Predicting Cessation From Intimate Assaultiveness (PDF)

Predicting Cessation of Intimate Assaultiveness after Group Treatment Mark Bodnarchuk, M.A. & P. Randall Kropp, Ph.D. B.C. Institute on Family Violence James R. P. Ogloff, Ph.D. & Stephen D. Hart, Ph.D. Simon Fraser University Donald G. Dutton, Ph.D. University of British Columbia Project #4887-10-91-106 Submitted to Family Violence Prevention Division Health Canada 1995 The B.C. Institute on Family Violence Suite 290, West Cordova Street, Vancouver, B.C. Canada, V6B 1G1 Tel: 604/669-7055 Fax: 604/669-7054 Acknowledgements Funding for this project was provided by the Family Violence Prevention Division, Health Canada, but the views expressed in this report are those of the authors. We are grateful to Health Canada, Family Violence Prevention Division, for its support. We also thank the following organizations and individuals for their contributions: the men and women who participated in this research, the Victoria Family Violence Project, the Vancouver Assaultive Husbands Program, the provincial Corrections Branch of the Ministry of the Attorney General, and the Royal Canadian Mounted Police. Contents may not be commercially reproduced, but any other reproduction, with acknowledgements, is encouraged. Summary Previous treatment outcome research has been criticized for using inadequate follow-up periods and for failing to focus on individual characteristics predicting treatment responsiveness. This study attempted to address these criticisms. In the first part, a lengthy follow-up of up to eleven years was conducted on 446 men referred for spousal assault treatment. In the second part, a close examination of the relationship between personality variables and treatment outcome is reported. Part 1: Criminal history follow-up Criminal records were obtained for 446 men referred to the Assaultive Husbands Program in Vancouver, British Columbia. The men were divided into four different participation groups: Treatment Completers, Noncompleters (Drop outs), No shows, and Rejects. Completers were compared to the other three groups on a number of criminal recidivism variables. There was little evidence to support a relationship between treatment completion and follow-up “success”. Completers had a tendency to commit fewer crimes at follow-up, but these differences were not statistically significant. In general, the study found that the best predictor of future arrests and convictions in spousal assaulters is past criminal activity. The limitations of this study, including the reliance on criminal justice statistics, the focus on “statistical significance”, and the lack of a true randomized design are discussed. Part 2: Personality characteristics and treatment outcome A small sub-sample of men were followed-up with considerable scrutiny focusing on men’s and their partners’ reports of physical and emotional abuse following treatment. Subjects had completed treatment in one of two treatment programs in British Columbia. Both men and their partners reported substantially less physical and emotional abuse at follow-up. The findings from this small sample were considered reliable since statistical analyses confirmed that the sample was representative of a much larger population of treated men. Importantly, the study evaluated the effect of several personality variables on treatment outcome. Three indices of personality disorder -- borderline personality, avoidant personality, and antisocial personality -- accounted for some amount of variance in follow-up reports of violence and emotional abuse. These three personality profiles tend to predict poorer treatment outcome. The implications of these findings for treatment providers are discussed. Table of Contents Topic Introduction Method Results Discussion References Page 1 24 36 64 72 Introduction Several evaluations of "treatment success" have been reported in the last decade evaluating diminution of intimate assaultiveness after group treatment. The focus of these evaluations has been on the question of whether the groups were successful in diminishing post-treatment violence. Considerable debate has been generated about the results of these studies, and their interpretation (e.g. Burns, Meredith & Paquette, 1991; Rosenfeld 1992; Hamberger & Hastings, 1993; Dutton 1995). No consensus has emerged about what constitutes a realistic level of expectation against which to gauge group treatment success. The present evaluation departs from past work in two potentially important ways. First, it includes up to an eleven year follow-up of treated and untreated men in order to establish potential long term between-group differences in recidivist intimate violence. In so doing, it discusses a new model for evaluating treatment based on empirical studies of change from other forms of addictive behavior (Prochaska, DiClemente & Norcross 1993). This model suggests that expectations of immediate change from chronic dysfunctional behavior may be unrealistic. Instead, a five stage spiral model of change is derived and applied to the problem of intimate violence. Secondly, past evaluations have done little to address the question of who benefits most and least from treatment. At the same time, considerable evidence now points toward the existence of high rates of personality disorders in assaultive populations (Hamberger & Hastings, 1988; Hart, Newlove & Dutton 1993; Dutton & Starzomski, 1994). The question of whether certain forms of personality disorder may mitigate positive effects of therapy is raised by these findings. The current study will attempt to answer this question. Additionally, study findings also shed light on who benefits the most from treatment. The treatment programs We begin this report with a brief history and description of the therapeutic rationale of the groups that supply client populations for this study. The treatment programs which supplied clients for this study are the Assaultive Husbands Program in Vancouver, B.C. and the Victoria Family Violence Project in Victoria, B.C. Both programmes use a cognitive-behavioural model augmented by confrontational exercises about gender attitudes and guided examinations of early learning about the use of violence (“family of origin work”). Victoria Family Violence Project clients are men with repetitive patterns of abusive and controlling behaviours towards female partners, including physical abuse. In this program, violence is conceptualized as a learned response to strong emotions such as fear, frustration and rage, that results in men getting their way. Two thirds of clients are referred by a wide range of community services; many are self-referred and approximately one-third have been charged with wife assault. Men must be willing to admit to having been abusive and commit to change to be eligible to join the treatment program. Clients often have long histories of intimate abusiveness and violence. The program consists of three phases. Phase one is a 12 week treatment group focused on the prevention of men’s violence and abuse toward female partners, led by a recovering man who is a trained lay leader and a professional woman counsellor. Men may recontract for six weeks in phase one based on the advice of the group. All recovering lay leaders continue to attend their own therapy and support group. Phase two is a minimum of 24 sessions, with the goal of continued recovery through exploration of a broader range of personal issues which underlie abuse and control, such as self-esteem and childhood experiences. The focus in phase two is on abuse and control issues. Phase three is six months or longer of advanced therapy, and includes training of phase one group co-leaders. An innovative aspect of the project is the use of recovering violent men to co-lead Phase one groups, to conduct men’s assessment interviews, and to speak in public about spouse abuse and treatment. The Assaultive Husbands Program draws its' clientele from court referrals. Often these men have long histories of intimate violence, are minimally motivated to work on personal problems, have little insight and are resistant to interventions. Despite this rather poor prognosis, the Assaultive Husbands Program has served for thirteen years as a option between jail and judicial admonishment. At the time of its inception these were the only options used judicially. Consequently, judges who were uncomfortable with sending a man to jail (which they frequently saw as counterproductive) would either admonish him (without teaching him how to manage his anger) or refuse to hear the case. This latter response discouraged Crown from proceeding with "family violence" cases and police from arresting or even reporting them. The Assaultive Husbands Program was established in 1982 to remedy this situation. Cognitive-behavioral therapies for anger Cognitive-behavioral therapies are based on three fundamental assumptions: (l) that cognition affects behavior, (2) that cognition may be monitored and altered, and (3) that behavior change can be generated through changing cognitions (Dobson and Block, l987). Stemming from a growing body of literature in the l970s emphasizing the role of cognition in anxiety (Lazarus and Averill, l972) and depression (Beck, l976), and from a dissatisfaction with the results of long term psychodynamically oriented psychotherapy (Eysenck, l969, Rachman, and Wilson, l980), cognitive behavioral treatment began to be applied to a wide variety of affective and behavioral disorders (Dobson and Block, l987). These included problems with "self-control" (Mahoney and Thoreson, l979), anxiety (Meichenbaum, l977), depression, (Beck, l976) and anger (Novaco, l975). Novaco's (l975) application of cognitive-behavioral therapy to anger management focused on the interrelationship of autonomic and cognitive determinants. (See Dutton, 1995, for an in-depth explanation). Novaco also emphasized the positive functions served by anger arousal: it energizes behavior, serves expressive functions (advertising potency and determination) and defensive functions (overrides feelings of anxiety, vulnerability and ego threat). If anger serves this variety of functions, therapy must include alternative means for clients to satisfy each function. However, this is not always easily accomplished in short term therapy. For example, if anxiety and vulnerability increase as a result of a client's learning to reinterpret his anger, limited therapeutic time may not permit adequate strategies for dealing with these alternative, and male sex-role dissonant, feelings. For this reason, a less ambitious therapeutic objective may be the development of awareness in the client that other feelings can provide a sub-stratum for anger. Novaco's (l975) "anger management" treatment was designed to alter clients anger-enhancing cognitions. To this end, Novaco focused on (l) changing client's perception of the "aversive stimulus" or incident from a personal affront to a task that requires a solution, (2) teaching clients to use their own arousal as a cue for non-aggressive coping strategies, (3) increasing clients' perceptions that they are in control of themselves in provoking circumstances, (4) teaching clients to dissect "provocation sequences" into stages with self-instructions for managing each stage, and (5) teaching relaxation techniques to enable clients to reduce anger-arousal. In other words, if we refer to Figure l, Novaco's techniques would apply to all four learned aspects a stimulus-response chain described in Dutton (1995). In social learning terms, anger management attempts to modify both the perception of the instigators to aggression and the cognitive regulators of aggression. Figure 1 Cognitive mediators of instigator-aggression relationship Perception of instigator as: (1) threatening to ego (2) malevolent in intent Labeling of arousal as anger Behavioural expression of anger: (1) assertiveness (2) withdrawal (3) verbal aggression (4) physical aggression (1) to self (2) to wife (3) to other Instigator Aversive arousal Anger Aggression In order to achieve the five objectives above, Novaco first assessed the impact of various provocations on clients by means of an anger inventory (Novaco, l975) containing 90 statements of provocation incidents. Clients rated on a five point scale how angry they would feel if that incident happened to them. Analysis of reactions to the anger inventory items generated provocations that were role-played with clients. Novaco (l975) reported that his treatment techniques successfully allowed subjects to lower self-report anger scores and physiological indices associated with anger (i.e., systolic and diastolic blood pressure, and galvanic skin response levels). In addition, subjects demonstrated improved scores on interpersonal reactions to provocation (i.e., constructive action scores were increased while verbal and physical antagonism scores were decreased). While these results were promising, they were obtained under role-play conditions which lend themselves to subjects occasionally cooperating to verify experimental hypotheses (Orne, l969). However, even if these "demand characteristics" (which cue subjects to the nature of the experimental hypotheses) were involved in Novaco's assessment, they still demonstrate that when angry clients want to lower their anger they can. They can effectively improve their affective, physiological and interpersonal responses to provocation. Novaco also tried to offset the artificial nature of role-played provocations by conducting "direct experience laboratory provocations" (unsuspected personal affronts that would be perceived as real-life provocations). However, given that his subjects were wired for physiological measurement, the deceptive nature of this design is questionable. Novaco reports comparisons of various "partial treatment" groups (e.g. cognitive control alone, relaxation training alone, etc.) in an effort to ascertain which components had the greatest effect on improved anger management. Subjects themselves reported the most important aspect of treatment to be task orientation when faced with a provocation. Task orientation requires clients to define the situation as one requiring a solution rather than an attack and directs attention away from internal stimuli associated with anger. The "automatic" perception of a provocation as a personal affront begins to change as clients learned that not becoming demonstrably angry did not mean that they had to give up their position or back down (i.e., that increased assertiveness was possible when anger was controlled). Treatment groups for wife assaulters The development of treatment groups to specifically work with wife assaulters was pioneered by Anne Ganley (Ganley & Harris, l978; Ganley, l98l). Ganley developed her treatment program from a social-learning orientation focusing on poor conflict resolution skills learned by wife assaulters in their family of origin: violence was often the only means of dealing with conflict-generated anger, listening skills were poor, verbal problem-solving skills were poor and emotional self-disclosure was equated with loss of control. As a step toward rectifying these deficits, Ganley included assertiveness training as part of her treatment model. Typically, her groups focus not only on physical abuse but on verbal and sexual abuse, and abuse of property as well. Since the abuse takes place in the context on an intimate relationship with a women, the man's perceptions about women, his wife, and the meaning of intimacy to him all must be explored. In particular, an assessment should be made of how the man uses abuse. That is, the control that he has over his use of violence should be assessed to determine to what extent he uses violence as a form of control. Some wife assaulters use violence with this function, while others appear to use it more reactively (either to a felt insurmountable conflict or to alleviate an internal tension state). Ganley viewed battering as a learned tension-reducing response that occurred in the family setting because that was the safest place to aggress without punishment and because batterers had stereotyped views of the man as being the absolute ruler at home. Ganley described the tendency of batterers to deny or minimize their violence and abuse and to externalize it by holding others responsible and culpable for their own moods, and outbursts. She recommended confrontation as a therapeutic strategy for dealing with these forms of "neutralization of self-punishment". She also developed a highly structured treatment format that stressed personal accountability to each participant. Exercises such as getting men to write a personal policy statement about their use of violence, and maintaining an anger diary emphasize to clients the need for personal responsibility in constant monitoring of anger. Batterers also tended to express emotions such as hurt, anxiety, excitement, sadness, guilt, humiliation and helplessness as anger. Ganley's treatment program develops batterer's motivation to change by helping the batterer identify other negative feelings besides anger. Men may use anger to ward off chronic "emotional scripts" of guilt, shame or depression. These men must learn how to identify and better cope with these emotions. Anger diaries help the man to identify the instigators of his anger, and his physical and cognitive responses to anger. Men list the "triggers" (instigators) of their anger (i.e., what another person did or said to anger them), how angry they became (on a l0-point scale), how they knew they were angry (physiological responses, etc.), "talk up" (i.e., what they said to themselves to increase their anger) and "talk down" (what they said to themselves to calm themselves down). Men list the "triggers" as "objective" recordings of events. They are taught to be specific, not to make assumptions about the other's motive and to record only what was seen or heard. This exercise forces them to analyze how frequently they impute negative motives to others and the extent to which these assumptions generate anger. Comparisons of the "trigger" and "talk up" columns emphasize the interpretative or subjective quality of their anger responses (since the talk up column generally contains blaming statements that serve to increase the man's self-generated anger). "Talk down" or anger decreasing statements need to be taught to most men. These statements contain the acknowledgment that the man is feeling angry. They serve to both improve his ability to detect anger cues and to generate self-control through changing his self-statements from external blame to acknowledgment of internal feelings. For a complete description and examples of anger diaries, the reader is referred to Sonkin and Durphy (l982), or to Sonkin, Martin, and Walker (l985). When men are successfully in treatment groups and consistently completing anger diaries, the diaries are used as a step to assertiveness training. Bower and Bower (l976) develop assertiveness by getting clients to verbalize a "DESC script". DESC is an acronym for (l) describe, (2) express, (3) specify and (4) consequences. Hence, clients are asked to describe what behaviors in the other bother them, to express how these behaviors make them feel, to specify what new behaviors they want, and to express the (positive) consequences for the other person if they perform these behaviors. This assertiveness exercise becomes a first step in teaching clients to negotiate interpersonal differences. The bridge from the anger diary to the DESC script is built as follows: the "triggers" from the anger diary (specific acts or statements) provide the Describe portion of the DESC script. The specificity learned in keeping an anger diary helps the man to focus his verbal statements on a behavior rather than a predisposition of the other person. The "talk down" column (statement of feeling) then becomes the Express part of the DESC script. Hence, the anger diary provides the first half of an assertive statement, often an improvement in communication without the Specify and Consequence portions. These latter steps teach the man to assert what changes he wants and what changes he is willing to make himself as a consequence. Hence, they develop a problem-oriented or negotiation approach to the communication of anger. After rehearsal and practice in the treatment group, men are encouraged to continue with couples communication therapy. It should be emphasized that the Assaultive Husbands Project recommends such treatment only when the use of violence by the man is under control. The communication aspect of treatment is designed to improve the conflict climate of his primary relationship, once his partner no longer feels at risk for further violence. Motivation to change Men who come to treatment through the court system may have no prior experience with treatment groups. The thought of disclosure of personal problems in front of other men is discomfiting. Denial and minimizing of abuse is commonplace, as is victim blaming, blaming the criminal justice system or externalizing blame to any other source in an attempt to ward off personal responsibility for violence. How then to motivate these men to work at reducing their violence? Prochaska, DiClemente & Norcross (1992) present a model of the change process applied to addictive behaviors. They define the stages of change as follows: 1. Precontemplation - No intention to change, no awareness or "underawareness" of the problem. Families and close friends are aware of problem. 2. Contemplation - Awareness of problem and serious thought about overcoming it but no action plan. ("Knowing you want to go (to treatment) but not quite ready yet." op. cit. p.1103). 3. Preparation - Intending to take action, with some minor attempts but no substantial behavioral change. 4. Action - Overt behavioral change with successful modification from one day to six months (depending on the prior frequency of the behavior). 5. Maintenance - Work to prevent relapse and consolidate the gains from action. Prochaska et al. describe the empirical studies of change for a wide variety of addictive behaviors. Based on these studies, most people do not successfully maintain their gains on their first attempt at change. Rather a type of spiral process occurs where gains and relapses follow each other as the person progresses from stage 1 to 4 as in Figure 2. Figure 2 A spiral model of the stages of change (Prochaska, DiClemente & Norcross, 1992) When we apply this analysis to wife assaulters, the following conclusions seem reasonable. Wife assault is a form of addictive behavior, the man has learned to use it habitually in ego threatening or stressful conflict situations. Men who enter treatment are anywhere from the precontemplative stage to the action stage. Denial and minimizing are similar to underawareness as reported by Prochaska et al. Unfortunately, Prochaska et al.'s conclusion is that some form of relapse is almost inevitable for men who are trying to quit for the first time and have entered therapy in the earlier (precontemplation-contemplation stages). This would describe almost all court mandated men. Again, we have to wonder whether it is realistic to expect court mandated men to become violence free immediately after a 16 week treatment group. Prochaska et al. also suggest that different forms of treatment might work better with men at varying stages. Consciousness raising, for example, is important at the precontemplation stage, whereas reinforcement management is more important at the action stage. In court mandated groups, a detailed examination of the range of abusive acts: physical, emotional, sexual and damaging property and pets constitutes part of the consciousness raising. Men frequently comment on how abuse is much more frequent and pervasive than they had formerly thought, and how they were more abusive than they had thought. (Another form of consciousness raising, is to get the men to attend to the language they use to describe women, especially when they are angry. What does this language which typically focuses on the woman as a "whore" or "slut" tell us about the deeper images or templates that man carries in his view of women?) Reinforcement management would include redefining power from control of the other to self-control and rewarding the self for self-control. According to the authors, the stages and processes used by Prochaska et al. (1992) were able to predict with 93% accuracy which patients would drop our prematurely from therapy. Below we will consider the question of whether treatment groups "work". According to the analysis just presented, it might be naive to expect court mandated men who are forced into treatment in the precontemplative stage to become completely violence free after 16 weekly treatment sessions. Sex-role socialization, power, and group processes While anger recognition and improved communication skills provide the essence of therapy for assaultive males, other issues also constitute important adjuncts to this treatment. Since treatment is for male-female violence, the contribution of sex role socialization to setting the stage for violence is important. Male socialization both narrows the range of "acceptable" emotions (Fasteau, l974; Pleck, l98l) and creates occasionally unrealistic expectations about family roles. This can be a source of chronic conflict in a relationship (Coleman & Straus, l986). Treatment must address these issues and attempt to develop empathy for the victim (see Edelson & Tolman 1992). One way to develop empathy is to have assaultive males describe their own experiences as victims of abuse, if it occurred, by their parents. Exploring the feelings connected to these experiences in the group, and explicitly relating those feelings to the wives’ experiences as victims, can serve to strengthen empathy. This process makes salient the negative consequences of violence for the victim. In Assaultive Husbands Program treatment groups, the discussion of abuse by parents has usually been a very emotional experience, and one in which strong group cohesiveness is developed. For many of clients, this is the first time in their lives that they have discussed these painful experiences. Treatment should also include an attempt to have males think about power in a different way. A man typically enters treatment thinking about power vis-a-vis his wife in an adversarial fashion: his gain is his wife's loss, and vice versa. The therapist encourages each man to view power in interdependent terms: that by diminishing his wife he loses a vital partner, and by accepting her empowerment he actually gains. This is done by making salient the personal losses he sustains as a result of violence toward his wife (e.g., her emotional and sexual withdrawal, mistrust, and chronic anger toward him). By repeatedly "yoking" the couple's gains and losses, the concept of power interdependence becomes clearer. In this way an objective of feminist therapy is addressed in a means less threatening than confronting men's attitudes about sexual politics. Feminist therapy works toward a recognition that violence is used to maintain the power imbalance in a patriarchal society, and that such an imbalance is unhealthy and unproductive for all concerned. However, to attempt to change assaultive males' attitudes about patriarchy in a limited therapeutic time period can be ineffective: it can, for most individuals, simply offer an opportunity to cast responsibility for their behavior onto the social system, rather than onto themselves. Furthermore, many assaultive males feel powerless, so trying to convince them that they are powerful and controlling can backfire therapeutically. What therapy can do is clarify the negative consequences of any individual's attempts to achieve greater power solely through control of another person, and the positive consequences of sharing that power and of finding noncoercive lifestyles that lead to a feeling of empowerment. Strangely, empowerment issues, so important for battered women, are also therapeutically salient for wife assaulters. As assaultive males find alternative methods of feeling inner power, their need for power over their wives is diminished (see Ng, 1980, for an extended discussion of this distinction). Thus, the assaultive male moves closer to a feminist perspective, but on an individual, rather than systems, level. Finally, since treatment for assaultive males usually occurs in a group setting, group process issues (Yalom, l975) are also important. Most clinical texts describe assaultive males as being isolated (Ganley, l98l; Sonkin et al., l985). They frequently feel anxious about describing personal problems and feelings in front of other men. Therapists do considerable "bridge building" by explicitly connecting the experiences of men in the groups in order to establish some camaraderie and a sense of safety in self-disclosure. On the other hand, the therapist cannot allow group cohesiveness to generate mutual protection in the service of denial and minimizing. Since men in these groups may have a shared sense of outrage at the "injustice" of an arrest for what they consider to be a minimal act, they may try to feed each other's tendency to blame the victim, women in general, or the criminal justice system. The therapeutic objective is to allow resentments to be expressed in the group, while still confronting the men's perceptions of blame. Having a reformed client, or clients, participate in group sessions can be extremely useful in diffusing the individually confrontative role the therapist would otherwise have to assume. Regardless of whether the therapist operates alone or with reformed clients when confronting individuals in the group, if some or all of the clients perceive the challenges as attacks, group polarization can occur as a form of defense. The therapist must make it clear that confrontation is used to help each individual accept responsibility for his actions and learn new ways to manage conflict; they are not used to attack or judge. Are treatment groups effective? Since the late 1970s there has been a proliferation of court-mandated treatment groups for men convicted of wife assault. Browning (l984) and Eddy and Meyers (l984) provide descriptive profiles of numerous treatment programs for assaultive males. Both reviews outline referral processes, treatment procedures and funding issues for such programs, and agreed on the need for an evaluation of treatment effectiveness. Treatment groups for wife assaulters originated from public pressure on the criminal justice system to respond more effectively to the problem of wife assault (U.S. Commission on Civil Rights, l978; Dutton, l981; Standing Committee of Health, Welfare and Social Affairs, Canada, l982). As Dutton (l994b) pointed out, the hopes for such groups were twofold. First, the groups were seen as a means of improving protection for women who opted to remain in a relationship with a husband who would not seek treatment voluntarily. Second, by providing a viable sentencing option for judges, treatment groups could create a salutory "ripple effect" throughout the criminal justice system by making judges more willing to convict, prosecutors more willing to proceed with cases (where they perceived their chances of gaining a conviction as having improved), and police more willing to proceed with charges that they perceived as being actionable by prosecutors. Clearly both of these hopes were based on the expectation that treatment groups would be effective. Furthermore, since incarceration for a first offense of wife assault is unlikely (Lerman, l981; Dutton, l987), treatment groups represent an addendum to probation which could provide convicted men with a means for managing anger (cf. Novaco, l975). The hazard of treatment groups lies in their offering false hope. If men remain at risk for violence despite treatment, then their wives may be imperiled while falsely believing that the man is "cured". Therapists who offer treatment to abusive men have an obligation to educate their partners as to the realistic chances of improvement. Offender treatments for other behavioral problems have had mixed results (Shore and Massimo, l979; Gendreau and Ross, l980). Clearly, there is a need to assess the effectiveness of court-mandated treatment for wife assault. Several evaluation studies have been done with mixed results. Randomized designs that allocate potential clients to treatment and non-treatment conditions were recommended by Dutton in 1981 but at that time the groups were still considered too experimental to warrant an allocation of research funds. Now that treatment groups are widely established an opposite problem exists, probation services do not want to deny a group of men their right to be treated after a conviction for wife assault. Hence, the time for a randomized assignment to treatment and control groups to assess conclusively treatment efficacy may have come and gone. Dutton (1986) and Rosenfeld (1992) lamented the lack of a randomized design and fell back on the next best available assessment technique; a matched group design. The problem with a matched group design is that one is rarely sure whether the experiential and control groups are matched on all relevant features. In assessing wife assault treatment, for example, it sometimes appeared that treated groups may have been comprised of men with higher motivation to change than were untreated controls. What constitutes "effective" treatment? The "effect size" of treatment (Rosenthal, l983) for wife assaulters can be established by estimating what percentage of men would not repeat assault without treatment. Schulman (l979) and Straus, Gelles, & Steinmetz (l980) reported that single events of severe violence occurred for 21.0% and 27.8% respectively in couple’s marriages; and the probability of repeat physical violence was 63% and 66% respectively for couples who had reported single incidents of severe violence. If we consider 24.4% (the mean of 21.0 and 27.8%) as a target baseline for treatment, we cannot compare the survey data to the recidivism rates for treated assaultive men. The Schulman (1979) and Straus et. al. (1980) data were collected on the general population, whereas men who report for treatment either through the courts or at their wives' behest are more extreme cases, typically with longer histories of violence than a general population sample. In the Assaultive Husbands Program the men's use of physical aggression, whether assessed by their own or their wives' reports, puts the men in the top one percent of population scores for violence. In determining effectiveness, any effective reduction with this population may seem successful to the therapist. If the outcome data for treated men is similar to the survey data for the general population, the outcome data should be interpreted with the severity of pre-treatment violence in mind. Another point to consider in assessing group effectiveness is that frequency of violence rates vary greatly for men in treatment, so pre-post individual comparisons should be made which compare post-treatment behavior to each man's pre-treatment individual frequency rate. How might we establish a baseline against which to judge treatment group success? A major problem in assessing wife assault treatment effects has been the lack of a baseline measure of recidivism for a matched group of untreated offenders. In the absence of randomized designs, the next best comparison group is a group of males demographically similar to the treated group, with similar arrest patterns prior to arrest, and (if available) similar patterns of frequency of wife assault. The Sherman and Berk (l984) study reported above included a six month follow up of 161 "domestics" where 80% of the perpetrators had chronic assault histories. Police contact indicated an overall recidivism rate of 28.9%, and a 19.0% rate for men who were arrested. Hence, the Sherman and Berk (1984) study allows us to estimate expected recidivism with surveillance (e.g., probation) after criminal justice intervention. Specifically, if a "treated" population has been arrested prior to treatment we might expect a 19.0% recidivism rate in the first six months. For reasons described above, it is extremely difficult to do a randomized experiment that assigns assaultive men to control and experimental conditions by chance. Many service providers would be concerned with the ethics of refusing treatment to a designated "control." If the man re-offended could the victim sue the criminal justice agency for putting her at risk? Could his defense lawyer for the re-offense argue that his client never received the treatment to which he was entitled? Researchers counter that until a randomized design is done, we will never know for sure if treatment really works. Another scenario also supports the need to conduct a randomized design. A man court-ordered for treatment could be rejected on the grounds that his motivation is low and he takes little responsibility for his violence. Shortly afterwards, he reoffends. His defense lawyer asks whether the criteria used to exclude him from treatment has been proven to be valid; that men who do not meet the intake criteria do not benefit from treatment. The clinical intuition and experience is that poorly motivated men do not benefit from treatment. experimentally validated in research. In lieu of a randomized design, several studies have compared treated men with matched groups who had similar histories of assault, were arrested and convicted, but never received treatment. For example, Dutton (1986) scrutinized police records of a group of 50 untreated men for up to 3.0 years post-conviction. During the same period, records for 50 men arrested, convicted and "treated" for wife assault were examined. Demographic comparisons of the treated and untreated groups were virtually identical and were similar to the men studied by Sherman & Berk (1984). During the first six months (while both groups were on probation) both samples behaved similarly as well. The Sherman and Berk sample reoffended in 13.0% of cases while the Dutton sample reoffended in 16.0%. However, when probation ended for the Dutton untreated sample, their recidivism rate (based on police arrest records) jumped to 40.0% within the next two years. For the treated group recidivism remained at 4.0%. Dutton (1995) argued that men can suppress physical abuse while under surveillance and when they have something to lose from reoffending (e.g. believe they will go to jail). However, if they have not learned the requisite skills to manage intimate-anger, they will relapse after probation ends and/or their abuse will transmogrify into less detectable forms. The Dutton (1987) data support this interpretation as do data from a study by Dutton, Hart, Kennedy & Williams (1992) that assessed perceived sanctions for reoffending in a group of men in treatment for wife assault. Rosenfeld (1992) presents an overview of the results of 26 such evaluations, including 11 evaluations of untreated or dropout offenders. For untreated offenders the recidivism rate, according to police reports, goes from 16.0% at six months (Dutton 1986), to 20.0% at one year (Waldo 1988), to 40.0% at 2.5 years (Dutton 1986). The average recidivism rate for untreated men is 23.4%, for men who dropped out of treatment 29.0%, and for men who complete treatment 8.4%. Hence, based on several studies, treatment seems to diminish recidivism by about 66% (8.4/23.4). This result should be viewed as positive by criminal justice system However, the veracity of this has not been professionals. It is important to keep in mind, however, that treatment “drop-outs” may differ in important ways from those who complete treatment. When we turn to wives' reports of recidivism we find that 33.0% of wives' of men who completed treatment reports some post-treatment violence compared to 47.3% of dropouts wives. (The 16% figure in the Dutton (1986) study seems low because only Severe Assaults on the Conflict Tactics Scale were counted, reasoning that they were the ones most likely to be injurious and requiring repeat police presence.) Interestingly, Rosenfeld concluded, despite his data presented above, "men arrested but not referred to treatment appear to resume their violent behavior no more frequently than men arrested and treated" (op. cit. p. 221). Dutton (1995) has argued that his data do not support that negative conclusion, pointing out that Rosenfeld erred in a few places in his analysis. For example, he presented the recidivism figure in the Dutton (1986) study for the untreated group as 20.0% when it was 40.0% (p. 215), leading to an underestimate of the recidivism rate in untreated men. He also described the Dutton (1986) untreated group as more "treatment resistant" because they had more prior arrests (it was only .8 versus .88 in the two groups) and because some had been untreated because of extreme resistance and denial (only 8 of 50). Dutton (1995) suggested that Rosenfeld’s interpretations lead to the inference that he was trying make a case against treatment groups rather than objectively read the data. Such is the recent history of assessment of group treatment. Three major criticisms can be made of previous studies: (1) they did not utilize random designs, (2) they did not use long term follow-up assessment periods (the longest was 2.5 years by Dutton 1986), and (3) they did not systematically examine characteristics of men who presented for treatment. For ethical and logistical reasons described above, the present study could not utilize a randomized control group. However, this study will attempt to remedy the remaining two faults. In the first part of this study we will conduct up to an 11 year follow-up of arrest records on treated and untreated men. In the second part, we will assess personality disorders in men who present for wife assault treatment and systematically relate the presence of those disorders to treatment success or failure. Method Part 1: Criminal History Follow-Up Subjects The initial sample consisted of 518 voluntary and court-referred assaulters assessed for the Vancouver Assaultive Husbands Program between June 1, 1982, and December 31, 1992. The majority of the subjects were court-referred and convicted of an assault against a female partner. Typically, convicted men receive a probation order of one to three years and are ordered for either “anger management” or spouse assault treatment. Less frequently, a no-contact (with the spouse) condition is imposed. Another referral method used by the courts was to specify treatment on a recognizance order. Attempts were made to locate criminal records for the entire sample using the procedure described below. Records were located for 86% of the sample (N=446). This reduced sample was used for the follow-up and analyses. It included: (1) men who were referred for treatment and did not appear for an AHP intake assessment; (2) men assessed by AHP and subsequently deemed inappropriate for treatment (rejected); (3) men who were accepted but did not complete the program (i.e., attended 0-11 sessions); and (4) those who completed treatment (12 or more sessions). Comparisons were made amongst these groups (see Results). Procedure Criminal records were checked via the Canadian Police Information Centre (CPIC) by the “E” Division of the Royal Canadian Mounted Police, Vancouver. To facilitate the location of the records, full names, dates of birth and Finger Print Section (FPS) numbers were collected for AHP files. Between April 6, 1993, and September 25, 1993, criminal records for 446 men were received from the RCMP. The records contained all criminal charges and convictions. Failures in locating CPIC information usually resulted from the absence of an FPS number or birthdate. Follow-up periods were calculated for each individual by subtracting the date of treatment “discharge” (last session attended) from the date of the record check (as indicated on the CPIC sheet). Eligible follow-up periods were expressed as the number of “Days at Risk”. The calculation can be expressed as follows: Days at Risk = Criminal Record Check Date - Discharge Date The Days at Risk variable was used in subsequent analyses examining recidivism rates over time (see Survival Analyses in the results section). For other analyses, it was necessary to calculate the number offenses occurring per year at risk. For this purpose, Days at Risk was simply divided by 365.25. Pre-contact risk periods were calculated in a similar fashion. The number of days between each subject’s eighteenth birthday and the date assessed for the AHP was divided by 365.25 to obtain the pre-contact years at risk. Various types of offenses were categorized according to a number of criteria. The most important distinction was between violent and non-violent offenses. Within the violent category further distinctions were made between assaults and other offenses. Finally, attempts were made to determine the nature of all assaults; specifically, it was desirable to know if assaults were against intimate partners. Thus, requests were made to RCMP offices to clarify whether assaults were (a) against women, and (b) against intimate partners. Part 2: Personality Characteristics and Treatment Outcome Research participants Fifty-two men who had been involved in the Vancouver Assaultive Husbands Program (AHP), 60 men who had been involved in the Victoria Family Violence Project (VFVP), and all the men’s current female partners were potential follow-up participants. These two samples of men and their partners were selected for follow-up because data on the men’s pre-treatment levels of abuse were available from a study conducted by Dutton and Starzomski (1994). Fifty of the 52 men (96.1%) in the AHP sample were court-ordered for treatment. The majority of the men participating in the VFVP were self-referred. Self-referred men often request treatment for their assaultiveness because of an ultimatum from their wives and/or recognition that their behaviour in the relationship is problematic. The VFVP staff estimated that approximately one-third of their clients were court-mandated. The 60 VFVP men had volunteered for the Dutton and Starzomski (1994) study between July 10, 1990 and September 29, 1992; the total number of men participating in treatment at the VFVP during this time period was 329. Thus, the potential follow-up sample represented 18% of the treated men. The AHP men were recruited from 11 treatment groups conducted between January, 1989 and September, 1992. This sample comprised 34% of approximately 154 participants involved in treatment during this time period. As Table 2.1 illustrates, complete data were obtained at follow-up for 52% (n=27) of the potential AHP men and 63% (n=38) of the potential VFVP men. Data from partners were available for 37% (n=19) and 42% (n=25) of the AHP and VFVP samples respectively. All analyses were conducted on these followed-up subjects. The Victoria and Vancouver samples of men did not differ significantly on any of the following demographic characteristics: age (combined samples M = 35.2, SD = 6.8), education (combined samples M = grade 12), alcohol use (76% of all subjects reported alcohol problems), ethnicity (both samples were predominantly white Anglo-Saxon Protestant), and occupational status (primarily blue collar, as assessed by self-report) (Dutton & Starzomski, 1994). Procedure All subjects were followed-up after completing treatment to determine their levels of post-treatment violence. The average follow-up time period was 27 months with a maximum of 56 months, and a minimum of four. Follow-up interviewers were provided with a client list with existing contact information. Interviewers were also provided with a script to be read to participants explaining: (a) the purpose of contact; and (b) the voluntary and confidential nature of participation. Potential research participants were offered a $20 honorarium for completion of the follow-up interview. If participants had changed partners since involvement in treatment, an attempt was made to contact the most recent partner of any relationship lasting more than six months. Otherwise, an attempt was made to contact the partner on record at the time of treatment. Interviewers were chosen for their experience in working with violent men or abused women. They were usually connected to the treatment program for which they conducted follow-up interviews. For example, all VFVP interviewers had co-led treatment groups. In Vancouver, the partner interviewer had conducted some of the pre-treatment interviews with the women and was a co-leader for the men’s treatment program. The men’s interviewer in Vancouver was a therapist with extensive experience with spousal assault treatment. The follow-up interviewing process was as follows. First, an attempt was made to contact the treatment client. If contact was successful, an interview was conducted and partner information was gathered (e.g., relationship status, contact information). Attempts were then made to interview the partner. Whenever a man could not be reached, an attempt was made to contact his partner listed at the time of treatment. These women were asked to participate if they met the criteria described above. If direct contact information from the men’s files was invalid (i.e., home and work phone numbers), alternative strategies included (1) contacting friends, relatives and acquaintances, (2) methodically dialing every surname in telephone directories that might apply, and (3) making inquiries to BC Corrections Branch probation offices for updated contact information. Generally, interviewers would discontinue phoning after six to ten unsuccessful attempts. Measures Conflict Tactics Scale (CTS). The CTS (Straus, 1979) is a standardized scale designed to measure the frequency and intensity of 19 different conflict resolution tactics. The scale includes measures of verbal and physical abuse. Respondents estimate both their own and their partner’s use of these tactics. In the current study, subjects reported the number of times during the past year that various tactics were used. Straus, Gelles, and Steinmetz (1980) have published population norms for the usage of each tactic in a variety of intimate relationships. This measure was used as a pre-treatment measure of physical violence and administered to men and their partners. Severity of Violence Against Women Scales (SVAWS). The SVAWS (Marshall, 1992) is a 46-item measure consisting of four subscales: threatening behaviour, mild physical violence, severe physical violence, and sexual violence. The SVAWS assesses the incidence and severity of physical violence towards women in the past 12 months. There are versions both for men and their partners (the latter focuses on the man’s behavior). This instrument was used to measure physical violence at follow-up. Modified CTS Scale (CTS-Mod). A flaw in the design of this study was the use of different measures of physical violence at the pre- and post- stages of assessment. In order to compare levels of Men’s pre-treatment violence, as assessed with the CTS, and Men’s follow-up violence, assessed with the SVAWS, scores on common items were converted to a “Modified” CTS score. Table 2.1 shows the common items on the CTS and SVAWS. Table 2.1 Comparable CTS and SVAWS items CTS k) Threw or smashed or hit or kicked something. l) Threw something at the other one. SVAWS 1. Hit or kicked a wall, door, or furniture. 2. Threw, smashed or broke an object. 4. Threw an object at her. 21. Pushed or shoved her. 22. Grabbed her suddenly and forcefully. 29. Slapped her with the palm of my hand. 30. Slapped her with the back of my hand. 31. Slapped her around the face and head. 28. Bit her. 33. Punched her. 34. Kicked her. 32. Hit her with an object. 38. Used a club-like object on her. 39. Beat her up. 19. Threatened with a knife or gun. 40. Used a knife or gun on her. m) Pushed, grabbed, or shoved the other one. n) Slapped the other one. o) Kicked, bit or hit with a fist. p) Hit, or tried to hit with something. q) Beat up the other one. r) s) Threatened with a knife or gun. Used a knife or gun. The CTS item scores were re-scaled to make them directly comparable to the SVAWS items. The CTS item categories are zero for “never”, one for “once”, two for “twice”, three for “three to five times”, four for “six to ten times”, five for “11 to 20 times”, and six for “more than 20 times”. The SVAWS items response categories are zero for “never”, one for “once”, two for “twice”, and three for “many times”. Thus, CTS item scores of four, five, and six were recoded as three’s in order to match the SVAWS category of “many times”. For those CTS items corresponding to two or three SVAWS items (i.e., items k, n, o, and p), SVAWS scores for each item were added up to a maximum of three (that is, the maximum score possible on the CTS item). These “modified” CTS scores were used as the principle measure of pre- and post-treatment physical violence in our analyses. Psychological Maltreatment of Women Inventory (PMWI). The PMWI (Tolman, 1989) contains 58 items measuring various forms of psychological or emotional abuse of women. Items describe behaviors that are rated from one ("never" occurs) to five (occurs "very frequently"). dimensions measured by the inventory. Tolman (1989) described two The first, “dominance/isolation”, includes items related to rigid Second, observance of traditional sex roles, demands for subservience, and isolation from resources. emotional/verbal abuse includes withholding emotional resources, verbal attacks, and behavior that degrades women. According to Tolman, Factor analyses support the separation of the two factors. There are forms both for men and their partners. The PMWI provides information on the man’s psychological abuse of a female partner in the last year. In the current study, this inventory was administered only to women. Multidimensional Anger Inventory (MAI). The MAI (Siegel, 1986) is a 38-item self-report scale assessing the following dimensions of anger response: frequency, duration, magnitude, mode of expression, hostile outlook, and range of anger-eliciting situations. Siegel (1986) reported the results of a factor analysis of this scale and the reliability of its subscales (coefficient alphas ranging from .51 to .83). He also reported internal reliability coefficients for the entire scale (coefficient alphas of .84 and .89 for two separate samples). The scale also correlates highly with other conceptually similar anger inventories. This scale was completed by men and their partners (focusing on the Men’s anger) prior to treatment and at follow-up. Michigan Alcohol Screening Test (MAST). The 25-item MAST (Selzer, 1971) measures the subject's perception of his/her own drinking, as well the types of problems which have arisen for the subject due to his drinking. According to Selzer, scores of five or greater are indicative of alcoholism. The MAST was administered pre- and post-treatment both to men and women (again focusing on the man’s behavior). Drug Abuse Screening Test (DAST). The DAST (Skinner, 1982) was used to assess substance abuse. It is a 28-item self-report scale which focuses on consequences of drug use. The DAST has a reported coefficient alpha of .92 and test-retest reliability of .85 (Skinner, 1982). It was administered in exactly the same manner as the MAST. Marlowe-Crowne Social Desirability Scale. Dutton and Strachan (1987) and Dutton and Hemphill (1992) emphasized that studies of wife assaulters need careful assessment of the degree to which respondents minimize violence or other socially undesirable behavior or attitudes. The Marlowe-Crowne Social Desirability Scale (Crowne & Marlowe, 1960), consisting of 33 items, was employed in this study as a measure of subjects' attempts to underreport their unfavorable conduct on self-report measures. This scale was administered at the pre- and post-treatment stages to men and women. Unlike the other measures, however, women were asked to complete the Marlowe-Crowne Scale with reference to themselves. Self-Report Instrument for Borderline Personality Organization (BPO). The BPO (Oldham et al., 1985) is a 30-item scale containing three subscales. The first, identity diffusion, refers to a poorly integrated sense of self or of significant others. The second measures primitive defenses such as splitting, idealization, devaluation, omnipotence, denial, projection, and projective identification. The final subscale, reality testing, examines external versus internal origins of perceptions, evaluation of own behavior in terms of social criteria of reality, differentiation of self from non-self, internal reality testing, and the cognitive process of reality testing. Coefficient alphas of .92 for the identity diffusion subscale, .87 for primitive defenses, and .84 for reality testing were reported by Oldham et al. (1985). Oldham et al. provided data on the scale's relationship to DSM-III (American Psychiatric Association, 1980) diagnoses, and the application of the scale within the context of differing theories of BPO. This scale was administered to the men prior to their participation in treatment. Millon Clinical Multiaxial Inventory-II (MCMI-II). The MCMI-II (Millon, 1992) contains 175 items that form 25 scales intended to measure a variety of Axis I (six clinical syndromes, three severe clinical syndromes) and Axis II disorders (10 clinical personality patterns and three severe personality pathologies) from the DSM-III-R (American Psychiatric Association, 1987), as well as three response styles (disclosure, debasement, and desirability). Scoring of the MCMI-II is complex: first, raw items are weighted to yield scale scores; second, scale scores are weighted to control for response styles; and finally, weighted scale scores are transformed into base rate (BR) scores. Base rate scores over 75 are indicative of the presence of a particular characteristic (e.g., passive-aggressive personality disorder), and those above 85 define a characteristic as a predominant feature in the respondent's personality. All MCMI-II protocols in the present study were machine scored by National Computer Systems. As with the BPO this measure was only administered to men at the pretreatment stage. Table 2.2 summarizes the measures used with the participants at each stage of the study. Table 2.2 Measures administered Participants Treatment clients • • • • • • • Pre-treatment measures Conflict Tactics Scale Modified Conflict Tactics Scale Multidimensional Anger Inventory Michigan Alcohol Screening Test Drug Abuse Screening Test Marlow-Crowne Social Desirability Scale Self-report Instrument for Borderline Personality Organization Millon Clinical Multiaxial Inventory-II Conflict Tactics Scale Modified Conflict Tactics Scale Psychological Maltreatment of Women Inventory Multidimensional Anger Inventory Michigan Alcohol Screening Test Drug Abuse Screening Test Marlow-Crowne Social Desirability Scale • • • • • • • Severity of Violence Against Women Scales Modified Conflict Tactics Scale Psychological Maltreatment of Women Inventory Multidimensional Anger Inventory Michigan Alcohol Screening Test Drug Abuse Screening Test Marlow-Crowne Social Desirability Scale • • • • • • Follow-up measures Severity of Violence Against Women Scales Modified Conflict Tactics Scale Multidimensional Anger Inventory Michigan Alcohol Screening Test Drug Abuse Screening Test Marlow-Crowne Social Desirability Scale • Clients’ partners (Note: All questionnaires addressed only the man’s behavior) • • • • • • • Results Part 1: Criminal History Follow-Up Analysis of criminal history data was done in four stages. First, background information (demographic characteristics, pre-contact crimes) was examined to determine the representativeness of the sample and participation group uniformity. Second, regression analyses were executed to determine the relative effect of completing treatment, referral type (voluntary versus court-mandated), and pre-contact crimes on recidivism. Third, supplementary analyses compared a number of participation sub-groups on several crime statistics. Finally, “survival analyses” were conducted to determine the recidivism rates for the AHP men. Before proceeding with the analyses, a priori definitions of group membership were determined. Thus, the total sample was divided into the four sub-groups described in Table 3.1. Table 3.1 Participation group definitions Participation group • • • • Completers Non-completers Rejections No shows Definition Completed at least 12 of 16 group treatment sessions Attended intake interview and completed less than 12 sessions Attended intake interview; did not meet group inclusion criteria Referred to program but did not attend intake interview The above breakdown provided three potential “comparison groups” for the treated wife assaulters (completers). It was determined, however, that the rejection group was inappropriate for comparison purposes since it was known that they differed from treated men on important variables such as “motivation for change” and “acceptance of responsibility for behavior”. Indeed, these variables were used as inclusion criteria for the group. Thus, most of the subsequent comparisons were made between the completers, non-completers, and no show groups. Data for the rejected group are included for the interest of the reader. It was necessary to further divide the sample into court-mandated referrals (n=372) and voluntary referrals (n=74) sub-samples. This division was made with the assumption that the two types of referrals could differ substantially on a number of pre-contact variables. Therefore, all analyses were conducted three times: once each for the entire sample, court-mandated sub-sample, and voluntary sub-sample. Background Information1 Demographic Information. Demographic information was gathered during the intake interview. This information was analyzed to determine if there were (a) differences between the reduced sample (N=446) and the complete sample (N=518), and (b) differences between the participation groups. An independent t-test analyses of age indicated no difference in mean age between the reduced and complete sample. Similarly, chisquare statistics computed for education, occupation, employment (full- versus part-time), marital status and ethnicity indicated no differences between these two samples. Thus, there was no evidence that the reduced sample was not representative of the entire population of AHP referrals. The demographic information for the reduced sample is included in Table 3.2. The same variables were examined for differences between the participation groups. summarizes this information. Table 3.2 1The analyses of background information focused on demographic and pre-contact criminal history. For court-mandated subjects, information also existed on the disposition of their most recent assault. There were no differences on disposition variables between (a) the final sample of court-mandated subjects (n=372) and the original sample (n=424), and (b) any of the participation groups. Table 3.2 Demographic characteristics of the participation groups Total Completers 156 34.4 59.6% 40.4% 82.6% 17.4% Non-completers 167 33.0 41.6% 58.4% 86.1% 13.9% Rejected 32 36.5 50.0% 50.0% 92.9% 7.1% No Shows 91 33.1 55.3% 44.7% 84.0% 16.0% n Mean age Education* • grade 12 or less • >than grade 12 Occupation • Blue collar • White collar/ professional Employment* • Full-time • Part time, unemployed, social. asst. Marital status • married/ common-law • separated/ divorced/single 446 33.7 51.2% 48.8% 85.0% 15.0% 67.0% 33.0% 75.9% 24.1% 62.9% 37.1% 46.4% 53.6% 65.9% 34.6% 58.7% 41.3% 62.6% 37.4% 52.5% 47.5% 66.7% 33.3% 61.3% 38.7% Ethnicity 74.1% 75.7% 77.3% 59.3% 67.9% • Ethno-cultural majority 25.9% 24.3% 22.7% 40.7% 32.1% • Ethno-cultural minority * Denotes statistically significant difference between completers and non-completers No differences were found between completers and no shows on any of the demographic variables. However, some differences existed between the completers and non-completers. For example, completers were more educated, X2 (1, N = 273) = 8.80, p < .005, and more likely to have full-time employment, X2 (1, N =296) = 5.82, p < .02. As well, completers were more likely to be in a relationship. This trend approached statistical significance, X2 (1, N = 305) = 3.14, p < .08. These same analyses were performed for the court-mandated and voluntary sub-samples. There were no differences found between completers and no shows in either subsample. Once again, however, in the court-mandated sample, completers were more educated and more likely to be employed full-time than the non-completers. In the voluntary sub-sample, completers were more educated than non-completers. Pre-contact criminal history. For the participants with valid CPIC records, it was possible to examine the complete criminal histories of the subjects prior to their contact with the Assaultive Husbands Program. Participation groups were compared to determine if their were any existing differences on a number of variables. All crime variables were corrected for the number of years at risk (i.e., length of follow-up). These data are summarized in Table 3.3. Table 3.3 Pre-contact criminal history variables Total n Pre-contact statistics Mean years at risk Mean # crimes / year* Mean # violent crimes / yr.* Mean # assaults / year* 15.72 .36 .09 .08 16.34 .43 .14 .12 14.92 .60 .18 .15 18.44 .36 .09 .08 15.06 .47 .14 .13 446 Completers 156 Noncompleters 167 Rejected 32 No Shows 91 * Denotes statistically significant difference between completers and non-completers. Again, there were no differences between the completers and no shows on any of the pre-contact measures. However, completers had lower pre-contact rates of criminal offenses, t (272) = 2.49, p.< .02, violent crimes, t (296) = 2.49, p. < .02, and assaults, t (300) = 2.00, p. < .05 than non-completers. Thus, there appeared to be a higher level of pre-contact criminality in the non-completers. Note that this relationship was identical in the voluntary sub-sample. In the court-mandated sub-sample, there were no differences in precontact criminality between the completers, non-completers and no shows. Two primary conclusions can be drawn from the preceding analyses: 1. The reduced sample of 446 can be considered representative of the total AHP sample of 518. There are no data to disconfirm this assumption. 2. In general, the completer and non-completer groups differ in some important ways, namely in education, employment, marital status and pre-contact criminality. The completer and no show groups are virtually identical on all of these variables. This pattern tends to hold true for the voluntary sub-sample and, to a lesser extent, the court-mandated sub-sample. Thus, the no show comparison group appears to be the most appropriate to use for the follow-up analyses. An important caveat, however, is that the no show individuals were never accepted for treatment, leaving some uncertainty about comparability of this group. All three comparison groups (i.e., rejections, non-completers, and no shows) will be included in the following analyses. Linear regressions on post-contact criminal behaviour Regression analyses were conducted to determine the relative contribution of various “predictor” variables on criminal recidivism. Dependent Variables. Four categories of offenses were used to represent post-contact criminal behavior. The categories are defined in table 3.4. Table 3.4 Categories of criminal offenses Offense Category • • Total crimes Violent crimes Definition Any criminal offense Crimes against persons, such as confinement, abduction, sexual violations, and serious person offenses (e.g., manslaughter); nonsexual assault excluded. Includes assault, assault causing bodily harm, and aggravated assault. Assaults perpetrated against intimate partners • • Assaults Wife assault Crimes statistics were tabulated in two ways. First the number of offenses in each category was computed for each participation. Second, the number of offenses was divided by the number of years at risk (see Method section). The latter method provided offense rates which correct for differences between individual follow-up periods. Thus, crimes per year is considered the most accurate measure of post-contact recidivism. The four crime categories were divided by years at risk to construct the outcome, or recidivism, variables for regression analyses. Independent Variables. Two sets of independent variables were constructed. The first set included variables labeled “completion” (completers versus non-completers), “mandated” (referral type: court-mandated vs. voluntary referral), “interaction” (the interaction between completion and mandated), and the “pre-contact” crime corresponding to the post-contact crime being predicted (i.e., pre-contact violence was used to predict post-contact violence, pre-contact assault to predict post-contact assault, and so forth). Pre-contact rates of partner assault were not available. All variables were regressed on the four separate crime categories. The second set of variables was identical except that “completion” became defined as completers versus no shows. Again, four separate regression analyses were conducted. In all the analyses, variables were entered simultaneously allowing the unique variance accounted for by each variable to be expressed as Beta weights. Table 3.5 outlines the results of these analyses. Table 3.5 Beta weights of variables predicting post-contact crime rates Independent Variable/“Predictor” Total Crimes Completer/Non-completer • Completion • Mandated • Interaction • Pre-contact crimes Completer/No show • Completion • Mandated • Interaction • Pre-contact crimes * Denotes significant at p < .05; -.132 -.022 .061 .426*** Predicted Crime Category Violent Assaults Partner Crimes Assaults -.031 .023 .022 .349*** -.029 .005 .001 .341*** -.123 .012 .115 --.014 .115 -.012 --- .008 -.021 .001 .010 .036 .060 .047 .097 .004 .468*** .132* .150* *** Denotes significant at p < .0001 The results of the regression analyses suggest that neither completion of treatment nor type of referral (court-mandated vs. voluntary) can account for post-contact recidivism. The only variables contributing to the prediction of outcome are the pre-contact crime rates. In fact, the Beta weights for pre-contact crimes are virtually identical to the R2 statistics of the linear regressions (R2 represents the total variance accounted for by all variables), suggesting that pre-contact crimes account for virtually all of the variance in post-contact recidivism. The best predictor of future crime appears to be past crime. Supplementary criminal recidivism analyses Post-contact crimes and crime rates were analyzed in more detail for the total, court-mandated, and voluntary samples respectively. A priori decisions were made to focus on differences between completers and non-completers, and differences between completers and no shows. Therefore, pairwise tests of statistical significance were conducted without correcting for family-wise error. Table 3.6 presents the post-contact crime statistics for the total sample. Table 3.6 Post-contact crime variables for TOTAL Sample (N=446) Total n Post-contact crimes Mean contact years Mean # crimes* Mean # violent-crimes* Mean # assaults* Mean # wife assaults* % age with at least 1 assault Post-contact crimes per year Mean # crimes / year* Mean # violent crimes / year Mean # assaults / year Mean # partner assaults / yr. 0.36 0.15 0.11 0.07 0.29 0.14 0.09 0.06 0.48 0.19 0.14 0.07 0.36 0.17 0.16 0.07 0.27 0.11 0.09 0.06 5.19 1.67 0.61 0.46 0.26 25.3 5.20 1.26 0.47 0.32 0.23 23.2 4.85 2.08 0.74 0.55 0.50 28.0 5.91 2.16 0.88 0.81 0.29 37.5 5.59 1.45 0.49 0.40 0.23 21.0 446 Completers 156 Noncompleters 167 Rejected 32 No Shows 91 * Denotes statistically significant difference between completers and non-completers. For the total sample, a statistically significant difference was observed between completers and noncompleters in the total crime category, t (234) = 2.12, p.< .04. The same result occurred when total crime was corrected for time at risk (mean crimes per year), t (258) = 2.46, p.< .02. For the other crime categories uncorrected for time at risk, differences between completers and non-completers approached significance. Thus, completers tended to commit more violent crimes, t (262) = 1.84, p.< .07, more assaults, t (254) = 1.93, p.< .06, and more partner assaults, t (299) = 1.69, p.< .10. When these figures are corrected for time at risk, these trends become less statistically significant. Finally, there were no differences between the completers and no shows on any of the crime indices. Identical analyses were conducted for the court-mandated sub-sample. presented in Table 3.7 on the following page. The crime statistics are Table 3.7 Criminal history variables for COURT-MANDATED sub-sample (n=372) Total n Post-contact crimes Mean # years at risk Mean # crimes Mean # violent-crimes Mean # assaults Mean # post-wife assaults % age with at least 1 assault Post-contact crimes per year Mean # crimes / year Mean # violent crimes / yr. Mean # assaults / year Mean # wife assaults / year 0.37 0.17 0.13 0.07 0.32 0.16 0.11 0.07 0.48 0.20 0.15 0.07 0.36 0.17 0.16 0.07 0.26 0.12 0.10 0.07 4.80 1.58 0.62 0.47 0.26 25.5 4.73 1.29 0.54 0.36 0.21 25.4 4.48 1.89 0.72 0.49 0.26 26.0 5.91 2.16 0.88 0.81 0.50 37.5 5.05 1.22 0.47 0.20 0.25 20.0 372 Completers 126 Noncompleters 138 Rejected 32 No Shows 76 For the court-mandated group there were no significant differences between completers and noncompleters on any of the measures. The difference between these groups approached significance for the mean number of crimes per year. On this variable, non-completers were more likely than non-completers to commit a post-contact offense, t (218) = 1.77, p.< .08. There were no observed differences between the completers and the no shows. The following table summarizes post-contact information for the voluntary sub-sample. Table 3.8 Criminal history variables for VOLUNTARY sub-sample (n=74) Total n Post-contact crimes Mean # years at risk Mean # crimes* Mean # violent-crimes* Mean # assaults Mean # post-wife assaults* % age with at least 1 assault Post-contact crimes per year Mean # crimes / year* Mean # violent crimes / yr.* Mean # assaults / year Mean # wife assaults / year 0.30 0.08 0.06 0.04 0.15 0.04 0.03 0.02 0.48 0.13 0.09 0.07 028 0.06 0.04 0.01 7.23 2.15 0.54 0.38 0.24 25.5 7.31 1.13 0.20 0.17 0.10 13.0 6.62 2.97 0.86 0.59 0.45 34.0 8.30 2.60 0.60 0.40 0.13 27.0 74 Completers 30 Noncompleters 29 No Shows 15 * Denotes statistically significant difference between completers and non-completers. It is clear that the most significant results occurred in the voluntary sub-sample. Completers differed from non-completers on virtually all of the criminal indices. Thus, voluntary completers committed fewer crimes, t (38) = 2.18, p.< .04, fewer violent crimes, t (34) = 2.33, p.< .03, and fewer partner assaults, t (36) = 2.34, p.< .04, than voluntary non-completers. As well, differences between these two groups approached significance for mean number of assaults, t (36) = 1.75, p.< .09, and percentage with at least one assault, X2 (1, N = 59) = 3.64, p < .06 (again with completers committing fewer offenses). Even when corrected by time at risk, completers had a lower rate of crimes, t (35) = 2.45, p.< .02, and violent crimes, t (44) = 2.03, p.< .05. The difference between the two groups in the rate of partner assaults approached significance, t (44) = 1.91, p.< .07, with completers committing fewer offenses. Once again, there were no differences between the completers and the no shows. In sum, it appears that the completers and no shows have virtually identical recidivism patterns. Noncompleters, on the other hand, have higher frequencies of post-contact crimes, but these differences appear to exist exclusively within the voluntary sub-sample. The rates of recidivism are examined more closely in the next section. Survival analyses Survival analyses, as described in Norusis (1992), were conducted to calculate the proportion of men remaining assault-free over time. “Assault”, as described earlier, was defined to includes assault, assault causing bodily harm, and aggravated assault. Thus, it was possible to examine the “shape” or survival distribution of the four sub-groups to the time of their first assault. As with the analyses above, the procedure was executed for the total sample, as well as the court-mandated and voluntary sub-samples. The curves were similar for all three samples. Figure 3 presents the survival curve for the total sample. Figure 3 Proportions of participation groups remaining assault free over time The survival analysis procedure also computes pairwise statistics to compare survival distributions of the sub-groups. The statistics assume a null hypothesis that the groups come from the same survival distribution. All possible pairwise comparisons were computed between completers non-completers, rejects, and no-shows. There were no significant differences between the groups (e.g., completers and non-completers) for the total sample or the court-mandated or voluntary subsamples. A Cox proportional hazards regression (Norusis, 1992) was performed as a means for determining which variables influence survival times for the total sample. A similar method to that used in the earlier linear regressions was used to enter the following independent (“predictor”) variables: “completion” (completers versus non-completers); “mandated” (referral type); “interactions” (interaction of completion and mandated); and pre-contact assaults per year. This process was repeated redefining “completion” as completers versus no shows. The objective was to determine the relative contribution of these factors to the prediction of assault recidivism. The resulting regression coefficients, representing the unique variance accounted for by each variable, are presented in table 3.9 on the following page. Table 3.9 illustrates, that only pre-contact assaults per year contributes significantly to the prediction of survival times. This result is consistent with the linear regressions discussed earlier. Table 3.9 Beta weights of variables predicting survival time to first assault Independent Variable Completer/Non-completer • • • • Completion Mandated Interaction Pre-contact assaults per year - .016 - .000 - .004 .162** .000 .000 .000 .094** Regression coefficient Completer/No show • Completion • Mandated • Interaction • Pre-contact assaults per year ** Denotes significant at p < .001 Part 2. Personality Characteristics and Treatment Outcome This part of analysis proceeded in four stages. First, an examination of the follow-up sample’s representativeness was conducted. Second, analyses of covariance were conducted to determine the influence on follow-up violence of (a) the treatment program (AHP vs. VFVP), and (b) pre-treatment violence. Third, supplementary comparisons were made between subjects pre- and post-treatment levels of physical and psychological abuse. Finally, personality variables were regressed on treatment outcome. Representativeness of AHP follow-up sample To determine the representativeness of the followed-up sample three groups of men were defined: Table 3.10 Definitions of samples used for representativeness analyses Sample Final follow-up Potential follow-up Initial sample Definition Men for whom complete data were obtained at follow-up Men who participated in the Dutton & Starzomski (1994) study for whom complete pre-treatment data were available All men accepted in the AHP program Note that the final sample is a subset of the potential sample, which in turn is a subset of the initial sample. The three groups are referred to as the representativeness samples. It was possible to compare these samples on several important variables2. The variables were divided into three categories: demographic, preand post-contact violence, and personality. Demographic variables. There were no differences between the initial and potential follow-up sample on demographic measures. However, men in the final follow-up sample were older, t (44) = -2.61, p.<.02, and more likely to be employed full-time X2 (1, N = 48) = 3.81, p < .051, than the potential sample. Thus, on demographic variables, with a couple of exceptions, the final sample appeared to be a representative subset of the larger population of treated men. Table 3.11 summarizes these comparisons. that it was only possible to determine the representativeness of the AHP sample. The necessary data were not available for the VFVP sample. 2Note Table 3.11 Demographic characteristics of representativeness samples Initial sample Potential followup sample 25 31.7 58.8% 41.2% 81.8% 18.2% 54.5% 45.5% Final follow-up sample 27 36.5 30.4% 69.6% 80.8% 19.2% 80.8% 19.2% n Mean age Education • grade 12 or less • >than grade 12 Occupation • Blue collar • White collar/ professional Employment • Full-time • Part time, unemployed, social. asst. Marital status • married/ common-law • separated/ divorced/single Ethnicity • Ethno-cultural minority 272 33.6 51.9% 48.1% 85.0% 15.0% 69.4% 30.6% 57.3% 42.7% 58.3% 41.7% 57.7% 42.3% 23.6% 22.7% 23.1% Criminal History. The same samples were compared with respect to pre- and post-contact crime rates. Table 3.12 summarizes these analyses. Table 3.12 Pre-and post contact criminal history of representativeness samples Initial sample n Pre-contact crimes per year Mean years at risk Mean # crimes / year Mean # violent crimes / yr. Mean # assaults / year Post-contact crimes per year Mean years at risk Mean # crimes / year Mean # violent crimes / yr. Mean # assaults / year Mean # wife assaults / year 5.61 .36 .14 .10 .06 1.68 .42 .21 .16 .09 2.04 .60 .40 .24 .18 15.51 .50 .14 .12 13.69 .51 .17 .15 18.48 .38 .10 .08 272 Potential followup sample 25 Final follow-up sample 27 There were no significant differences between the initial and potential samples on any of the crime indices. This pattern held true in comparisons between the potential and final samples, with one exception: the final sample tended to commit more assaults per year than the potential follow-up sample, t (39) = 2.01, p. = .051. These samples were also compared on pre-contact levels of anger and violence as measured by the Multidimensional Anger Inventory and the “modified” Conflict Tactics scales. The two groups did not differ on these measures. Personality. Finally, the final and potential follow-up samples were compared on a number of scales of the Millon Clinical Multiaxial Inventory (MCMI-II). Again, these data were available both for the AHP and VFVP samples. The data, presented as scale scores (standardized scores) are presented in Table 3.13. Scale scores of 75 or above represent relatively high levels of personality pathology. Table 3.13 MCMI-II mean scale scores for representativeness samples MCMI-II subscale Schizoid Avoidant Antisocial Aggressive/sadistic Passive-aggressive Self-defeating Potential follow-up sample 56.4 66.3 81.0 80.3 74.3 57.4 Final sample 60.0 55.8 83.2 85.0 85.9 69.7 The final follow-up sample had higher levels, on average, of self-defeating personality characteristics, t (90) = 2.30, p. < .03, than the potential sample. Otherwise, there were no differences between the samples on MCMIII measures of personality. Summary of representativeness samples. Taken together, the demographic, criminal history, and personality analyses indicate that the final follow-up sample is essentially similar to the potential follow-up and initial samples. In the absence of compelling evidence to the contrary, an assumption was made that the final follow-up sample was a representative subset of the total population of treated men. With this in mind, subsequent analyses focused exclusively on the follow-up sample. Effects of “program” membership and pre-treatment violence It was recognized that the treatment populations may have differed in some important ways. For example, the majority of men in the AHP program were court-mandated. On the other hand, the majority of the VFVP were self-referred. Analyses of covariance (ANCOVA) were conducted to determine the effect of “program” membership on follow-up reports of abuse. In each ANCOVA, a covariate measure of pre-treatment violence was used to determined its relative contribution in predicting follow-up violence. Three analyses were conducted. The first examined the relationship between program membership on men’s self-reported physical violence as measured by the SVAWS. Men’s pre-treatment violence as measured by the CTS was used as a covariate. The next two analyses examined the relationship program membership to partners’ reports of physical abuse (as measured by the SVAWS) and psychological abuse (PMWI). Partner’s pre-treatment CTS scores were covaried in these latter two analyses. Table 3.14 summarizes these procedures. Table 3.14 Analyses of covariance Dependent variable Main effect tested Covariate Men’s self-reported pretreatment physical violence (CTS) Partner’s reports of men’s pre-treatment physical violence (CTS) Partner’s reports of men’s pre-treatment physical violence (CTS) Men’s self-reported physical Program membership violence (SVAWS) at follow-up (AHP vs VFVP) Partner’s reports of men’s physical violence (SVAWS) at follow-up Partner’s reports of men’s psychological abuse (PMWI) at follow-up Program membership (AHP vs VFVP) Program membership (AHP vs VFVP) None of the main effects or covariates in these analyses were statistically significant. Thus, it appears that levels of follow-up abuse was not determined by program membership or pre-treatment reports of violence. The analyses below explore influence of other factors on follow-up violence. Because program membership did not appear to influence violence, the AHP and VFVP samples were combined for all subsequent analyses (note, also, that the two samples were similar demographically -- see Method). Physical and psychological abuse at follow-up Men’s self-report. Table 3.15 contrasts the pre-treatment and follow-up levels of men’s self-reported anger (Multidimensional Anger Inventory: MAI), and physical violence (modified Conflict Tactics Scale: CTS-Mod). Table 3.15 Men’s pre-treatment and follow-up self-reported anger and physical abuse. Pre-treatment Modified CTS (n=55) 6.83 (SD=5.74) Multidimensional (n=62) Anger Inventory 83.00 (SD=16.18) Follow-up 2.15*** (SD=3.21) 72.19*** (SD=15.84) *** Denotes significance at p < .0001 The mean follow-up period for the men was 26.8 months (SD=13.3). Note that possible MAI scores range from 17 to 135. The potential score on the CTS-Mod ranges from 0 to 27. The emotional/verbal subscale of the PMWI ranges from 19 to 95; dominance/isolation ranges from 28 to 140. The total PMWI scores can range from 47 to 235. It can be seen that the men reported significantly less anger, physical violence, and psychological abuse at follow-up. Partners’ reports. Women were asked to indicate their partner’s pre-treatment and follow-up levels of physical and psychological abuse: results are presented in Table 3.16. The mean follow-up time for partners was approximately the same as that for the men. Table 3.16 Partners’ reports of men’s physical and psychological abuse at pre-treatment and follow-up Pre-treatment Modified CTS (n=12) 9.75 (SD=5.58) Dominance/isolation (n=25) 84.88 (SD=27.07) Emotional/verbal (n=25) 46.76 (SD=15.93) Total psychological abuse (n=25) 135.60 (SD=34.60) Follow-up 2.08** (SD=2.78) 50.40*** (SD=31.99) 23.16*** (SD=15.71) 73.56*** (SD=47.13) ** Denotes significance at p < .01, *** denotes significance at p < .001 Pre-treatment CTS-Mod data was available for 16 female partners; follow-up data were available for 12 of these women. Pre-treatment PMWI data was available for 41 partners; follow-up data were available for 25 of these women. Despite the relatively small numbers in the sample, highly significant statistical differences were observed on all the measures. In sum, partners reported less physical violence and psychological abuse at follow-up. Importantly, partners’ reports of men’s pre-treatment psychological abuse, as assessed with the PMWI (M = 135.6 on total score), were extremely high in comparison to mean scores of other populations. For example, Dutton, Landolt and Starzomski (1995) found a mean score of 93.7 for 50 female partners of male clinical outpatients (not treated for wife assault), and a mean score of 91.0 for 33 female partners of male blue collar workers (not treated for wife assault). However, the mean follow-up score of 73.6 reported by partners in this study was substantially lower than these untreated populations. In order to address the concern that men and female partners may have underestimated their reports of violence and abuse by answering in a socially desirable manner, men’s and partners’ scores on the Modified CTS, the Severity of Violence Against Women Scales (SVAWS), and Psychological Maltreatment of Women Inventory were correlated with the Marlowe Crowne Social Desirability Scale. Table 3.17 Correlations between measures of social desirability and measures of abuse CTS-Mod Marlowe Crowne, men, n=65 Marlowe Crowne, women, n=44 - 0.08 - 0.18 SVAWS - 0.03 - 0.18 PMWI - 0.16 n/a None of the correlations were statistically significant. Specifically, men’s self-report of physical abuse does not appear to be related to social desirability as measured by the Marlowe-Crowne instrument. Similarly, women’s reports of their partners’ physical and psychological abuse were not related to social desirability. It is, of course, impossible to conclude that men and women are not under-reporting abuse. Interestingly, however, there is a weak, non-significant, relationship between higher abuse scores and lower social desirability scores. Another interesting finding is that partner’s total SVAWS and PMWI scores are significantly correlated (r=.78, p <.001). This indicates, that for this sample of women, changes in physical abuse are highly related to changes in psychological/emotional abuse. Thus, the men who are abusive towards women physically are often abusive to them psychologically and emotionally. Importantly, lower levels of physical abuse are correlated with lower levels of psychological abuse. This appears to be contrary to the widely held belief that reductions in physical abuse are followed by increases in psychological or emotional abuse. Personality variables and follow-up abuse T-scores on the MCMI-II subscales were correlated with partner reports of physical violence and psychological abuse. Table 3.18 shows several statistically significant relationships. Table 3.18 Correlations between MCMI-II subscales and partner reports of abuse MCMI-II Subscale SVAWS CTS-Mod PMWI Dominance /isolation .35 .42 .41 .38 .37 .25 PMWI Emotional /verbal .22 .35 .47 .35 .39 .32 Total PMWI .31 .41 .44 .38 .39 .28 Schizoid Avoidant Antisocial Aggressive/sadistic Passive-aggressive Self-defeating .41 .42 .38 .38 .32 .34 .29 .31 .30 .32 .23 .26 r=.30, p < .05; r=.38, p < .01; r=.47, p < .001 Further analyses were required to determine which personality variables were most strongly related to levels of post-treatment violence and abuse. Two multiple linear regression analyses were performed to determine the degree to which these personality variable(s) predicted levels of post-treatment violence and abuse. Linear regression. Six MCMI-II scales measuring personality disorder and the Borderline Personality Organization scale (BPO) were regressed on partner’s follow-up SVAWS scores (physical violence). The stepwise method of exclusion was utilized. The only variable left in the regression equation was the total BPO score. This variable accounted for 23% of the variance in follow-up violence scores, R2 = .234, p < .001. An identical procedure was used to predict partner’s reports of psychological abuse at follow-up. Two variables remained in the regression equation, avoidant personality and antisocial personality. accounted for 28% of the variance in the women’s reports of psychological abuse, R2 = .276, p < .002. The two Discussion This project has made two important contributions to the treatment outcome literature. First, it examined criminal histories of four groups of men with prior convictions for wife assault: Completers, Noncompleters (or Drop outs), No shows, and Rejects. This assessment was based on national police arrest records (CPIC), an admittedly insensitive measure of all assaults. The follow up period for these individuals ranged from four months to eleven years. Previous studies had not tracked men for such a long period. The longest previous known follow-up was two and a half years (Dutton, 1987). Second, the current project has attempted to ascertain whether certain men, on the basis of their psychological profile, might be better (or worse) candidates for treatment. The psychological profiles have focused on personality disorders found in previous studies to be both prevalent amongst abusive men and statistically related to abusiveness. In particular we focused on borderline personality, antisocial personality and avoidant personality disorders, all which have some prior established relationship to abusiveness. Part 1: Criminal History Follow-up Regression and survival analyses (time to first assault) Turning first to the criminal history data, we found in regression analyses of recidivist assault, that group completion and court mandated referral did not significantly account for recidivist assault. Only precontact crimes were significant predictors of recidivism. This is consistent with the axiom, existing in the literature, that past criminal behavior tends to be a robust predictor of future criminal behavior. In this study, men who completed group treatment "survived" somewhat longer until first assaults than did dropouts until first assault (a difference of 148 days) but the difference was not statistically significant. Similarly, comparisons amongst all four groups were not significant. Also, men who never presented for assessment (no shows) actually had the longest survival time of all four groups. This result is intriguing, but difficult to interpret since little is known about the no show group (i.e., they were never assessed). One possible interpretation of this result is that the no show group could have been perceived by probation officers to be lower risk (perhaps correctly), and therefore allowed to avoid treatment. It was also found that voluntary men survived longer until first assault. Voluntary men survived for an average of 2,320 days (about 6.4 years) while court mandated men survived 1,435 days (about 3.9 years). In general, the conclusion of the survival analyses do not support a case for treatment effects for court mandated men. The finding that survival was better for voluntary men may reflect differences in initial motivational levels between the two groups. Voluntary men typically come to treatment with a somewhat higher level of responsibility for violence and lower levels of denial. In the present study, a subgroup of voluntary of men (n =24) were perceived by therapists to have much higher levels of motivation and acceptance of responsibility than court-referred men. Thus, it is likely that in many cases, although their attendance might be prompted by an ultimatum from their partner, voluntary men are typically past the "precontemplation" phase described by Prochaska et al.(1992). Court referred men, on the other hand frequently present in the precontemplation phase at initial assessment. Therapists report their major and initial treatment task to be building motivation in this group. Time to last assault The Prochaska et al. (1992) spiral model of change described in the introduction suggests than an important treatment outcome measure would be time to last assault. In other words, does treatment shorten the "assaultive career" of these men? We attempted to test this model by assessing time to last assault. Unfortunately, from an statistical point of view, very few men in our sample committed multiple recidivist assaults (rejects 6/32, (20%), no-shows 8/91 (9%), drop outs 18/167 (11%), completers 9/156 (6%). Hence, a time to last assault analysis would not be informative, based on such small samples. In order to test the Prochaska model, an initial sample in the 4000+ range would be required (our initial sample was 446) to provide adequate multiple recidivist sample sizes by group. Recidivism ratios Survival analysis is not necessarily the most sensitive or complete measure of recidivism since it tells us only the time lapse until first repeat offense but not how many offenses may occur after the first re-offense. To establish the total amount of repeat assaults, ratios were calculated based on the total number of repeat assaults by each group, divided by the sample size for that group. When these ratios are calculated for each group a somewhat more favorable picture of treatment effects emerges (see Table 3.6). The following ratios were calculated for all assaults and for woman assaults respectively: No shows .40, .23; Rejects .81, .29; Dropouts .55, .50; Completers .32, .23. Hence, when total number of repeat assaults are calculated, Completers have the lowest overall ratios of all groups. Similar differences are observed for annual assault rates. Only No shows, with an identical woman assault ratio are comparable. This ratios become potentially important when they are projected to rates per thousand clients (which is the approximate number of men treated in British Columbia per year through all programs). A group of Rejected men, for example, would generate 810 new assaults in a eleven year period, Dropouts would generate 550, No shows 400, and Completers 320. For woman assaults, the comparable figures would be Rejects 290, Dropouts 500, No shows 230 and Completers 230. Viewed from this perspective, treated men commit fewer assaults than the others. It is important to remember, however, that important differences might exist between the participation groups, and without a randomized design, it is difficult to conclude that it is treatment per se that accounts for the differences. Moreover, we found that when the above ratios were corrected for time at risk, there were no significant difference between recidivism rates between groups. Actual number of assaults, of course, is underestimated by arrest. Our sample of women interviewed post treatment indicated that for every arrest, there were 35 actual assaultive actions reported by the victim (as assessed with the Severity of Violence Against Women Scale). Hence, while the difference between Dropouts and Completers may be 270 arrests, this may indicate a difference of 270 x 35 assaultive acts. Whether treatment groups are successful or not depends on which statistical measure one wants to give the greatest credence. Regression analyses and survival functions do not show significant improvements in post treatment drops in recidivism attributable to treatment. A recidivism ratio, on the other hand, shows a nonsignificant trend for treatment completers to commit fewer offenses than non-completers. It is important to keep in mind that Rosenthal (1983) has argued that effect size (the size of the difference between treated and untreated groups) is more important than statistical significance. The finding, based on recidivism ratios, that group completers did better than men in other groups can be viewed as occurring for at least three reasons. The first is that the content of the treatment groups had an ameliorative effect on their violence, the second is that they exhibited higher motivation to change as demonstrated by their attendance, the third is that this group has a higher employment rate. This third factor, referred to as "marginality" by Sherman et al. (1992) provides a rough index of what the man has to lose by repeated infractions of the law. Sherman et al. (1992) found that arrest deterred recidivist assault except for a marginal group who was more likely to be unemployed and living common law with their partner. Dutton (1994b) argued that this finding exemplified how criminal justice system intervention could only succeed within parameters established by the broader sociopolitical context. One final note about treatment outcome is anecdotal. One man in the completer group had six recidivist assaults. He was a man whom the therapists suspected had neurological problems requiring intervention beyond the scope of the treatment group. They decided to take him as client anyway because other avenues of treatment were not available to him. He single handedly accounted for one fifth of all post treatment wife assaults in the court mandated completer group. Without him the post treatment recidivism ratio would have been .16. Should therapists be put into a position of demonstrating group outcome success, one consequence may be that men such as this will be denied treatment because they are too high risk and may inflate failure rates. It is also true that outliers such as this man existed in all groups. One man in a dropout group had 11 recidivist assaults. Surely, from a policy perspective these men represent not treatment problems but problems in the systems' failure to respond with increasing severity to what are, in effect, serial batterers who are probably untreatable. Part 2. Personality Characteristics and Treatment Outcome Reports of change with treatment Men completing treatment reported pre-post treatment drops in both anger and use of violence (Table 3.15). Typically, such reports are viewed as potentially self serving. More weight is put on partners' reports of changes in the men's behavior. As shown in Table 3.16, partner’s reports also show significant drops in the man's use of abusiveness after treatment. This was true for both physical and emotional abuse and for his use of controlling behaviors. While these reports are available only for a self selected group, this group is demographically and psychologically similar to other, non-contacted men. Pre-treatment assault did not predict post treatment assault. One possible interpretation of this interrupted prediction is that the combination of criminal justice system intervention and treatment has had some impact on assaultiveness. randomized design however, the distinct contributions of each component cannot be assessed. Without a Personality and outcome: Implications for treatment The present study assessed whether aspects of the clients personality influenced post treatment outcome. Three categories of perpetrators' personality had significant effects on women's reports of post-treatment abusiveness. Borderline personality organization was most strongly related to post treatment abusiveness assessed on the Severity of Violence Against Women Scale, while antisocial and avoidant personality were most strongly related to post treatment abusiveness assessed by the Psychological Maltreatment of Women Inventory. Dutton (1994a, 1995) and Dutton & Starzomski (1993, 1994) described an "abusive personality " prone to cyclical tension buildups and abusiveness. The original, central aspect of this personality was borderline personality organization (BPO). In this empirical work BPO assessments of men were significantly related to their wives' reports of both physical and verbal abusiveness. Again, in the current study, BPO was strongly related to abusiveness assessed post treatment on the Severity of Violence Against Women Scale, an omnibus measure of a variety of forms of abusiveness. When post treatment abusiveness was assessed via the Psychological Maltreatment of Women Inventory, however, antisocial and avoidant personality were the strongest predictors. Dutton (1994a) found avoidant personality to be the second strongest contributor to a discriminant function on abusiveness after BPO (and before anger). Hart, Dutton & Newlove (1993) found a high prevalence of all three forms of personality disorder in a group of assaultive men. In assessing what percentages of assaultive men met criterion (>84) on the MCMI-II for each disorder, they found avoidant personality in 35% of men, antisocial in 44% and borderline in 38%. Some writers (Dutton 1988, Saunders 1992) have suggested that three distinct groups of assaulters may exist and the descriptions given by these writers correspond to avoidant (overcontrolled), antisocial (generally violent) and borderline (emotionally volatile) categories. In the present study, these categories again appear, this time as predictors of post treatment abusiveness. One implication of this finding, is that treatment groups regularly assess men prior to treatment using a standardized global assessment instrument such as the MCMI-II. Various forms of individualized treatment might be considered for men who show elevations on personality disorder profiles. Linehan (1987), for example, has employed a behavior therapy technique for borderlines that could be incorporated into most cognitive-behavioral treatment models. Avoidant personality presents a somewhat different clinical picture. The Diagnostic and Statistical Manual of the Mental Disorders (DSM-IV) describes avoidant personality disorder as avoiding social contact that may generate shame or humiliation. A recent study by Dutton, Starzomski & van Ginkel (1995) showed shame to be a prominent feature of abusiveness. Abusive men reported significantly higher rates of shaming experiences in childhood. Recollections of shaming experiences were significantly correlated with borderline personality measured on the MCMI-II (r = .52, p <.001), antisocial personality (r = .33, p <.001), and avoidant Personality (r = .32, p <.001). They were also significantly correlated with what is sometimes referred to as "negativism" (passive-aggressive and self-defeating personality) but not with other measures of personality disorder. Typically these shaming experiences were comprised of public humiliations by parents, random punishment and global criticisms (i.e., of the whole self). Wallace & Nosko (1993) have described the treatment issues involved in working with shame in groups of abusive men. Having to attend a group and disclose violence can itself be shame inducing. The exercise of confession, typical in most treatment groups can allow men to overcome their shame through what the authors call "vicarious detoxification"; listeners encounter their own shame as each member of the group confesses. Projective identification allows men to encounter parts of themselves otherwise split off through hearing other men's confessions and sets the stage for working through shame issues and "owning" violence and abusiveness. Furthermore, the requirement that each man confess builds group solidarity. Eventually, anger, rage and violence, the original defenses against shame become less necessary. These issues may require special attention in all groups with high levels of early shaming experiences such as borderlines, antisocial personality and avoidant personality. Of course, antisocial men present a more complicated treatment problem. Hare (1995) argues that there is a high degree of overlap between antisocial personality and psychopathy. The former includes behavioral indices such as frequent law breaking while the latter refers to a more psychological profile comprised of lack of empathy and emotional shallowness. Since so many antisocial personalities are also psychopathic, and since psychopaths are poor treatment risks, the question arises whether antisocial men can be helped by group treatment at all. If they cannot, should they be screened out of treatment? Would treatment effects be greater if antisocial clients were dropped from the Completer groups? At present we have no answers to these questions. An answer is possible however, in future research. The present study, which has assessed personality disorders in treated men since 1990 could be used as a basis for future assessment of these questions by examining criminal histories of the present cohort in 1997. At that time we could cross index recidivism by personality profile. Even our present study with its reliance on reports of self selected samples hints at an answer. The "big three" personality profiles of abusiveness (borderline, antisocial and avoidant) represent difficult clients and are more likely to abuse post treatment. While other research has identified this triad of abusive personality types (e.g. Dutton 1988, Saunders 1992), it had not been related to treatment outcome until the present study. References American Psychiatric Association. (1980). Diagnostic and statistical manual of mental disorders (3rd ed.). Washington, DC: Author. American Psychiatric Association. (1987). Diagnostic and statistical manual of mental disorders (3rd ed., revised). Washington, DC: Author. Beck, A.T. (1976). Cognitive theory and the emotional disorders. New York: International Universities Press. Bower, S.A., & Bower, G.H. (1976). Asserting yourself: A practical guide for positive change. Reading, MA: Addison-Wesley. Browning, J. J. (1984). Stopping the violence: Canadian programmes for assaultive men. Ottawa: Health and Welfare Canada. Burns, N., Meredith, C., & Paquette, C. (1991). Treatment programs for men who batter: A review of their success. Report for the Policy, Programs, and Research Sector of the Department of Justice Canada. Coleman, D.H., & Straus, M.A. (1986). Marital power, conflict, and violence in a nationally representative sample of American couples. Violence and Victims, 1(2), 141-157. Crowne, D. P., & Marlowe, D. A. (1960). A new scale of social desirability independent of psychopathology. Journal of Consulting Psychology, 24, 349-354. Dobson, K.S., & Block, L. (1987). Historical and philosophical bases of the cognitive-behavioral therapies. In K.S. Dobson (Ed.), Handbook of cognitive-behavioral therapies. New York: Guilford Press. Dutton, D.G. (1981). The criminal justice response to wife assault. Ottawa: Solicitor General of Canada, Research Division. Dutton, D.G. (1986). The outcome of court-mandated treatment for wife assault: A quasi-experimental evaluation. Violence and Victims, 1(3), 163-175. Dutton, D.G. (1987). The criminal justice response to wife assault. Law and Human Behavior, 11(3), 189206. Dutton, D.G. (1988). Profiling wife assaulters: some evidence for a tri-modal analysis. Violence and Victims, 3(1), 5-30. Dutton, D.G. (1994a). Behavioural and affective correlates of Borderline Personality Organization in wife assaulters. International Journal of Law and Psychiatry, 17(3), 265-277. Dutton, D.G. (1994b). Intervention strategies in wife battering. Paper presented at the International Conference on Violence in the Family, Vrije University, Amsterdam. Dutton, D.G. (1995). The domestic assault of women: Psychological and criminal justice system perspectives. (Rev. ed.). Vancouver: University of British Columbia Press. Dutton, D.G., Hart, S.D., Kennedy, L.W., & Williams, K.R. (1992). Arrest and the reduction of repeat wife assault. In E.S. Buzawa and C.G. Buzawa (Eds.). Domestic violence: The changing criminal justice response (pp. 111-127). Westport, CT: Greenwood Press Inc. Dutton, D. G., & Hemphill, K. J. (1992). Patterns of socially desirable responding among perpetrators and victims of wife assault. Violence and Victims, 1, 29-40. Dutton, D.G. Landolt, M., & Starzomski, A. (1995). Cross-validation of the Propensity for Abusiveness Scale. Report to the Solicitor General of Canada, Research Division. Dutton, D.G. & Starzomski, A. (1993). Borderline Personality Organization in perpetrators of psychological and physical abuse. Violence and Victims, 8(4), 327-338. Dutton, D.G. & Starzomski, A. (1994). Personality differences in court-referred and self-referred wife assaulters. Criminal Justice and Behavior, 21(2), 203-222. Dutton D. G., &. Starzomski, A., & van Ginkel, C. (1995). The role of shame and guilt in the intergenerational transmission of abusiveness. Manuscript submitted for publication. Eddy, M.J., & Meyers, T. (1984). Helping men who batter: A profile of programs in the U.S. Texas Council on Family Violence, Arlington, Texas. Edelson, J.L., & Tolman, R.M. (1992). Intervention for men who batter: An ecological approach. Newbury Park, CA: Sage. Eysenck, H. (1969). The effects of psychotherapy. New York: Science House. Fasteau, M.F. (1974). The male machine. New York: McGraw-Hill. Ganley, A. (1981). Participant's manual: Court mandated therapy for men who batter: A three day workshop for professionals. Washington, DC: Center for Women Policy Studies. Ganley, A., & Harris, L. (1978). Domestic violence: Issues in designing and implementing programs for male batterers. Paper presented at the 86th annual convention of the American Psychological Association, Toronto. Gendreau, P., & Ross, R. (1980). Correctional potency: Treatment and deterrence on trial. In R. Roesch & R. Corrado (Eds.). Evaluation and criminal justice policy. Beverly Hills: Sage Publications. Hamberger, K. L., & Hastings, J. E. (1988). Personality characteristics of spouse abusers: A controlled comparison. Violence and Victims, 3, 31-48. Hamberger, K. L., & Hastings, J. E. (1993). Court-mandated treatment of men who assault their partner: Issues, controversies, and outcomes. In N.Z. Hilton, (Ed.), Legal Responses to Wife Assault: Current Trends and Evaluation. Newbury Park: Sage. Hare, R.D. (in press). Psychopathy: A clinical construct whose time has come. Criminal Justice and Behavior. Hart, S. D., Dutton, D. G., & Newlove, T. (1993). The prevalence of personality disorder amongst wife assaulters. Journal of Personality Disorders, 7(4), 328-340. Lazarus, R.A., & Averill, J.R. (1972). Emotion and cognition: With special reference to anxiety. In C. D. Spielberger (Ed.), Anxiety: Current trends in theory and research (Vol. II). New York: Academic. Lerman, L.G. (1981). Prosecution of spouse abuse: Innovations in criminal justice response. Washington, DC: Center for Women Policy Studies. Linehan, M. (1987). Dialectical behavior therapy for Borderline Personality Disorder. Bulletin of the Menninger Clinic, 51(3), 261-276. Mahoney, M.J., & Thoreson, C.E. (1979). Self-control: Power to the person. Monterey, CA: Brooks/Cole. Marshall, L.L. (1992). Development of the Severity of Violence Against Women Scales. Journal of Family Violence, 7(2), 103-121. Meichenbaum, D. (1977). Cognitive behavior modification. New York: Plenum Press. Millon, T. (1992). Millon Clinical Multiaxial Inventory: I and II. Journal of Counseling and Development, 70, 421-426. Ng, S. H. (1980). The social psychology of power. New York: Academic Press. Nourusis, M.J. (1992). SPSS for Windows: Advanced statistics, Release 5. Chicago: SPSS Inc. Novaco, R. (1975). Anger control: The development and evaluation of an experimental treatment. Lexington, MA: Lexington Books. Oldham, J., Clarkin, J., Appelbaum, A., Carr, A., Kernberg, P., Lotterman, A., & Haas, G. (1985). A selfreport instrument for borderline personality organization. In T. H. McGlashan (Ed.), The borderline: Current empirical research (pp. 1-18). Washington, DC: American Psychiatric Press. Orne, M. (1962). On the social psychology of the psychological experiment: With particular reference to demand characteristics and their implications. American Psychologist, 17, 776-783. Orne, M. (1969). Demand characteristics and the concept of quasi-controls. In R. Rosenthal & R. Rosnow (Eds.), Artifact in behavioral research. New York: Academic Press. Pleck, J.H. (1981). The myth of masculinity. Cambridge, MA: M.I.T. Press. Prochaska, J.O., DiClemente, C.C. & Norcross, C.C. (1992). In search of how people change: Application to addictive behaviors. American Psychologist, 47(9), 1102-1114. Rachman, S.J., & Wilson, G.T. (1980). The effects of psychological therapy (2nd Ed.). Oxford: Pergamon Press. Rosenfeld, B.D. (1992). Court-ordered treatment of spouse abuse. Clinical Psychology Review, 12, 205-226. Rosenthal, R. (1983). Assessing the statistical and social importance of the effects of psychotherapy. Journal of Consulting and Clinical Psychology, 51(1), 4-13. Saunders, D. G. (1992). A typology of men who batter women: Three types derived from cluster analysis. American Journal of Orthopsychiatry, 62, 264-275. Schulman, M. (1979). A survey of spousal violence against women in Kentucky. Washington, DC: U.S. Department of Justice, Law Enforcement. Selzer, M. L. (1971). The Michigan Alcoholism Screening Test: The quest for a new diagnostic instrument. American Journal of Psychiatry, 127, 89-94. Sherman, L.W., Schmidt, J.D., Rogan, D.P., Smith, D.A., Gartin, P.R., Cohn, E.G., Collins, D.J. & Bacich, A.R. (1992). The variable effects of arrest on criminal careers: The Milwaukee domestic violence experiment. Journal of Criminal Law and Criminology, 83(1), 137-161. Sherman, L.W., & Berk, R.A. (1984). The specific deterrent effects of arrest for domestic assault. American Sociological Review, 49, 261-272. Shore, M., & Massimo, J. V. (1979). Fifteen years after treatment: A follow up study of comprehensive psychotherapy. American Journal of Orthopsychiatry, 49(2), 240-245. Siegel, J. M. (1986). The Multidimensional Anger Inventory. Journal of Personality and Social Psychology, 51, 191-200. Skinner, H. A. (1982). The Drug Abuse Screening Test. Addictive Behaviors, 7, 363-371. Sonkin, D.J., & Durphy, M. (1982). Learning to live without violence: A handbook for men. San Francisco: Volcano Press. Sonkin, D. J., Martin, D., & Walker, L. E. (1985). The male batterer: A treatment approach. New York: Springer. Standing Committee on Health, Welfare, and Social Affairs. (1982). Report on violence in the family: Wife battering. House of Commons, Ottawa. Straus, M. A. (1979). Measuring family conflict and violence: The Conflict Tactics Scale. Journal of Marriage and the Family, 41, 75-88. Straus, M. A., Gelles, R. J., & Steinmetz, S. (1980). Behind closed doors: Violence in the American family. New York: Anchor Press. Tolman, R.M. (1989). The development of a measure of psychological maltreatment of women by their male partners. Violence and Victims, 4(3), 159-177. U.S. Commission on Civil Rights. (1978). Battered women: Issues of public policy. Washington, DC: U.S. Government Printing House. Waldo, M. (19880. Relationship enhancement counseling groups for wife abusers. Journal of Mental Health Counseling, 10, 37-45. Wallace, R., & Nosko, A. (1993). Working with shame in the group treatment of male batterers. International Group Psychotherapy, 43(1), 45-61. Yalom, T.D. (1975). The theory and practice of group psychotherapy. New York: Basic Books.