up.

  • Jeff Coben, a physician and injury researcher, and his staff at the Center for Injury Control, University of Pittsburgh Medical Center, are reviewing medical records of the battered women and the medical assistance they obtained during follow-up.

  • Ron Laporte, an epidemiologist from the Graduate School of Public Health, University of Pittsburgh, is developing a capture-recapture analysis of the outcomes for the 15-month follow-up.

  • Robert Ackerman, an alcohol specialist at MAATI, and Edward Gondolf of MAATI will complete a paper on the assessment of alcohol use which contrasts the results of the MCMI alcohol dependence subscale, MAST, men's self-reports, and women's reports about the men's alcohol use.

  • Edward Gondolf will complete a report on the diversity of program organization, structure, and linkages at the four research sites in our study.

Multi-site Evaluation of Batterer Intervention Systems funded by Center of Disease Control, 1994-1997

Investigators

  • Edward Gondolf, Principal Investigator,Mid-Atlantic Addiction Training Institute

  • Jeff Coben, Co-Investigator, Allegheny Univ. of the Health Sciences

  • Ron LaPorte, Co-investigator, University of Pittsburgh

  • Edmund Ricci, Co-invesitgator,University of Pittsburgh

Analysts

  • Alison Snow Jones, Statistical Analyst, Johns Hopkins University

  • Alex Heckert, Statistical Analyst, Indiana University of Pennsylvania

  • Samuel Forjuoh, Statistical Analyst, Allegheny Univ. of the Health Sciences

  • Yue-Fang Chang, Statistical Analyst, University of Pittsburgh

Program Directors

  • Robert Foster, DACC, Pittsburgh

  • Paul Burchfield, DACC, Pittsburgh

  • Toby Myers, Director, PIVOT, Houston

  • Robert Gallup, AMEND, Denver

  • Diane McGauley, Family Place, Dallas

  • Sherry Lundberg, Family Place, Dallas

  • Debra Cosimo, Family Place, Dallas

Consultants

  • Barbara Hart, Battered Women' Justice Project

  • Oliver Williams, Univ. of Minnesota

  • Ellen Fisher, Texas Council on Family Violence

  • Ellen Pence, Domestic Abuse Intervention Proj.

  • Estella De Anda, Mujeres Unidas, McAllen, TX

  • Edward Mulvey, University of Pittsburgh

Assisting Organizations

  • Penna. Coalition Against Domestic Violence

  • Texas Council on Family Violence

  • Colorado Coalition Against Domestic Violence

  • Center for Injury Research and Control,

  • University of Pittsburgh

  • Mid-Atlantic Addiction Training Institute,

  • Indiana University of Pennsylvania

  • Centers for Injury Control, DHHS

Administration

  • Jewel Lee Doherty, Project Coor.

  • Crystal Deemer, Admin. Asst.

  • Robert Ackerman, MAATI Director

  • Lois Brown, Finance Office

Research Assistants

  • Robert Gregory

  • Gayle Moyer

  • Mimi Prada

  • Steven Cohen

  • Dan Schaeffer

  • Lisa Watson

  • Lanette Moloney

  • Steven Olish

  • Ken Guerino

  • Mike McSkimming

  • Sheree Rogers

  • Jim Cassels

  • Vincent Miles

Site Researchers

  • Don Boziich, DACC

  • Jennifer Daly, PIVOT

  • Jody Scruggs, Family Place

  • Gary Gibbons, AMEND

Graduate Assistants

  • Bob White, Clinical Consultant

  • Danielle Matula

  • Patty Pierson

  • Neth Chanthavisoik

  • Sean McCaskill

  • Rosalind Scholfield

  • Julia Nunez

Data Managers

  • Randy Hutzler

  • Marc States

  • Loretta Fouse

  • Cori Bosack

  • David Somola

  • Neil Fulton

Coding/Data Entry and Tracking/General

  • Heather Bosley

  • Michelle Shandick

  • Ruth Ann Yanoschick

  • Linda Thompson

  • Melanie Shandick

  • Cybil Peoples

  • Alicia Dohn

  • Kyle Bernard

  • Krista Erceg

  • Adel Lesniak

  • Kimberly McMillan

  • Erin Ziemba

  • Joann Hooper

  • Jim Sekoch

  • Jessica Loughrey

  • Marty Pilgrim

Summary of the 15-months Follow-up

Introduction

One of the most pressing questions for policy makers and practitioners in the domestic violence field is whether counseling programs for men who assault their female partners---or "batterer programs"---are effective in preventing further assaults on women. A second question is also being vigorously debated. That is, what is the most effective referral procedure, program length, and program format? For instance, are 9-month programs, that are mainly discussion sessions for men who have been convicted for assaulting their female partners, more effective than 3-month programs, that are mainly instructional classes for men sent to the program as part of a pre-trial diversion. A third emerging question is whether different types of batterers warrant specialized treatment or intervention. Related to the question of "batterer types" is whether we can identify "risk markers" or predictors of reassault that would help identify the potential reassaulters.

Our initial expectation, based on previous research and practitioner experience, was that batterer programs do help interrupt the violence for the majority of men referred to them. We would also expect the longer and more comprehensive programs to have lower rates of reassault, especially given the social reinforcement for abuse and compounding problems of many batterers (e.g., alcohol abuse). Finally, we might expect "antisocial" men and men with more severe psychopathology to be the most likely to reassault. We posed a situational model of reassault, however, which suggests that contact with the woman, additional services or intervention, unemployment, and further drinking would influence the outcome.

Method

Four batterer intervention systems were evaluated using a naturalistic, comparative research design, as opposed to a clinical trial or experimental design.Over 30 program evaluations have been published in recent years, but most of these are limited by their single-site design and low-response rates. Few studies comparing programs or program approaches have been conducted, and there remain some difficult questions about the ethics, feasibility, and reliability of experimental or "clinical trials" in this field.

The four research sites were selected to represent a continuum of intensive and extensive intervention.The sites include 1) a pre-trial, 3-month, didactic program with court liaisons (Pittsburgh), 2) a post-conviction, 3-month, process program with women's services (Dallas), 3) a post-conviction, 5-month, didactic program with legal advocacy (Houston), and 4) a post-conviction, 9-month, process program with complementary services (including substance abuse treatment, individual mental health counseling, and women's services coordinators) (Denver). A background questionnaire, an alcohol test (MAST), and a personality test (MCMI) were administered to 210 men at each of the four sites for a total sample of 840 (82% of the sample were court-ordered as opposed to being "voluntary" participants.) The men's partners were also interviewed by phone at the time of intake.

The men and their battered partners (and 170 identified new partners) were interviewed by phone every 3-months for 15 months.Police arrest reports, program records and participation reports, rearrest records during follow-up, and women's medical records were reviewed and analyzed to help verify self-reports. Site-visits, which included staff interviews and observations of group sessions, were conducted to verify the program approach and system.

A female partner was interviewed for 79% of the batterers at least once during the 15-month follow-up,and we were able to account for 67% of the sample for the full 15-month follow-up. The women respondents are slightly biased toward program completers and Anglo subjects. A similar response rate was achieved for the men. In 64% of the cases, both the men and their partners were interviewed. The outcome findings are based on women's reports, unless otherwise indicated.

Major Findings

Reassault and Other Outcomes:The reassault rate for the participants in the four programs falls between 32% and 39%.Nearly one-third (32%) of the men reassaulted a female partner according to the women's reports. Adjusting the women's reports with available reports of their male partners increases the reassault rate to 33%, and adjusting for male reports and arrest records increases the rate to 36%. A capture-recapture analysis of all available men, women, and arrest reports projects a reassault rate of 39% for the entire sample.

A much lower percentage of women were actually physically injured during the follow-up.Over half of the men (60%) who reassaulted a partner (19% of the total with responding partners) physically bruised or injured their partner, but only 20% of the injured partners (4% of the total respondents) sought medical attention for their injuries.

A surprising number of the reassaults occur shortly after program intake.Nearly half (44%) of the men who reassaulted a partner did so within 3-months of program intake. The total percentage of new reassaults drops from 14% at the 3-month follow-up to 3% at the 9-month follow-up and continues at that level through the remainder of the follow-up.

A small portion of men appear unresponsive to the intervention and severely abusive.Over half (59%) of the men who reassaulted (19% of the total with responding partners) committed more than one reassault, and a third (32%) of the men who reassaulted a partner did so during more than one of the 3-month follow-up intervals.

A higher percentage of women reported other forms of abuse.Seventy percent of the men were verbally abusive; nearly half (45%) used controlling behaviors; over one-third (43%) of the men threatened their partners, and 16% stalked them. The arrest records indicate that only 10% of the batterers had been rearrested for domestic violence, 14% had been arrested for some other violence, and 16% had been arrested for some other offense during the follow-up period (40% had been arrested for some offense during the follow-up).

Most women, however, feel their circumstances have improved as a result of the batterer intervention.Two thirds (66%) of the women indicated at the 15-month follow-up that they were "better off" overall than when their partner was sent to the batterers program, but 12% indicated they were "worse off." The area of the least improvement was financial. Almost three-fourths (72%) of the women reported feeling "very safe" at each follow-up interval.

Program Comparisons and Batterer Types:The program participation appears to contribute to reducing reassault.Program dropouts (35% of the respondent sample) were approximately 13% more likely to have reassaulted their partners (40% vs. 28%; X2[1]=10.59; p is less than .001). The dropouts were also more likely to have been rearrested for domestic violence (20% vs. 5%). Program attendance remained a significant predictor when controlling for background factors in a logistic regression for reassault. Moreover, the voluntary participants were significantly more likely to dropout of the programs and more likely to reassault than the court-ordered men.

There was no significant difference in the outcomes of the four different programs(i.e., reassault rate, men making threats, domestic violence rearrests, and victim "quality of life"). The longer, more comprehensive program did have a significantly lower rate of "severe" reassault; however, this apparent "site effect" disappears when background variables (i.e., demographics, personality, behavioral, relationship status variables) are controlled in a logistic regression and "survival" analysis. Also, separate subscales for antisocial, borderline, or avoidant tendencies were not of themselves predictive of reassault.

The outcomes also did not significantly vary for different personality types.A factor analysis of scores on the MCMI subscales was used to generate four types that reflect the prevailing personality types for batterers. The types might be characterized as 1) little psychopathology, 2) antisocial/narcissistic, 3) avoidant/dependent, and 4) severe pathology. Also, there was no evidence that one "type" of batterer did better in one program approach or another (i.e., didactic vs. process).

The only substantial predictor of outcome was a man being drunk during the follow-up.General estimating equations were used to test a situational model of reassault. This analysis entered a variety of background variables and intervening situational variables (e.g. unemployment of the man or woman, additional services for the batterer, victim services for the woman) for the longitudinal data (each of the 3-month follow-up intervals). Men who were reportedly "drunk" during the follow-up were 3 times more likely to reassault than those who were not drunk. Behavioral and personality factors assessed at program intake and the situational factors during follow-up were not significant or substantial predictors.

Conclusions

Summary:Our evaluation of four batterer programs produced some unexpected findings. The reassault rate was comparable to that reported in previous single-site evaluations, and may be a substantial accomplishment given the personalities problems, criminal history, and substance abuse among substantial portions of the men. Moreover, there does appear to be a "program effect" in that the program completers are significantly less likely to reassault than the dropouts even when controlling for background factors. The severest reassault is committed by those who reassault early and repeatedly during the follow-up. Unfortunately, we can not readily identify these more "dangerous" men or the reoffenders in general.

The rate of new reassaults during the follow-up decreased substantially rather than increased as deterrence theories might suggest. We may find that this rate of reassault increases beyond a threshold of a year and a half, and after the chronic offenders (i.e., those who reassault early and repeatedly) are removed.

The most surprising finding is that the shorter programs appear to be as "effective" as the longer, more comprehensive programs. The benefits of longer programs may not become evident until after our relatively short 15-month follow-up. The longer programs may, furthermore, be compromised by other features in their communities, such as the slow response of the police to men who dropout, or the delays in some courts between arrest and referral to a batterer program. Some programs may be unique to their communities. That is, they reflect the resources, personnel, legal procedures, and expectations of their particular locale.

Our findings also reinforce how difficult it is to predict reassault among clinical populations. Different personality types did not have worse outcomes than others and did not appear to do better in one program or another. Additionally, behavioral characteristics assessed at intake and situational factors during follow-up did not substantially predict outcome. Only "drunkenness" during follow-up was significantly and substantially related to program outcome. This finding does not necessarily mean that alcohol abuse is the "cause" of reassaults. "Drunkeness" may be an indicator of unresponsive and defiant men.

Program Implications:The programmatic implications of our findings remain tentative without a longer follow-up and additional confirming research. The 15-month follow-up does suggest that program format in terms of length, court-linkage, and additional services warrants more consideration. Shorter programs with essential linkages and curriculum may be as effective as longer programs faced with court delays and uncertain sanctions for dropout or reassault. Additionally, more intensive (e.g., 3-4 days a week for 3 months) rather than extensive (e.g., once a week for 9 months) programming might be warranted for the potential early and repeat assaulters identified in our study. We did not, however, find evidence for specialized programming to address different personality types, beyond the prevailing gender-based cognitive-behavioral approach. There may be a need to more carefully monitor alcohol abuse after the program and sustain alcohol treatment for some men over the long term.

Future Research:We are continuing our investigation into a number of other topics related to the batterers and victims in our study. These topics include several studies about the characteristics of the batterers and their victims, including the reliability of alcohol assessments, the reliability and validity of abuse self-reports, the structure and topics of incident narratives, and the women's use of services. We are also examining more about the programs and their outcomes in topics, such as, the variations in program structure and linkages, the influence of race on program participation and outcome, the outcome of "voluntary" participants as opposed to court-mandated participants, predictors of injury and medical assistance, predictors of program dropout, the perceptions of sanctions and outcome, and violence avoidance techniques used by batterers.

We are also beginning an additional 3-year grant that will extend our 15-month follow-up to 4-years, and reveal the long-term impact of the programs for the initial and new partners of the men. This new research will additionally explore the reasons for the relatively small portion of women who receive additional services or support. This research will, furthermore, examine the effect of the new welfare reform on the relationships between the men and women in our initial study. The overall purpose of the new research is to confirm or revise our current findings based on a relatively short-term follow-up.

Completed and Proposed Papers

Completed Papers

Research Design

Gondolf, E. (1997). Batterer programs: What we know and need to know. Journal of Interpersonal Violence, 12,83-98.

Gondolf, E. (1997). Expanding batterer program evaluations. In G. K. Kaufman & J. Jasinski (Eds.), Out of darkness: Contemporary research perspectives on family violence(pp. 208-218). Thousand Oaks, CA; Sage.

Gondolf, E., Yllo, K., & Campbell, J. (1997). Collaboration between researchers and advocates. In G. K. Kaufman & J. Jasinski (Eds.), Out of darkness: Contemporary research perspectives on family violence(pp. 255-261). Thousand Oaks, CA; Sage.

Heckert, A. & Gondolf, E. (1997). Agreement of assault among batterer program participants and their partners.Paper presented at the 5th International Family Violence Conference, University of New Hampshire, Durham, NH, June 29-July 2 (submitted to Journal of Interpersonal Violence).

Heckert, A. & Gondolf, E. (1997). Predictors of batterer-partner disagreement on assaults. Journal of Family Violence(under review).

Myers, T., & Daly, J. (1997). A battering intervention project's reaction to participation in a multi-site program evaluation.Paper presented at the 5th International Family Violence Conference, University of New Hampshire, Durham, NH, June 29-July 2.

Matula, D. (1997). Women's accounts of domestic violence: A contextual "story" of life.Masters thesis, Department of Sociology, Indiana University of Pennsylvania, Indiana, PA.

States, M. (1997). The utility of logistic regression in predicting reassault among batterers.Masters thesis, Department of Mathematics and Statistics, Indiana University of Pennsylvania, Indiana, PA.

Batterer Characteristics

Gondolf, E. (1996). Characteristics of court-mandated batterers in four cities: Diversity and dichotomies.Paper presented at the Annual Meeting of the American Society of Criminology, Chicago, IL, November, 20-23 (submitted for publication to Journal of Family Violence).

Gondolf, E. (under review). MCMI results for Batterer Program Participants in Four Cities: Less "pathological" than expected. Journal of Family Violence..

Gondolf, E. (under review). Batterer types based on the MCMI: A less than promising picture. Violence Against Women.

Gondolf, E. & White, R. (under review). A clinical-based assessment of Batterer MCMI-III prfiles: A diluting of distinct types. Journal of Family Psychology.

White, R., & Gondolf, E. (under review). The implication of MCMI profiles for batterer treatment.: Support for the gender-based, cognitive-behavioral approach. Journal of Interpersonal Violence.

Victim Characteristics

Gondolf, E. (in press). Victims of court-mandated batterers: A different kind of battered woman? Violence Against Women.

Coben, J., Forjuoh, S., & Gondolf, E. (in press). Injuries and health care use by women with partners in batterer intervention programs. Journal of Family Violence.

Forjuoh, S., Coben, J., & Gondolf, E. (under review). Risk factors for injury in battered women: A case-control analysis. Women & Health.

McFarlane, J., & Gondolf, E. (1988). Preventing abuse during pregnancy: A clinical protocol (includes insert on "Abuse and assistance patterns of pregnant and non-pregnant women"). American Journal of Maternal & Child Nursing,23, 22-26.

Program Outcomes

Gondolf, E. (April, 1997). Multi-site evaluation of batterer intervention systems: A summary of findings for a 12-month follow-up.Report submitted to Centers for Disease Control (CDC), Atlanta, GA.

Gondolf, E. (November, 1997). Multi-site evaluation of batterer intervention systems: Summary of a 15-month follow-up.Report submitted to Centers for Disease Control (CDC), Atlanta, GA.

Daley, J. (1996). Predicting compliance among men who batter: The contributions of demographics, violence-related factors, and psychopathology.Dissertation submitted to Department of Psychology, University of Houston, Houston, Texas.

Gondolf, E. (in press). Patterns of reassault in batterer programs. Violence and Victims.

Gondolf, E. (under review). A comparison of reassault rates in four batterer programs: Do court referral, program length and services matter? Journal of Interpersonal Violence

Jones, A. & Gondolf, E. (1997). Post-program predictors of Re-assault for Batterer Program Participants.Paper presented at the 5th International Family Violence Conference, University of New Hampshire, Durham, NH, June 29-July 2 (submitted to American Journal of Public Health).

Gondolf, E., Chang, Y., & Laporte, R. (under review). Capture-recapture analysis of batterer reassaults: An epidemiological innovation for batterer program evaluation. Violence and Victims.

Papers in Progress/Proposed

Papers in Progress

  • Edward Gondolf will complete a report on the diversity of program organization, structure, and linkages at the four research sites in our study.

  • Jeff Coben, a physician and injury researcher, and his staff at the Center for Violence and Injury Control, Allegheny University of the Health Sciences, are reviewing medical records of the battered women and the medical assistance they obtained during follow-up.

  • Robert Ackerman, an alcohol specialist at MAATI, and Edward Gondolf of MAATI will complete a paper on the assessment of alcohol use which contrasts the results of the MCMI alcohol dependence subscale, MAST, men's self-reports, and women's reports about the men's alcohol use.

  • Edward Gondolf with Ed Ricci, School of Public Health, University of Pittsburgh, are working on three methodology papers: a protocol for the safety of battered women, tracking and follow-up procedures for batterer program evaluation, and procedural and conceptual issues in batterer program evaluation.

  • Alison Snow Jones, Johns Hopkins University, is developing a longitudinal analysis of helpseeking and service use of the female partners of batterers.

  • Edward Gondolf and Robert White, a clinical psychologist at Indiana University of Pennsylvania, are studying the distinguishing characteristics of men who repeatedly reassault their partners during follow-up.

Papers Proposed for 1998

  • Types and predictors of dropout in batterers programs--Ed Gondolf, MAATI

  • Domestic violence incidence according to Police Reports--Alex Heckert, IUP

  • Decision-making factors in assessing reassault among batterers--Alex Heckert, IUP

  • Techniques used for avoiding violence by batterers--Bob White, IUP

  • Batterer program recommendations from batterers and battered women--Bob White, IUP

  • Racial differences among batterer program participants--Oliver Williams, Un. of Minn.

  • The effect of perceptions of sanctions on batterer program outcome--Martin Schwartz, Un. of Ohio

Investigators

  • Edward Gondolf, Principal Investigator Mid-Atlantic Addiction Training Institute

  • Jeff Coben, Co-Investigator Allegheny Univ. of the Health Sciences

  • Ron LaPorte, Co-investigator University of Pittsburgh

  • Edmund Ricci, Co-invesitgator University of Pittsburgh

Analysts

  • Alison Snow Jones, Statistical Analyst Johns Hopkins University

  • Alex Heckert, Statistical Analyst Indiana University of Pennsylvania

  • Samuel Forjuoh, Statistical Analyst Allegheny Univ. of the Health Sciences

  • Yue-Fang Chang, Statistical Analyst University of Pittsburgh

Program Directors

  • Robert Foster, DACC, Pittsburgh

  • Paul Burchfield, DACC, Pittsburgh

  • Toby Myers, Director, PIVOT, Houston

  • Robert Gallup, AMEND, Denver

  • Diane McGauley, Family Place, Dallas

  • Sherry Lundberg, Family Place, Dallas

  • Debra Cosimo, Family Place, Dallas

Consultants

  • Barbara Hart, Battered Women's Justice Project

  • Oliver Williams, Univ. of Minnesota

  • Ellen Fisher, Texas Council on Family Violence

  • Ellen Pence, Domestic Abuse Intervention Proj.

  • Estella De Anda, Mujeres Unidas, McAllen, TX

  • Edward Mulvey, University of Pittsburgh

Assisting Organizations

  • Penna. Coalition Against Domestic Violence

  • Texas Council on Family Violence

  • Colorado Coalition Against Domestic Violence

  • Center for Injury Research and Control,

  • University of Pittsburgh

  • Mid-Atlantic Addiction Training Institute, Indiana University of Pennsylvania

  • Centers for Injury Control, DHHS

Administration

  • Jewel Lee Doherty, Project Coor.

  • Crystal Deemer, Admin. Asst.

  • Robert Ackerman, MAATI Director

  • Lois Brown, Finance Office

Research Assistants

  • Robert Gregory

  • Gayle Moyer

  • Mimi Prada

  • Steven Cohen

  • Dan Schaeffer

  • Lisa Watson

  • Lanette Moloney

  • Steven Olish

  • Ken Guerino

  • Mike McSkimming

  • Sheree Rogers

  • Jim Cassels

  • Vincent Miles

Site Researchers

  • Don Boziich, DACC

  • Jennifer Daly, PIVOT

  • Jody Scruggs, Family Place

  • Gary Gibbons, AMEND

Graduate Assistants

  • Bob White, Clinical Consultant

  • Danielle Matula

  • Patty Pierson

  • Neth Chanthavisoik

  • Sean McCaskill

  • Rosalind Scholfield

  • Julia Nunez

Data Managers

  • Randy Hutzler

  • Marc States

  • Loretta Fouse

  • Cori Bosack

  • David Somola

  • Neil Fulton

Coding/Data Entry and Tracking/General

  • Heather Bosley

  • Michelle Shandick

  • Ruth Ann Yanoschick

  • Linda Thompson

  • Melanie Shandick

  • Cybil Peoples

  • Alicia Dohn

  • Kyle Bernard

  • Krista Erceg

  • Adel Lesniak

  • Kimberly McMillan

  • Erin Ziemba

  • Joann Hooper

  • Jim Sekoch

  • Jessica Loughrey

  • Marty Pilgrim

References

Beasley, R., & Stoltenberg. (1992). Personality characteristics of male spouse abusers. Professional Psychology: Research and Practice, 23,310-317.

Brown, P., & O'Leary, D. (1995). Marital treatment for wife abuse: A review and evaluation.Paper presented at the 4th International Family Violence Research Conference, July 21-24, 1995

Dutton, D. (1986). The outcome of court-mandated treatment for wife assault: A quasi-experimental evaluation. Violence and Victims, 1,163-175.

Dutton, D. (1995). Trauma symptoms and PTSD-like profiles in perpetrators of intimate abuse. Journal of Traumatic Stress, 8,299-316.

Dutton, D., & Starzomski, A. (1993). Borderline personality in perpetrators of psychological and physical abuse. Violence and Victims, 8,327-337.

Edleson, J. L., & Grusznski, R. (1989). Treating men who batter: Four years of outcome data from the Domestic Abuse Project. Journal of Social Service Research, 12,3-22.

Edleson, J., & Brygger, M.P. (1986). Gender differences in reporting of battering incidences. Family Relations, 35,377-382.

Edleson, J., & Syers M. (1990). The relative effectiveness of group treatments for men who batter. Social Work Research and Abstracts, 26,10-17.

Edleson, J., & Tolman, R. (1993). Intervention for men who batter: An ecological approach.Newbury Park, CA: Sage.

Eisikovits, Z., & Edleson, J. (1989). Intervening with men who batter: A critical review of the literature. Social Service Review, 37,85-414.

Fagan, J. (1989). Cessation of family violence: Deterrence and dissuasion. In L. Ohlin & M. Tonry (Eds.), Family violence.Chicago: University of Chicago Press.

Gondolf, E.W. (1988). Who are those guys? Towards a behavioral typology of men who batter. Violence and Victims, 3,187-203.

Gondolf, E.W., Fisher, E.R., & McFerron, R. (1991). Racial differences among shelter residents: A comparison of Anglo, Black, and Hispanic battered women. In R. Hampton (Ed.), Black Family Violence: Current Research and Theory.Thousand Oaks, CA: Sage.

Gondolf, E.W., & Foster, R.A. (1991). Preprogram attrition in batterer programs. Journal of Family Violence, 6,337-350.

Grusznski, R.J., & Carrillo, T.P. (1988). Who completes batterer's treatment groups? An empirical investigation. Journal of Family Violence, 3,141-150.

Hamberger, L.K., & Hastings,, J. E. (1988). Skills training for treatment of spouse abusers: An outcome study. Journal of Family Violence, 3,121-130.

Hamberger, L.K., & Hastings, J.E. (1989). Counseling Male Spouse Abusers: Characteristics of Treatment Completers and Dropouts. Violence and Victims, 4,275-286.

Hamberger, L.K., & Hastings, J.E. (1991). Personality correlates of men who batter and nonviolent men: Some continuities and discontinuities. Journal of Family Violence, 6,131-148.

Harrell, A. (1991). Evaluation of court-ordered treatment for domestic offenders.Washington, DC: The Urban Institute.

Hart, S., Dutton, D., & Newlove. T. (1993). The prevalence of personality disorders among wife assaulters. Journal of Personality Disorders, 7,329-341.

Holtzworth-Munroe, A., & Stuart, G. (1994). Typologies of male batterers: Three subtypes and the differences among them. Psychological Bulletin, 116,476-497.

Hutchison, I., Hirschel, J., & Pesackis, C. (1994). Family violence and police utilization. Violence and Victims, 9,299-313.

Kivel, P. (1992). Men's work: How to stop the violence that tears our lives apart.Center City, MN: Hazelden.

Pence, E., & Paymar, M. (1993). Education groups for men who batter: The Duluth model.New York: Springer.

Roberts. A. (1987). Psychosocial characteristics of batterers: A study of 234 men charged with domestic violence offenses, 2,81-93.

Russell, M.N. (1995). Confronting abusive beliefs: Group treatment for abusive men.Thousand Oaks, CA: Sage.

Saunders, D.G. (1992). A typology of men who batter: Three types derived from cluster analysis. American Journal of Orthopsychiatry, 62,264-275.

Saunders, D.G., & Parker, J.C. (1989). Legal sanctions and treatment follow-through among men who batter: A multivariate analysis. Social Work Research and Abstracts, 25,21-29.

Selzer, M. (1971). The Michigan Alcoholism Screening Test (Mast): The quest for a new diagnostic instrument. American Journal of Psychiatry, 3,176-181.

Stordeur, R., & Stille, R. (1989). Ending men's violence against their partners.Thousand Oaks, CA: Sage.

Straus, M. (1979). Measuring intrafamily conflict and violence: The Conflict Tactics (CT) scales. Journal of Marriage and the Family, 41,75-88.

Tolman, R.M., & Bennett, L.W. (1990). A review of quantitative research on men who batter. Journal of Interpersonal Violence, 5,87-118.