Research Reports

Assessing Treatment Change among Family Violent Offenders:
Reliability and Validity of a Family Violence Treatment Assessment Battery

Kelley Blanchette, David Robinson, Christine Alksnis and Ralph Serin
Research Branch,
Correctional Service Canada
December, 1997

Executive Summary

While many new treatment programs have been developed for offenders who have been abusive toward members of their family, there is little agreement regarding the measures that should be used for assessing impact of family violence programs on offenders. This paper investigates the utility of three self-report scales for measuring treatment-related changes in men who have participated in programs for family violent offenders. Correctional Service of Canada (CSC) implemented six Family Violence Treatment Demonstration Projects between 1990 and 1993 in both institutional and community settings. The treatment programs reflected both cognitive-behavioural and feminist models of family violence treatment. Within this broad framework, each demonstration site adopted its own specific approach to treatment. Nevertheless, the overriding aim of all six programs was to reduce or eliminate abuse of family members in all its forms. This validation study is based on data pooled from these projects between April 1, 1994 and March 31, 1995.

In 1994, a standard set of assessment instruments was established across the six sites in order to allow for a comparison of research findings across the treatment programs. The three principle instruments were administered as self-report questionnaires. The instruments measured attitudes and behaviours related to woman abuse and various dimensions of anger arousal and expression. The fourth instrument (Balanced Inventory of Desirable Responding (BIDR)) measured the tendency to present oneself in a socially desirable manner while responding to questionnaire items.

In addition to the self-report measures, a detailed intake information form and case closure form were completed by program delivery staff. The intake instrument included demographic factors, criminal history variables, measures of family violent behaviours, and items related to a variety of criminogenic needs. The closure instrument was used to record information about the offenders’ participation in the program and the therapist’s judgements about progress made during treatment. The intake and closure information supplied by program delivery staff was used in the process of validating the self-report instruments.

The three self-report instruments and the social desirability measure are listed below along with the sub-scales used to measure different dimensions within each scale:

The usefulness or validity of the three principle self-report measures was assessed by examining the following criteria:

1. susceptibility of the self-report instruments to socially desirable responding
2. ability of the self-report instruments to detect changes in offenders from the beginning to the end of the program
3. ability of the self-report instruments to detect early termination or drop-out from the program
4. ability of the self-report instruments to detect treatment gains as rated by the therapists who conducted the treatments
5. ability of the self-report instruments to exhibit reasonable patterns of correlation with various demographic variables, criminogenic need, and childhood experiences of victimization and/or witnessing of violence

The sample was comprised of 336 men. Half participated in institutionally-based programs while the other half participated in community-based programming. About 80% of the sample reported perpetrating one or more types of abuse against women. Sixty-two percent had abused their current partners while about half (51. 9%) indicated that they had abused previous partners. Some offenders also indicated that they had abused their children and/or stepchildren (13. 2%). Less than half of the men (45. 5%) had experienced any legal consequences or sought treatment for their abusive behaviour. About two-thirds of the treated offenders (67. 8%) had witnessed or experienced violence in their family of origin.

Before assessing the degree to which the three instruments detected treatment-related change in the family violent offenders, special analyses were conducted to determine whether there was any association between these measures and the social desirability measure (BIDR). The social desirability measure served as a check on the extent to which offenders presented their attitudes and behaviours in a positive light, thus disguising their true characteristics. Two measures of socially desirable responding are measured by the BIDR: impression management (i. e. , willful desire to present oneself in an overly favorable light to impress or mislead others) and self-deception (i. e. , innocent but overly positive representation of the self). The analyses showed that most of the scales were susceptible to socially desirable response biases. The pattern of results indicated that offenders were inclined to downplay the degree to which they were abusive. Generally, the effects of impression management on the scales were more substantial than the effects of self-deception.

The Impression Management (IM) subscale of the BIDR was significantly correlated with 3 of 5 IBWB subscales, 3 of 4 ICB subscales and 4 of 5 MAI subscales. At post-test, the influence of socially desirable responding was less pronounced but did not disappear completely. Specifically, 4 of the 5 MAI subscales were correlated with the IM subscale at post-test. It is not clear whether these findings were indicative of offenders’ tendencies to deny and minimize the extent and seriousness of their abusive behaviour or whether the offenders were attempting to convince therapists that treatment had been effective.

These findings suggest that assessments at pre-treatment and post-treatment should be interpreted with caution for offenders who score high on the BIDR sub-scales. Importantly, the BIDR (social desirability) scores of offenders in the current study were comparable to those found in research with other groups of offenders (Kroner & Weekes, 1996). Notably, however, there are no published clinical guidelines regarding what constitutes a "high" BIDR score for offenders. It is therefore suggested that those offenders who score one standard deviation above their group mean be considered "high" scorers. Accordingly, other self-report assessments completed by those who score high on social desirability should be interpreted with additional caution.

At pretest, men in the institutional treatment programs exhibited more attitudes supporting family violence than their community couterparts. Specifically, the IBWB total, and 4 of 5 subscale scores, showed significant between-group differences. Compared with community-based men, those in institutions were more likely to agree that wife beating is justified and that wives gain from abuse. Accordingly, institutional men were less likely to believe that abusers should be punished, and that abusers are responsible for their violent behaviour. Although at post-test IBWB total scores maintained significant between-group differences, those for all subscales disappeared. It should be noted, however, that a disproportionate number of community-based men dropped out of treatment. The lower drop-out rate for offenders in institutional programs highlights the utility of offering treatment while incarcerated.

Two of the three anger/violence inventories, the IBWB and the MAI, showed some promise for measuring treatment-related change. Specifically, offenders’ scores changed over time, suggesting that the scale is reasonably dynamic. IBWB total scores, plus "justify", "punish", and "responsible" subscales demonstrated changes in offenders’ attitudes from pre- to post-treatment. The most robust effect was found in the IBWB total scores, demonstrating that offenders endorse fewer attitudes supporting family violence post-treatment. One MAI subscale detected changes in abusers’ attitudes and behaviours after program participation. A significant pre- to post-test difference score on the "anger in" subscale is suggestive of positive (post-treatment) changes in attitudes and/ or behaviours related to the expression of anger.

The relationship between pre-test/background information and post-test scores on the IBWB were consistent with expectations regarding how abusers deal with conflict and with the intergenerational transmission of abuse. Men who reported using reasoning to resolve conflicts were less likely to endorse attitudes supporting wife beating. Those who were unable to express their anger or who had experienced a greater variety of abuse themselves or who had been charged/incarcerated for family violent acts were significantly more likely to endorse wife beating. Anger (as measured by the MAI) was associated with greater use of emotional and sexual abuse tactics (as measured by the ICB), a greater variety of woman abuse perpetrated, and the presence of substance abuse problems. Substance abuse was also significantly related to pre- to post-program difference scores on IBWB and MAI subscales. However, further investigation revealed that the effect was due to dissimilar pre-test scores by substance abusers and non-abusers rather than to differential ability to benefit from treatment.

The self-report measure of abusive behaviour, the ICB, failed to demonstrate any pre- to post-test changes. It seems plausible that no decrease was detected because the endorsement of controlling tactics was already quite low at intake. This low endorsement could be due to the operation of denial/minimization or to social desirability response bias. Furthermore, the proportion of the sample that had data on this measure at both pre- and post-test was less than 25% -- the relatively small sample size may have prevented detection of any treatment-related differences that did exist. There were also some problems with the reliability of the scales in that offenders were often inconsistent in admitting to ICB behaviours from pre-test to post-test. Nevertheless, as a pre-test measure, the ICB may be helpful in the clinician’s assessment of the client’s level of admission or denial of abusive behaviour prior to treatment.

With respect to early program termination, just over three-fifths of the sample completed the program (61. 1%), the remaining 38. 9% dropped out early. The anger/violence measures were less successful at predicting early program termination than they were at measuring change. The "wife beating is justified" subscale of the IBWB and the "anger out" subscale of the MAI were significantly correlated to early program termination. Those who terminated their program participation early had lower (pretest) scores on the "justify" subscale, indicating that at the outset, they were less likely to view wife beating as justified. Program drop-outs were also more likely to express their anger than those who completed the program according to their pre-test "anger out scores" on the MAI. However, it should be noted that the "anger out" subscale of the MAI consists of only two items, and internal reliability was poor (a =.11). As such, its utility for predicting attrition from family violence programming is dubious at best.

Finally, early program termination was also reliably associated with two background variables: number of victimization experiences in childhood, and number of current problems. Specifically, those who did not complete the program reported fewer experiences of childhood victimization than the program completers. Multi-need offenders (as measured by number of current problems) were also more likely to terminate their program participation early, suggesting that they are at greater risk for negative outcomes and may require higher intensity services, or a multi-faceted approach.

Of all attitudinal and behavioural measures, the only reliable predictor of treatment outcome (therapist-rated success) was the IBWB total scale score. These findings indicate that high endorsement of attitudes supporting family violence (pre-treatment) are associated with more positive ratings post-treatment. However, in interpreting this finding it is also important to consider that those who showed more negative attitudes pre-treatment also had a wider margin for showing improvement at post-test. The modest results on the therapist outcome ratings may be due to the small number of men for whom this data was available (n=72). This represents only 38% of the 190 men who completed treatment.

Although the present study was severely hampered by missing data, the results obtained thus far show some promise for continued use of the four measures in family violence research and practice. It is noteworthy that the IBWB, the ICB, and the MAI are diverse (i. e. , not highly interrelated) in that they appear to measure distinct aspects of family violence attitudes and behaviour. Moreover, some subscales of the IBWB and the MAI demonstrated an ability to measure change over time (pre- to post-treatment). The design of the current study precludes the ability to draw firm conclusions regarding whether these changes actually reflected a treatment effect. However, the observed changes in family violent attitudes/behaviour were consistent with anticipated treatment effects. Specifically, all pre-to post-treatment changes were in the expected direction, with men showing improved attitudes/behaviour post-treatment. Prospective research is recommended to confirm the efficacy of these tools in the assessment of treatment progress for individual offenders.

While the pre-test/post-test data reported here provide only minimal evidence of program effectiveness, the results suggest that offenders who participate in family violence programming develop attitudes, over the course of treatment, which are less supportive of abuse. An important remaining question concerns the extent to which diminished endorsement of attitudes condoning family violence predict reductions in actual abuse behaviour.

Table of Contents

• Executive Summary
• Table of Contents
• LIST OF TABLES
• Introduction
• Method
  • Description of the Sample
  • Comparisons with normative samples of offenders
  • Abuse experienced and perpetrated by clients
  • Instruments
  • Closure Information
• RESULTS
  • Socially Desirable Responding
  • Pre-program to post-program differences
  • Early termination
  • Therapist-rated outcome
• References
• Appendix A
• Appendix B
• Appendix C
• Appendix D

LIST OF TABLES

• Table 1: Sample Size by Treatment Site
• Table 2: Background Information
• Table 3: Childhood History and Victimization Experiences
• Table 4: Abuse Perpetrated by Offender
• Table 5: Reasons for Failure to Complete Family Violence Program
• Table 6: Relationship between Social Desirability and Violence/Anger Measures
• Table 7: Significant Differences between Pre-test and Post-test on IBWB & MAI
• Table 8: Mean scores on Attitudinal and Behavioural Measures at Pre-test: Institutional and Community Samples

Introduction

The Correctional Service of Canada (CSC) joined a number of federal departments in developing new programs under the federal government’s Family Violence Initiative. The purpose of the initiative was to respond to the needs of individuals affected by violent situations through education, prevention, intervention and treatment. Recognizing that offenders are at a disproportionately high risk of committing domestic violence (Dutton & Hart, 1992; Alksnis & Robinson, 1995; Robinson & Taylor, 1995), CSC introduced a number of interventions for men who act abusively in intimate relationships. These included staff training and workshops, psychoeducation, and offender treatment programs. Intervening with offenders who are at risk of perpetrating family violence is consistent with CSC’s Mission to protect the public and to prepare offenders for successful reintegration into society (CSC, 1995; Rondeau, Brochu, Lemire, Brodeur & Schneeberger, 1994).

In recent years there have been a number of programs developed to treat offenders who have been perpetrators of abuse against family members. However, there is a dearth of research on the effectiveness of such programs. Without sound empirical data, researchers and practitioners have been unable to agree on what types of instruments should be used to measure treatment gains and how to assess the effectiveness of specific programs targeting this population. The current study attempted to contribute to research on the assessment of family violent offenders.

This report provides a summary of findings from assessment data pooled (from April 1, 1994 to March 31, 1995) from six treatment demonstration programs.

These programs were specifically introduced by CSC for offenders identified as at risk for family violence. The Family Violence Demonstration Projects were implemented between 1990 and 1993 in both institutional (n=2) and community (n=4) settings. These treatment programs draw heavily on two major models: the cognitive-behavioural model and the feminist model (Vandenburg & Knoll, 1996; CSC, 1994). The first emphasizes the learned nature of violence and the notion that abusive behaviour can be un-learned. The second involves analysis of the unequal relationship between men and women and the notion that abusive behaviour is an attempt to establish control over a partner via domination techniques (rather than a loss of control over behaviour). Notably, these two approaches are not mutually exclusive. Within this broad framework, each demonstration site adopted its own specific approach to treatment. For instance, Quebec embraced a systematic approach (see Fortin & Devault, 1993) while Newfoundland employed the Readiness Model (see Institute for Human Resource Development, 1994). Despite some variations in treatment delivery across sites, the basic goals of all six programs remained the same: to encourage men to take responsibility for their violent behaviour, to modify sexist attitudes, to enhance the capacity to express emotions and to develop relational abilities. The overriding goal is to reduce or eliminate abuse of family members in all its forms.

Beginning in 1994, the demonstration projects implemented a standard set of instruments in order to compare research findings across the treatment programs (Taylor & Robinson, 1994). Selection of instruments was made on the basis of consensus from the treatment directors of the six demonstration projects. The purpose of this report is to assess these instruments for (i) usefulness in measuring treatment-related change, (ii) correlation with early termination, and (iii) correlation with therapist-rated outcome. In addition, the study was designed to assess the extent to which social desirability response bias contaminates the results of the pre-test/post-test self-reports of offenders. As a further test of the validity of the scales, the study also examined whether the family violence treatment measures demonstrated a reasonable pattern of correlation with demographic, criminogenic need, and family violence history measures.

Although this report is not intended to provide a direct evaluation of program effectiveness (i.e., measure whether the programs reduce family violence), the results contribute to the on-going examination of factors affecting treatment outcomes such as program completion and the assessment of treatment-related

change in offenders (e.g., attitudes, interpersonal skills). For other treatment resistant offender subgroups, treatment compliance is an important intermediate treatment goal (Preston & Murphy, 1997).

An effort was made to obtain detailed information about the characteristics of the men who were enrolled in the treatment programs in correctional settings. It is hoped that by learning more about the characteristics of this population, it will be possible to tailor the treatment delivery system to the multiple needs of this unique group (i.e., increase treatment responsivity). With this goal in mind, information about demographic variables and criminogenic needs (e.g., substance abuse, education/ employment, etc.) was gathered using a program intake information tool (Appendix A). The information from the family violence treatment sample was compared with demographic and criminogenic need data from normative samples of offenders from institutions and community settings. This provided an opportunity to examine how offenders undergoing treatment for family violence problems compare to the larger population of federal offenders.

Finally, a great deal of attention has been directed to the relation between abuse in the family of origin and later aggressive behaviour by the witness/victim (Widom, 1989b; Malinosky-Rummel & Hansen, 1993; Oliver, 1993). This intergenerational transmission of abuse has been detected in both non-offender samples (see Alksnis & Taylor, 1995) and offender samples (Alksnis & Robinson, 1995; McCord, 1983; Dutton & Hart, 1992; Widom, 1989a; Widom, 1991). The link between childhood experiences such as victimization and/or witnessing violence and offenders’ treatment outcomes on the selected instruments was explored in the current investigation.

Method

Description of the Sample

Data were collected by the six family violence treatment programs contracted by CSC to deliver services in community and institutional settings across the five regions of the jurisdiction: Victoria, B.C.; Edmonton, Alberta; Toronto, Ontario; Ottawa, Ontario; Laval, Quebec; and St. John’s, Newfoundland. The Victoria and Laval programs were institutionally based whereas those in Edmonton, Toronto, Ottawa and St. John’s were community based. The combined sample size was 336: 158 (47.0%) from institutions and 178 (53.0%) from community-based programs.

A percentage distribution of the total sample, by treatment site, is presented in Table 1. Data from this section was derived from the Client Information Form, an adaptation of an intake instrument used by the Ottawa program. The Client Information Form has been described by Taylor and Robinson (1995) and is included in Appendix A of this report.

Table 1: Sample Size by Treatment Site

	n	%
Community Sites:	178	53.0
Edmonton	36	10.7
Toronto	63	18.7
Ottawa	47	13.9
St. John’s	32	9.5
Institutional Sites:	158	47.0
Victoria	132	39.3
Laval	26	7.7
Total Sample	336	100

Comparisons with normative samples of offenders

To examine whether offenders in treatment for family violence differ from normative samples of federal offenders, a series of between-group comparisons were conducted. Specifically, chi-square statistics compared groups on variables associated with demographic characteristics and criminogenic need.

The normative (comparison) group was obtained by automation through CSC’s Offender Management System (OMS). Selection criteria included all male federal offenders who: 1) were incarcerated at some point in 1994/95; 2) had completed the Offender Intake Assessment (OIA) process, yielding data for all or most need indicators, and 3) were not included in the project family violence sample. This technique provided a normative sample of 632 federal offenders, 341 (54%) of whom were in the community (i.e., released), and 291 (46%) of whom were incarcerated as of 1995.

Briefly, the OIA is a comprehensive and integrated evaluation of the offender at the time of admission to the federal system. It considers a wide variety of information pertaining to offender demographics, risk and needs. It was therefore regarded as the most appropriate source from which proximal comparative data could be derived. However, since the OIA was first implemented in 1994, one potential limitation to the methodology for obtaining the normative sample is acknowledged: The community group is over-represented by offenders who had served shorter sentences.

Three series of between-group comparisons were performed. First, the entire project sample was compared to the normative sample on 15 demographic and need variables. The second and third series of comparisons focused on community and institutional subsamples, each comparing the family violence (i.e., project) group to the normative group (see Table 2). It is worth noting that 14% of the normative comparison group comprises offenders with a known history of family violence perpetration. A decision was made not to exclude these data, since family violent offenders naturally appear as a subgroup in normative samples of offenders.

As mentioned, three series of analyses compared the family violence sample to the normative sample across several demographic and need variables. Demographic comparisons included: age, language, Aboriginal origins, marital status, dependents, and level of education. Between-group comparisons of needs examined current problems with: finances, employment, education, housing, suicide risk, psychiatric disorder, substance abuse (any), alcohol abuse, and drug abuse.

Taken together, between-group analyses on demographics and needs totaled 15 comparisons. Thus, while an acceptable level of statistical significance is traditionally set at a probability of a =.05, stringent criteria suggest dividing this by the number of comparisons performed. Therefore, for total- and community sample comparisons, the probability to achieve statistical significance was set at .003 (which is a /15). For the institutional groups, three need areas were not relevant (financial, employment, housing), and therefore probability was set at .004 (a /12).

Table 2: Background Information

	Total Sample	Community Samples		Institutional Samples
	% (n)	Project Sample % (n)	Normative Sample % (n)	Project Sample % (n)	Normative Sample % (n)
Age:
Under 25	8.9 (124)	6.9 (58)	18.6 (145)	10.6 (66)	19.3 (88)
25 - 39	64.5 (124)	69.0 (58)	66.9 (145)	60.6 (66)	65.9 (88)
40 and Over	26.6 (124)	24.1 (58)	14.5 (145)	28.8 (66)	14.8 (88)
Language: a,b
English	88.1 (194)	100.0 (100)	66.2 (145)	74.3 (94)	69.0 (88)
French	11.9 (194)	0.0 (100)	33.8 (145)	24.7 (94)	31.0 (88)
Aboriginal origins:	17.2 (232)	16.0 (144)	18.1 (144)	19.3 (88)	15.9 (88)
Current marital status: a,b,c
Single	24.6 (244)	20.8 (149)	49.7 (141)	30.5 (95)	50.0 (86)
Common-law	34.4 (244)	39.6 (149)	38.3 (141)	26.3 (95)	37.2 (86)
Married	12.3 (244)	14.8 (149)	6.4 (141)	8.4 (95)	4.7 (86)
Divorced	13.9 (244)	9.4 (149)	2.1 (141)	21.1 (95)	5.8 (86)
Separated	11.5 (244)	12.8 (149)	3.6 (141)	9.5 (95)	2.3 (86)
Widowed	3.3 (244)	2.7 (149)	0.0 (141)	4.2 (95)	0.0 (86)
Have children at home c	45.3 (179)	58.3 (103)	46.7 (321)	30.3 (76)	56.6 (279)
Highest level of education:
Under grade 8	22.7 (88)	20.7 (58)	25.9 (247)	26.7 (30)	28.6 (210)
Grade 8 or 9	31.8 (88)	37.9 (58)	43.3 (247)	20.0 (30)	40.5 (210)
Completed high school (Grade 12 or 13)	45.5 (88)	41.4 (58)	30.8 (76)	53.3 (30)	31.0 (210)
Some post-secondary	34.5 (235)	29.1 (141)	---	42.6 (94)	---
Current problems:
Financial (on assistance)	38.5 (130)	38.5 (130)	30.0 (313)	n/a	n/a
Employment	46.6 (148)	46.6 (148)	41.6 (341)	n/a	n/a
Housing a,b	8.6 (151)	8.6 (151)	39.9 (316)	n/a	n/a
Education b	15.3 (235)	12.1 (149)	29.8 (322)	20.9 (86)	28.5 (281)
Suicidal thoughts	1.7 (237)	0.7 (141)	3.1 (324)	3.1 (96)	0.4 (278)
Psychiatric disorder	5.7 (248)	3.3 (152)	3.8 (315)	9.4 (96)	2.9 (278)
Substance abuse:
Any (alcohol or drugs) a,b,c	37.3 (298)	22.0 (150)	80.6 (320)	52.7 (148)	77.1 (280)
Alcohol problem a,b,c	26.6 (297)	18.4 (152)	58.0 (320)	35.2 (145)	51.8 (280)
Drug problem a,b,c	26.5 (298)	10.0 (150)	69.1 (320)	43.2 (148)	65.1 (278)

Note: ^a p<.003 for comparison between project and normative (entire) samples;

^b p<.003 for comparison between community subsamples;

^c p<.004 for comparison between institutional subsamples

It is noteworthy that while chi-square analyses on age groups (as shown in Table 2) showed no significant differences, a between-group t-test for continuous variables rendered reliable differences. Specifically, significant differences emerged, where the family violent (project) sample was older than the normative group. Mean age, by group, were 34.5 and 31.4, respectively (p<.0004).

Between-group differences noted in language were an artifact of the sampling procedures. There was no community project sample in the Quebec region, which is reflected in a much higher proportion of French offenders in the normative group.

As shown in Table 2, analyses revealed reliable differences between groups on marital status. Moreover, this was true for all three between-group comparisons: total, community, and institutional samples. Considering the nature of the treatment program, it was not surprising to find that offenders in the project sample were more likely to be either married, divorced, separated, or widowed than those in the normative sample. In turn, this latter group was much more likely to be either single (50%) or living common-law (38%). Interestingly, however, within the institutional subsamples, the normative group was almost twice as likely to have children at home.

While data showed a trend for the project sample to have more education than the normative sample, analyses yielded no statistically significant findings. Nonetheless, it is noteworthy that almost half (45.5%) of the family violent offenders had completed high school, compared with about one-third (31%) of the normative sample.

Finally, data presented in Table 2 also demonstrate fewer collateral problems in the family violence sample, as compared to the normative group. Specifically, statistically reliable between-group differences were noted in the areas of housing, education, and substance abuse. Differences were particularly robust in the area of substance abuse, where the normative sample was about twice as likely to have problems.

Abuse experienced and perpetrated by clients

Program participants were asked about their experiences of abusive behaviour, both as victims in childhood (includes witnessing abuse and being victimized) and as perpetrators in adulthood. This information is presented in Tables 3 and 4.

With respect to abuse suffered, about two-thirds (67.8%) of the men indicated that they had witnessed or experienced violence in their family of origin. Forty-seven percent of clients indicated that they had been physically abused while 17.5% revealed sexual abuse and 51% reported emotional abuse. Over half (56.2%) had witnessed the abuse of their mothers and/or their siblings in childhood. Men from institutionally-based programs reported significantly more abuse than those participating in community-based programs (M_Inst = 3.1, SD=2.1, M_Comm = 1.8, SD=1.9; p<.0001).

Table 3: Childhood History and Victimization Experiences

	%	n/N
Parents Divorced	47.5	116/244
Siblings Removed from Parents’ Care	28.1	68/242
Childhood Experiences of Abuse
Abuse witnessed	56.2	136/242
Father physically abused mother	40.3	96/238
Father sexually abused mother	4.7	11/232
Father emotionally abused mother	49.6	116/234
Offender witnessed abuse of a sibling	27.9	62/222
Victimization	56.4	137/243
Offender physically abused as a child	47.3	114/241
Offender sexually abused as a child	17.5	42/240
Offender emotionally abused as a child	51.1	121/237
Number of different abuse experiences (victimized and/or witnessed)
None	32.2	79/245
1 to 3	34.7	85/245
4 or more	33.1	81/245

The vast majority of offenders in the current sample had been abusive with female partners. About 80% reported perpetrating one or more types of woman abuse; community-based offenders were not significantly different from institutionally-based men with respect to number of abusive acts perpetrated. Sixty-two percent had abused their current partners while about half (52.3%) indicated that they had abused past partners. In total, 81% had abused either a previous partner, their current partner, or both. Some offenders indicated that they had also abused their children and/or stepchildren (13.2%).

A substantial proportion of offenders experienced legal consequences for their abusive actions (40.6% had charges laid against them while 33% were incarcerated for some period of time). Nevertheless, in over half of all cases there were no sanctions, either as treatment or punishment (54.5%). Institutionally-based offenders did not differ from community-based men with respect to the number of consequences received. The number of different interventions/consequences experienced (legal or therapeutic) served as a predictor variable in the analyses that follow, as did the number of different abusive acts perpetrated in adulthood and the number of different childhood abuse experiences reported.

Table 4: Abuse Perpetrated by Offender

	%	n
WOMAN ABUSE PERPETRATED BY OFFENDER
Abused current partner in some way	62.2	150/241
Abused current partner physically	63.3	93/147
Abused current partner emotionally	69.7	106/152
Abused current partner sexually	8.5	10/117
Abused previous partner(s) in some way	52.3	127/243
Abused previous partner(s) physically	49.1	84/171
Abused previous partner(s) emotionally	60.8	104/171
Abused previous partner(s) sexually	6.6	10/152
Abuse continued after separation	10.0	17/170
Number of different woman abuse acts perpetrated
None	19.2	44/229
1 to 3	55.9	128/229
4 to 6	21.8	50/229
7 or more	3.1	7/229
CHILD ABUSE PERPETRATED BY OFFENDER
Abused (step)child(ren) in some way	13.2	29/220
Abused (step)child(ren) physically	16.8	17/101
Abused (step)child(ren) emotionally	9.0	17/96
Abused (step)child(ren) sexually	9.2	9/98
CONSEQUENCES OF ABUSIVE BEHAVIOUR
Received treatment for abusive behaviour	20.3	49/241
Charges laid for abuse of family member	40.6	99/244
Incarcerated for abuse of family member	33.0	74/224
Number of consequences/interventions experienced
None	54.5	133/244
One	13.1	32/244
Two	19.3	47/244
Three	13.1	32/244

Instruments

There were two phases to the study. During the first phase, the six sites operated without uniformity in the use of data collection instruments. Although each site prepared individual research reports, the numbers were insufficient to draw firm conclusions. In order to permit comparisons across programs and to capitalize on the volume of data being collected on this unique treatment population, the second phase of the project witnessed the implementation of a standardized national battery of instruments at the beginning of 1994 (Taylor & Robinson, 1995). Selected sites were already using some of the instruments at the time standardization was established. Hence, where possible, the analyses reported here draw on data collected at both phases of the project. Given that the standardization was established relatively late in the progress of the demonstration projects, complete data is not available for all clients. Even with the use of data from both phases, there was a substantial proportion of missing data. Complete data were available for only 20.3% of the sample.

Three instruments were used to assess a series of attitudes and behaviours which were targeted by the demonstration projects as they were believed to be related to violence against women. In addition, a measure of social desirability response bias was included because of previous research which suggested that self-reports of violence against women are likely to be contaminated by response bias (Dutton & Hemphill, 1992, Sanders, 1991). The instruments are described in Taylor and Robinson (1995) and are included in Appendix B. For each instrument, higher scores indicate more of the attribute in question, e.g., higher scores on the Index of Controlling Behaviour indicate more frequent use of controlling tactics, higher scores on the Inventory of Beliefs about Wife Beating indicate greater endorsement of attitudes in support of wife beating. Analysis of the instruments’ internal reliability at pretest was conducted for each of the scales as well as their component subscales. These analyses were carried out for the national sample, as well as for each demonstration site. The scales exhibited acceptable levels of reliability in most cases (See Table 2A, Appendix C).

Balanced Inventory of Desirable Responding (BIDR; Paulhus, 1991). This 40-item inventory (a =.81) assesses the tendency to engage in (i) self-deceptive enhancement (a =.58) and (ii) impression management (a =.85). There are 20 items in each of the two subscales. Although the items must be dichotomized to obtain subscale scores, reliability analyses were conducted on the items in 7-point scale formats. Some of the observations from the Victoria site were not included in the reliability analysis because the data were only available in dichotomous format. However, these data were utilized in the remaining analyses. Correlational techniques were used to assess the degree to which responses on the anger/violence measures were associated with tendencies to present a positive image and/or engage in self-deception.

Inventory of Beliefs about Wife Beating (IBWB; Saunders, Lynch, Grayson, & Linz, 1987). The 31 items in this scale (a=.89) measure beliefs about wife abuse. The items are grouped to obtain 5 subscales: wife beating is justified (a=.83); wives gain from (or are responsible for) the abuse (a =.87); abused women deserve assistance (a=.54); abusers should be punished (a=.73); and, abusers are responsible for their violent behaviour (a=.59). Items in the last 3 subscales are reverse-coded as needed so that they will be consistent with the pattern established in the first two (i.e., high scores indicating attitudes supporting wife beating).

Index of Controlling Behaviour (ICB; based on Meredith & Burns (1990)). This inventory measured the degree to which the respondent utilized controlling tactics in his interactions with his partner. Two versions of the scale were administered: one that asked the offender to use a 7-point scale to estimate the frequency with which coercive tactics were used within a given time period and one that asked whether or not a particular tactic had ever been used by the offender (yes or no). Examination of pre- to post-test differences on the "ever" question revealed that in a substantial proportion of cases (25 to 40%), offenders’ responses at post-test indicated that they had never used a tactic which they had, in fact, admitted to at pretest. The questionable accuracy of the information obtained with the dichotomous "ever" version of the measure rendered it unusable for further analyses. Reliability analyses were conducted on the 7-point scale version but not on the "ever" items.

The ICB consists of 42 items (a =.88), of which all but 5 can be logically grouped into four subscales. The "reasoning" subscale is comprised of 3 items that assess the degree to which the respondent uses reasoning and positive dispute behaviours (a =.68). The 18 item "emotional" subscale assesses emotional and psychological abuse tactics (a =.94). Finally, physical and sexual abuse are assessed using 11 items (a =.89) and 5 items (a =.80) respectively. The remaining five items were included to address the impact of abuse.

Multidimensional Anger Inventory (MAI; Siegal, 1986). This 38 item scale is designed to assess the respondent’s level of anger (a=.92). The inventory is composed of 5 subscales measuring different aspects of anger. The first subscale, labeled "anger arousal," (a =.90) is comprised of 8 items that measure the degree to which the respondent is prone to feelings of anger. The second subscale, "range of anger eliciting situations," is made up of 7 items which present a variety of situations and ask the respondent to rate whether his/her anger would be provoked (a =.86). The "hostile outlook" subscale is composed of 4 items assessing how likely it is that others tend to cause the respondent to get angry (a =.75). The 5 items of the "anger-in" subscale assess the likelihood that the respondent internalizes his anger, e.g., in the form of grudges, difficulty letting go of anger (a =.81). The last scale, "anger-out" consists of two items measuring whether the respondent can let others know that he is angry. This scale had the poorest internal reliability of all the measures used (a =.11).

According to the initial design, all four measures were to be administered at intake and at closure. However, some of the demonstration sites chose to administer the ICB and the BIDR at intake only. Where possible, the differences between offenders’ pre-program and post-program responses on each of the measures was analyzed: pre-test scores were subtracted from post-test scores such that negative difference scores indicate improvement on a particular measure, i.e., less tendency to be angry (MAI), less tendency to use controlling tactics (ICB), less tendency to espouse attitudes supportive of wife beating (IBWB) and less tendency to engage in self-deceptive and impression management behaviours (BIDR). The higher the negative change score, the greater the improvement. Pre-test and post-test scale scores by demonstration site are presented in Tables 3A and 4A of Appendix C.

Closure Information

Early termination. Using the Case Closure Information Form (See Appendix D) developed for this project, therapists provided information about the program completion status of clients. Therapists indicated if the client completed the program or terminated participation early. Therapists were also asked to note the reasons for early termination, if they were known. The clients’ responses on the pre-test/post-test measures collected at intake were analyzed with respect to their usefulness in predicting early termination/withdrawal from the program.

Just over three-fifths of the sample completed the program (61.1% (n=190)), yielding a 38.9% (n=121) treatment attrition rate. The rate of attrition from treatment programs would have been considerably higher if those men who were initially referred to the program, but who never completed any of the intake measures had been included in these calculations. However, because one of the aims of this paper was to ascertain whether any of the four measures predicted early termination, those men who did not complete pre-test measures were omitted from further analyses.

A much higher rate of early termination was recorded for the community-based programs. Specifically, while 87.5% of those in institutions completed the program, only 35.9% of those in community programs did so. Table 5 shows the reasons for early termination of the program.

Table 5: Reasons for Failure to Complete Family Violence Program

Reason for non-completion of program:	Institutional	Community
Total non-complete	12.5% (19/152)	64.2% (102/159)
Reincarcerated	---	25.5% (26/102)
Community supervision terminated	---	2.9% (3/102)
Client involved in family violent behaviour	---	5.9% (6/102)
Placed in institutional segregation	10.5% (2/19)	---
Transferred to new institution	10.5% (2/19)	---
Released from institution	5.3% (1/19)	---
Participating in alternative institutional program	5.3% (1/19)	---
Client uncooperative/disruptive	5.3% (1/19)	4.9% (5/102)
Client doubted program could help	5.3% (1/19)	5.9% (6/102)
Satisfied with gains made	0.0% (0/19)	2.0% (2/102)
Referred elsewhere	10.5% (2/19)	2.0% (2/102)
Other reasons	10.5% (2/19)	2.9% (3/102)
Client did not return- no reason offered	5.3% (1/19)	9.8% (10/102)
Reason unknown- no data	31.5% (6/19)	38.2% (39/102)

Note: ‘Other reasons’ include client unlawfully at large, and attendance no longer enforced by Parole Officer.

As shown in Table 5, there were 12 possible reasons for early termination. Reincarceration was the most frequently cited reason for failure to complete the programs. A method for collapsing this information into larger categories was devised, and reasons were classified into one of two types: (1) those that represented systemic factors beyond the client’s control, e.g., client referred elsewhere, client transferred to a new institution, (4 variables) and (2) those that indicated either the client’s unwillingness to continue in the program or his engagement in inappropriate and/ or unlawful behaviour, e.g., client doubted that the program would help, client reincarcerated (8 variables).

The collapsed data revealed that about 90% of dropouts were accounted for by client (rather than systemic) issues. This was not surprising in light of the fact that the vast majority of those who dropped out were in community programs. Moreover, significant between-group differences emerged, where 57% of those who dropped out of institutional programs did so for reasons beyond their control, compared to less than 3% of the community group.

Therapist outcome measures. The therapist evaluated the degree to which the client showed improvement on a variety of indices over the course of the program. The 11 items that therapists used to assess offenders’ improvement were combined into one scale. Two items were reverse-coded (those dealing with the offender’s lack of motivation and tendency to externalize responsibility for his abusive behaviour). All other items were keyed in a positive direction such that a higher score on this measure was indicative of improvement.

Although 190 men completed the program, complete therapist outcome measures (all 11 items) were only available for 54 participants. For an additional 30 men, partial data were available. In cases where less than 20% (i.e., one or two items) of the data were missing, the individual’s mean score was used to supplement missing item scores. This procedure resulted in obtaining outcome data for an additional 18 cases, rendering a total of 72 participants.

The composite therapist outcome scale showed high internal consistency (a =.89), despite the fact that the analysis was based on a fairly small number of respondents. The total scale had a minimum possible value of 0 and a maximum possible value of 44. Therapist-assigned offenders’ scores ranged from 7 to 40, with a mean of 23.0 (SD=8.2). Attitude and behaviour measures taken at pre-test were correlated to outcome scores in order to ascertain whether they were reliably associated with therapist-rated improvement.

RESULTS

Socially Desirable Responding

Before assessing the degree to which the IBWB, ICB and MAI detected treatment-related change, analyses were conducted to determine whether there was any association between these measures and the social desirability measure. These findings are presented in Table 6. At pre-test, the tendency to engage in impression management was related to lower scores on 11 subscales (significant correlations ranged from -.21 to -.49) while the tendency to engage in self-deceptive enhancement was related to lower scores on 6 subscales (significant correlations ranged from -.28 to -.41).

These results are consistent with those found by Dutton and Hemphill (1992) in an independent sample of family violent offenders. The negative correlations indicate that offenders were inclined to downplay the degree to which they were violent, held inappropriate attitudes toward wife beating, and inappropriately managed anger. It is notable that impression management appeared to produce a larger response bias than self-deception on these measures.

The MAI measures appeared to be most susceptible to social desirability response bias with a range of significant correlations from -.34 to .-49. The Hostile Outlook scale showed the highest correlation with both BIDR measures. This implies that respondents had a tendency to deny having feelings of anger or hostility towards others, despite evidence (MAI scores) showing proclivity to anger. The Anger Out scale, on the other hand, was not significantly correlated with social desirability response bias. This was surprising, considering that Dutton and Hemphill (1992) found an extremely high negative correlation between Anger Out and BIDR scores (r -.68 and r=-.46 for impression management and self deception, respectively) for family violent offenders. However, recall that this (Anger Out) subscale is based on only two items, with poor internal reliability.

The Reason, Emotional and Sexual ICB sub-scales exhibited some degree of social desirability response bias with a range of significant correlations from -.28 to -.38. It is notable, and perhaps surprising, that admission of the use of physical abuse tactics (e.g., kicked, bit, hit, etc.) was unrelated to impression management or self deceptive enhancement. It is possible that offenders who agree to participate in family violence programs realize that their physically violent behaviour has already been identified and that they cannot "fool" the system into believing that they have no problems. On the other hand, they may minimize what is perceived to be more subtle and less verifiable abuse tactics (e.g., emotional and reasoning tactics).

The attitudinal measure, IBWB, exhibited the least susceptibility to social desirable response bias. There was a range of significant correlations from -.21 to .-28, but only for the impression management subscale of the BIDR. This contradicts findings with the ICB, and suggests that social desirability response bias is likely to exert a greater impact on self-report scales which are intended to measure behaviour (e.g., anger, abuse tactics). Accordingly, results of these analyses suggest that offenders may be relatively more truthful about disclosing their attitudes (i.e., condoning family violent behaviour).

While the effects of social desirability on the ICB and the IBWB appear to be contradictory, we suggest that the nature of these two measures may partially explain the apparent discrepancy. More specifically, the ICB requires self-disclosure by respondents, demanding admission or denial of particular behaviours. The IBWB, on the other hand, queries attitudes regarding acceptable behaviour in others.

Although the tendency for offenders to display a positive self-image was less pronounced at post-test, it did not disappear completely. Questions about feelings of anger were still influenced by social desirability: impression management scores were negatively correlated with 4 of the MAI subscales (significant correlations ranged from -.35 to -.55) and self-deceptive enhancement scores were associated with the same 4 subscales of the MAI (correlations ranged from -.33 to -.45). Anger measures appeared particularly susceptible to the influence of social desirability, relative to the measures of abusive behaviour and attitudes toward abuse.

Table 6: Relationship between Social Desirability and Violence/Anger Measures

Violence/Anger Measure	BIDR: PRE-TEST				BIDR: POST-TEST
	Impression Management		Self-Deception		Impression Management		Self-Deception
	r	n	r	n	r	n	r	n
IBWB
TOTAL	-.23*	144	-.14	152	-.20	76	-.22	77
Wife Beating is Justified	-.10	151	-.09	159	-.07	82	-.23*	85
Wives Gain from abuse	-.17	153	-.10	161	-.12	82	-.14	84
Help	-.25*	153	-.09	161	-.12	79	-.10	83
Punish	-.28**	157	-.19	164	-.20	82	-.11	86
Responsibility	-.21*	155	-.12	161	-.16	81	-.14	84
ICB
Reason	-.38**	100	-.18	102	-.10	51	-.04	51
Emotional	-.33**	95	-.28*	98	-.03	46	.04	45
Sexual	-.35**	103	-.30*	106	-.18	47	-.01	47
Physical	-.05	105	-.15	108	-.10	49	.16	49
MAI
Anger Arousal	-.34***	163	-.38***	171	-.35*	83	-.44***	87
Range of situations	-.41***	159	-.40***	167	-.38**	83	-.33*	87
Hostile outlook	-.49***	160	-.41***	170	-.55***	83	-.40***	87
Anger in	-.43***	162	-.41***	171	-.53***	80	-.45***	84
Anger out	.11	164	.08	172	-.06	83	-.02	87

*p<.01, **p<.001, ***p<.0001

The findings do indicate that some caution should be exercised when interpreting treatment-related change on the anger measures for offenders who score high on measures of socially desirable responding. However, the trends exhibited in Table 6 suggest that the impact of the response bias on self-reported violent behaviour using the ICB may diminish over the course of treatment. Such a finding would be consistent with a treatment gain process whereby offenders overcome initial denial of their abusive behaviour during treatment (Daniels & Murphy, 1997).

Interestingly, community-based offenders had higher scores on the impression management subscale than their institutionally-based counterparts at both pre-test (M_Comm = 6.7, SD=4.7, M_Inst = 5.0, SD=3.5; t₍₁₆₅₎ = -2.56; p<.01) and post-test (M_Comm = 7.4, SD=4.5, M_Inst = 5.1, SD=3.9; t₍₈₃₎= -2.40; p<.05).

Pre-program to post-program differences

Changes in offender responses from intake to closure were assessed on the attitudinal and behavioural measures as well as on the BIDR. As noted earlier, these analyses were hampered by the fact that some demonstration sites collected ICB and BIDR data at intake only. Although previous analyses demonstrated significant correlations between impression management (BIDR) and various behavioural/ attitudinal measures, it is important to note that impression management scores were not significantly associated with any of the pre- to post test difference scores. Moreover, pre-test and post-test BIDR scores were highly correlated (r = .79; p<.0001). In combination, these findings may be interpreted to mean that impression management affects pre- and post-test scores on attitudinal and behavioural measures equally. Additionally, these data imply that positive (post-treatment) changes in family violent attitudes and behaviour are not explained by increased impression management.

For the cases in which both pre- and post-treatment data were available, analyses revealed that three subscales of the IBWB and the ‘anger in’ subscale of the MAI showed significant differences. The significant findings are displayed in Table 7. More detailed information on the pre-and post-test scores and comparisons for all of the attitudinal and behavioural measures appears in Table 5A in Appendix C.

Table 7: Significant Differences between Pre-test and Post-test on IBWB & MAI

SUBSCALE	Pretest Score (n)	Posttest Score (n)	Difference Score (n)	t	Effect Size
IBWB (higher score = greater endorsement of wife beating)
TOTAL (min. = 31, max. = 151)	74.9 (148)	68.0 (148)	-6.87 (148)	-4.34***	0.26
Justify (min. = 12, max. = 74)	24.7 (165)	22.2 (165)	-2.48 (165)	-2.94*	0.19
Punish (min. = 5, max. = 35)	17.5 (168)	15.9 (168)	-1.61 (168)	-3.54**	0.25
Responsible (min. = 4, max. = 28)	11.9 (163)	10.6 (163)	-1.29 (163)	-3.20*	0.26
MAI (higher score = more anger)
Anger in (min. = 5, max. = 25)	11.1 (116)	9.9 (116)	-1.19 (116)	-2.95*	0.29

*p<.01, **p<.001, ***p<.0001

These results indicate that following treatment, offenders were less likely to endorse attitudes supportive of wife beating and more likely to report healthy changes in some attitudes and/or behaviours related to the expression of anger (i.e., diminished tendency to hold anger in). Post-treatment changes in IBWB subscales specifically demonstrate a diminished tendency to view wife beating as justified, and increased endorsement of beliefs that abusers are responsible for their violent behaviour, and that abusers should be punished. The two subscales of the IBWB that did not meet stringent criteria for statistical significance did show trends in the same direction. Specifically, there was less endorsement of the belief that wives gain from abuse (p <.03), and greater support for the belief that abused women deserve assistance (p<.05).

With respect to the MAI, reliable differences also emerged in the ‘anger in’ subscale. This finding suggests that, post treatment, men were less likely to internalize their anger. Also approaching statistical significance was the ‘range of anger’ subscale (p<.04), which implies that program participants endorsed fewer situations where anger would be elicited.

Further analyses were conducted to determine which pretest measures and/or background variables predicted post-program improvement on these 7 measures. Institutionally-based men had higher scores than community men on all of the IBWB scales at pre-test, the only exception being the "help" subscale on which there were no significant differences. Moreover, no between-group differences were found on any of the ICB subscales at pretest. Only the ‘anger-in’ subscale of the MAI showed reliable between-group differences at pre-test, though the community group possessed higher scores in this case. See Table 8 for a comparison of scores at pre-test.

Table 8: Mean scores on Attitudinal and Behavioural Measures at Pre-test: Institutional and Community Samples

Violence/Anger Measure	Institutional		Community
IBWB	score	(n)	score	(n)	t	p
TOTAL	77.8	(144)	63.5	(107)	4.35	.0001
Wife Beating is Justified	25.9	(153)	20.0	(113)	3.84	.0002
Wives Gain from Abuse	16.7	(153)	13.0	(114)	3.04	.0026
Abused Wives Deserve Help	12.0	(153)	10.9	(114)	1.78	.0762
Punish	18.1	(155)	15.1	(116)	3.82	.0002
Responsibility	12.2	(152)	10.5	(115)	2.89	.0042
ICB
Reason	5.3	(50)	5.8	(97)	-0.55	.5856
Emotional	7.6	(48)	8.0	(93)	-0.17	.8695
Sexual	1.1	(49)	1.7	(104)	-1.00	.3167
Physical	1.4	(51)	1.2	(102)	0.26	.7948
MAI
Anger Arousal	15.0	(105)	16.7	(176)	-1.97	.0493
Range of Situations	19.5	(91)	19.6	(163)	-1.02	.9190
Hostile Outlook	10.3	(92)	10.1	(165)	0.44	.6629
Anger In	10.8	(93)	11.5	(173)	-1.08	.2792
Anger Out	6.4	(95)	7.3	(176)	-4.10	.0001

For statistical rigor, a was adjusted for all post-test between-group comparisons. For example, between-group comparisons on IBWB subscales (n=5) required differences to be statistically significant at (a /5), which is .01. The MAI contains five subscales, and alpha was adjusted accordingly (.01). Finally, the ICB contains four subscales, and therefore required significance was set at (a/4), which is .013.

Only two between-group differences (i.e., comparing community and institutional samples) were noted post-treatment. It is important to consider, however, that analyses were hampered by the fact that, as mentioned, a large proportion of the community group dropped out of treatment. Regardless, results of comparisons showed that the institutional group maintained a significantly higher mean total score on the IBWB at post-test. Specifically, the mean scores were 70.9 and 57.9 for the institutional and community groups, respectively (p<.01). Similar to group results obtained at pre-test, the community sample had significantly higher scores on the ‘anger out’ subscale of the MAI: mean post-test scores were 6.2 and 7.2 for the institutional and community groups, respectively (p< .01).

The process of adjusting a for statistical rigor rendered all other post-test between-group comparisons not significant. However, it is worth noting that the ‘justify’ and ‘gain’ subscales of the IBWB and the ‘hostile outlook’ subscale of the MAI showed trends for higher scores in the institutional subsample. More specifically, the overall post-test mean for the ‘justify’ subscale was 23.2 for the institutional group, and 18.0 for the community group. This difference approached statistical significance at (p<.02). Similarly, there was a trend for those in institutional programs to be more likely to endorse beliefs that wives gain from abuse: mean scores on the ‘gain’ subscale were 15.2 and 11.6 for the institutional and community groups, respectively (p<.05). Finally, means for the ‘hostile outlook’ subscale of the MAI were 10.5 and 8.9 for institutional and community groups, respectively (p<.03).

Correlates of IBWB difference scores. The substance abuse variable was the only significant predictor of IBWB difference scores. Those who had drug and/or alcohol problems displayed less improvement on the "justify" subscale (M_Subs = 0.5, SD=9.6) than those who reported no such problems (M_NoSubs = -4.7, SD=11.4; t₍₁₅₈₎ = -3.04, p<.01). The same pattern emerged for total scale scores (M_Subs = -2.0, SD=17.6 and M_NoSubs = -10.4, SD=20.1 respectively; t₍₁₄₃₎ = -2.62, p<.01).

Interestingly, examination of the mean scores for each group indicated that men with substance abuse problems started with lower scores on these scales than did non-abusers ("justify": M_Subs = 21.8, M_NoSubs = 24.9; total: M_Subs = 69.1, M_NoSubs = 74.5). These pre-test differences approached statistical significance. However, by post-test, the two groups had similar scores ("justify": M_Subs = 22.2, M_NoSubs = 22.5; total: M_Subs = 68.5, M_NoSubs = 69.0).

Correlates of IBWB post-test scores. The analysis of raw post-test scores yielded a greater number of significant relationships. Pre-test scores on the "reason" scale of the ICB were significantly associated with the post-test "justify" (r₍₇₄₎ = -.28, p<.02) and "gain" subscales (r₍₇₃₎ = -.27, p<.02) as well as total scores on the IBWB (r₍₆₉₎ = -.28, p<.02). These results indicate that greater use of positive dispute-resolving behaviour at the beginning of treatment was associated with less post-program endorsement of the attitude that wife beating is justified or that wives gain from abuse. Men who used reasoning in the attempt to solve interpersonal problems were also less likely in general to endorse attitudes supporting wife abuse.

Pre-test scores on the "anger-out" scale of the MAI were negatively associated with the "punish" (r₍₁₁₆₎= -.22, p<.02) and "responsible" (r₍₁₁₄₎= -.28, p<.003) subscales of the IBWB but positively associated with the "justify" (r₍₁₁₅₎= .21, p<.03) and "gain" (r₍₁₁₄₎= .25, p<.01) subscales of the IBWB. This pattern indicates that those men who were unable to let someone know that they were angry were more likely to endorse, even after treatment, the views that wife beating was justified, that wives benefited from being abused, that abusers were not deserving of punishment and that abusers should not be held responsible for violent behaviour. The relation between inability to express anger appropriately and attitudes condoning woman abuse appears plausible.

Turning to background variables, individuals who had experienced a greater variety of abuse themselves were significantly more likely to report that wives gained from abuse (r₍₁₁₅₎= .22, p<.02) and more likely to endorse attitudes supporting wife beating (r₍₁₀₈₎= .19, p<.05). Men who had a wider variety of contacts with legal or therapeutic interventions as a result of past violent behaviour were more likely to indicate that wife beating was justified (r₍₁₀₇₎= .23, p<.02), that wives gain from abuse (r₍₁₁₄₎= .26, p<.005) and that generally, wife beating was defensible (r₍₁₀₇₎= .29, p<.02). However, more detailed analyses indicated that this relationship was attributable to the number of legal interventions experienced. Men who had been charged and/or incarcerated for perpetrating family violence were more likely to have higher scores on the three scales mentioned above (correlations ranged from .23 to .30; p<.01) than those who had no family violence-related contact with the criminal justice system. There were no significant differences attributable to prior therapy: men who had received treatment for family violent behaviour in the past had similar scores to men who had received no such interventions.

Correlates of MAI difference scores. As was the case with difference scores on the IBWB, only the substance abuse variable served as a significant predictor. Those who had drug and/or alcohol problems displayed a greater post-treatment decrease in proneness to anger (M_Subs = -3.4, SD=6.1) than those who reported no such problems (M_NoSubs = -0.1, SD=6.0; t₍₁₁₈₎ = 2.53, p<.01). Accordingly, substance abusers also showed a larger pre- to-post difference in the "range" subscale of the MAI (M_NoSubs = -0.2, SD=6.1; t₍₁₀₉₎ = 3.36, p<.001).

Further examination indicated that these results were due to the substance abusers starting out with higher scores on these measures at pre-test. Specifically, at pre-test, men with substance abuse problems showed more proneness to anger (M_Subs = 17.8, SD=6.6, M_NoSubs = 15.0, SD=6.6; t₍₂₄₀₎ = -2.93; p<.004), and were likely to get angry in a wider variety of situations than were non-abusers (M_Subs = 22.0, M_NoSubs = 18.5; t₍₂₂₃₎ = -3.8; p<.0002). Perhaps substance abuse and volatility are both symptoms of more generalized coping and stress management problems. At post-test, the group scores on these subscales were not significantly different (Anger Arousal: M_Subs = 14.9, SD=5.8, M_NoSubs = 14.5, SD=5.5; Range of Situations: M_Subs = 18.2, SD=5.1, M_NoSubs = 18.5, SD=5.7). Thus, there is a good possibility that family violence programming is meeting some of the aforementioned coping and stress management needs.

Correlates of MAI post-test scores. Pre-test scores on the "emotional" scale of the ICB were significantly correlated with the "anger arousal" (r₍₇₁₎ = .30, p<.01), "range" (r₍₇₁₎ = .36, p<.002), "hostile outlook" (r₍₇₀₎ = .34, p<.004), and "anger in" subscales (r₍₆₈₎ = .26, p<.03) at post-test. Specifically, greater use of emotionally abusive tactics was associated with higher levels of anger. Sexual abuse perpetrated (as measured by the "sexual" scale of the ICB) was also associated with the offenders’ identification of a wider range of anger-arousing situations (r₍₇₂₎= .37, p<.002), and higher scores on the "hostile outlook" subscale of the MAI (r₍₇₁₎= .29, p<.02).

Two historical variables were predictive of MAI anger scores at post-test. Offenders who had perpetrated a greater variety of abusive acts against partners reported higher anger arousal (r₍₁₁₅₎ =.23, p<.02). Interestingly, the number of abuse acts perpetrated was also significantly correlated with both the "anger in" (r₍₁₁₂₎ = .22, p<.02) and the "anger out" (r₍₁₁₆₎ =.27, p<.004) post-test subscales of the MAI. Those who had substance abuse problems also reported greater tendency to keep anger in (M_Subs = 11.6, SD=4.1) than those without such problems (M_NoSubs = 9.2, SD=3.5; t₍₁₁₅₎ = -3.05, p<.003).

Early termination

Men who participated in community-based programming were more likely to drop-out (64.2%) than those who were in institutions (12.5%; c ² = 87.2, p<.001). This finding is not surprising when one considers that there are fewer mechanisms to ensure the continued participation of men in the community compared to institutions. These data also emphasize the added benefits of offering correctional intervention within institutional settings, where other demands (e.g. employment, family) are less likely to interfered with program participation.

Importantly, two attitudinal/behavioural measures were predictive of early termination: the "wife beating is justified" subscale of the IBWB (t₍₂₂₄₎ = 2.26, p<.03) and the "anger out" subscale of the MAI (t₍₂₄₇₎ = -3.79, p<.0003). Those who terminated their program participation early had lower scores on the "justify" subscale at pre-test (M_Drop-out = 20.6, SD=10.3), indicating that at the outset of treatment, they were less inclined to believe that wife beating was justified than those who finished the program (M_Complete = 24.7, SD=13.6). Those who dropped out of the treatment program were also more likely to express their anger than those who completed treatment according to their pre-test "anger-out" scores (M_Drop-out = 7.4, SD=1.7 and M_Complete = 6.6, SD=1.8 respectively). These results were confirmed with a stepwise regression procedure, where pre-test total scores for the IBWB, the ICB, and the MAI, and all subscales were entered as independent variables to predict early termination. Only three measures met statistical significance (p<.05) for entry into the model. These were: the "justify" subscale of the IBWB, the "anger out" subscale of the MAI, and the "physical" subscale of the ICB. These results may be relevant to the development of strategies to enhance treatment compliance in community-based programs.

Two background variables were also significantly related to early termination: number of abusive experiences in childhood (t₍₂₂₀₎ = 3.07, p<.003) and number of current problems being experienced (t₍₂₈₀₎ = -3.51, p<.0005). Those who followed through with the program until completion had endured a greater variety of abusive experiences in childhood (M_Complete = 2.7, SD=2.2) than those who terminated early (M_Drop-out = 1.8, SD=1.9). It is possible that men who had suffered more abuse may have appreciated that the "cycle of violence" (Walker, 1990) was operating in their lives and were seeking to break it. However, it is also important to consider that, as previously mentioned, men in institutions had suffered significantly more abuse in childhood than those in community-based programs. In turn, those in institutional programming were less capable of voluntarily dropping out of treatment. Finally, men who terminated early experienced a greater number of problems (e.g., financial, educational, legal) at the time of intake (M_Drop-out = 1.5, SD=1.6) than those who completed the program (M_Complete= 1.2, SD=1.1).

Again, these findings were confirmed with a stepwise regression procedure, where background variables were used to predict early termination from treatment. The only two variables that met significance levels (p<.05) for entry into the equation were those previously mentioned: number of current problems, and number (variety) of previous abuse experiences.

Findings reveal that those with less extensive histories of childhood victimization are more likely to remain in family violence programming. However, these data also suggest that offenders with the highest levels of criminogenic need (e.g., multi-problem cases) are at higher risk for negative outcomes and may be less cooperative in completing programs. Perhaps those with multiple problems, who have not been victimized themselves, view other needs as priorities.

Therapist-rated outcome

Although 190 men successfully completed the family violence program(s), complete therapist outcome ratings (i.e., all 11 items) were only available for 54 participants. In many other cases, only partial data were available. To supplement missing data in cases where only one or two items were incomplete, the individual’s mean score (i.e., that of non-missing items) was used to replace missing data. This procedure augmented therapist-outcome data, yielding total scores for 72 men.

Correlational analyses were conducted in order to ascertain the degree of association between the attitudinal/behavioural measures and background information collected at intake and treatment performance, as rated by the therapist. Again, stringent criteria for statistical significance were applied to control for Type 1 error.

The first set of analyses correlated total therapist-rated outcome scores with the IBWB, ICB, and MAI total and subscale scores. For the IBWB total and (five) subscale scores, stringent criteria for statistical significance was set at (a /6) .0083. Results of these analyses showed statistically reliable relationships between therapist-rated outcome and IBWB total scores (r = .44), as well as the "justify" (r = .43) and "gain" (r = .43) subscales. The positive associations between outcome scores and these pretest measures suggest that those men who showed the most negative attitudes pre-treatment were rated as most improved post-treatment. For the MAI total and (five) subscale scores, again alpha was set at .0083 (i.e., a /6). Analyses revealed no statistically significant relationships between MAI pretest scores and therapist-rated outcome measures. ICB total and (four) subscale scores were also correlated with therapist-rated outcome, and criteria for statistical significance was set at (a /5), which is .01. Again, no statistically reliable associations were noted.

All pretest total scores and subscales, for each instrument, were entered into stepwise regression equations (3 separate) to predict therapist-rated outcome. The only scale that met .05 significance for entry into the model was the IBWB total pretest score. This finding suggests that high endorsement of attitudes supporting family violence (pre-treatment) is associated with more positive progress post-treatment. However, in interpreting these results it is important to consider that more negative attitudes at pretest also have a wider margin for showing improvement at post-test. Another issue is that of whether therapist-rated outcome reflects change or threshold scores of treatment performance. This concerns the question of rater (therapist) training for scoring, and the reliability of the scores assigned.

To maintain statistical power in analyses with background variables, only three measures were selected for correlation with therapist-rated outcome: number of current problems, variety of experiences of childhood abuse, and variety of previous interventions for family violence perpetration. Stringent statistical criteria were applied, such that an acceptable level of significance was set at (a /3) .017. Results showed statistically reliable associations between therapist-rated outcome and variety of experiences of abuse in childhood (r = .41). There was no reliable relationship noted between the number of current problems or variety of previous interventions for family violence and success (therapist-rated) in treatment.

The positive correlation between abuse in childhood and success in treatment denotes that men who endured a greater variety of childhood abuse were rated more positively by therapists at the end of treatment. This result was supported in a stepwise regression procedure, where variety of childhood abuse experiences was the only background variable to reliably predict therapist-rated outcome (R² = .16; p<.05). These findings lend support to the "cycle of violence" hypothesis expressed in the previous section. An alternative explanation is that both measures (i.e., childhood abuse and therapist ratings) reflect the offenders’ ability to be introspective and self-disclosing.

Though not statistically significant, a negative relationship was found between number of current problems and therapist-rated outcome (r = -.23). Again, this suggests that multi-need offenders did not fare as well in treatment as their lower-need counterparts. Perhaps offenders with multiple competing needs have more difficulty focusing in treatment, or view intervention for abuse in intimate relationships as a low treatment priority.

Participants who were in institutionally-based programs were more likely to receive positive outcome ratings than community-based men (M_Inst = 28.8, SD=5.3, M_Comm = 20.5, SD=5.9; t₍₇₂₎ = 4.56, p<.0001). However, the former group consisted of only 12 men for whom therapist ratings were available. Moreover, closer examination of the data revealed that this effect was possibly mediated by a third variable: substance abuse. Specifically, of those where therapist-rated outcome data were available none of the 12 men in the institutional sample had substance abuse problems, compared with over one-quarter (28.0%) of those in the community (p<.05). A comparison revealed that, relative to non-substance abusers, those with alcohol and/ or drug problems had poorer ratings on outcome (M_Subs = 19.8, SD=5.2, M_{No Subs} = 22.6, SD=6.8). However, this difference was not statistically reliable.

To verify whether the difference in therapist-rated outcome (community vs. institutional groups) was mediated by substance abuse, an analysis of covariance (ANCOVA) was performed. Briefly, the ANCOVA method enables a statistical comparison between groups, while removing the effect of a third variable - in this case, substance abuse. Means on therapist-rated outcome for the community and institutional groups, after adjusting for the effect of substance abuse, were 19.1 and 24.0, respectively. Despite controlling for the effect of substance abuse, statistical significance for the between-groups difference in means was still achieved.

This result implies that, notwithstanding more substance abuse in the community sample, men in institutional treatment for family violence are rated more positively by therapists on outcome criteria. One potential reason for this is that incarcerated men have a more vested interest in showing treatment-related improvement. Specific implications include earned privileges and early release (e.g., day parole, full parole). Alternatively, without clear training or criteria for scoring, perhaps institutional therapists (n=2) were more optimistic than their community counterparts.

DISCUSSION & CONCLUSIONS

As expected, the men in this sample had a high rate of perpetration of abuse -- about 80% acknowledged having abused a partner and about 13% had abused a child. Yet less than half of these men had experienced any prior interventions, therapeutic or legal, as a result of their abusive behaviour. This finding attests to the reluctance on the part of both the abuser and his victims to admit that domestic violence is a problem that needs to be addressed by either the criminal justice system or by mental health professionals.

Background information collected from program participants indicated that just over two-thirds had experienced and/or witnessed some type of abuse. The percentages for the various types of abuse were comparable, if not higher, than those reported in a recent file review study of randomly sampled federal offenders (see Alksnis & Robinson, 1995). It is perhaps not surprising that men who are involved in a treatment program for abuse in intimate relationships would have a higher likelihood of having been raised in a family where abuse occurred; these results are consistent with the "cycle of violence" hypothesis (Walker, 1990).

The finding of numerous negative relationships between the anger/violence measures and the social desirability (BIDR) highlights the importance of taking this into account when interpreting offenders’ self-reports. These correlations suggest that changes in offenders’ scores on the anger measures (to reflect treatment gain) should be viewed with caution. The tendencies to engage in impression management and self-deceptive enhancement in responding to the anger items were present at pre-test and, to a lesser extent, at post-test. The pre-program correlations may be attributable to offenders’ proclivity to deny and/or minimize the existence and seriousness of their abusive acts. The post-test pattern of results may also be due to denial/minimization or to offenders’ attempts to convince therapists that the program had a positive impact.

At the same time, there was some evidence that the influence of socially desirable responding was diminished following treatment. This suggests that offenders may have become more willing to admit to their offending behaviour and less likely to hide their anti-social attitudes toward women as a result of their involvement in the program. Regardless of what accounts for this phenomenon, the pattern of results suggests that assessors should be especially vigilant about social desirability response bias at the program intake stage.

Interestingly, men in community-based programs were more likely to engage in impression management than men in institutional programs at both pre-test and post-test. However, the reasons for this finding remain unclear. Cumulatively, the ‘social desirability’ results confirm that self-report measures should represent but one criterion for admission into family violence treatment, and but one index of treatment gain.

Two of the three anger/violence inventories, the IBWB and the MAI, showed promise for measuring treatment-related change. IBWB total scores, three of its subscales, and one of the MAI subscales detected changes in abusers’ attitudes and behaviours following program participation. It is not clear what accounts for the failure of the four remaining MAI subscales to detect changes. For the "anger-out" subscale, it could be that few items (n=2), and the low internal consistency of the measure (a = .11) played a role. However, the "arousal", "hostile", and "range" subscales did not suffer from the problem of inadequate reliability. An alternative explanation is that the treatment programs simply did not impact on the dimensions of anger measured by these scales. Specifically, constructs reflected in these subscales are not targeted in these particular treatment programs. Further research is needed to identify which aspects of abusers’ cognitive and behavioural repertoires are influenced by program participation.

The self-report measure of abusive behaviour, the ICB, failed to detect pre- to post-test changes. As mentioned earlier, problems were encountered with the "ever" items when respondents appeared to misinterpret what the questions were asking; a substantial proportion of men failed to indicate at post-test that they had ever used tactics which they had admitted to at pre-test. As for the frequency questions, it seems plausible that there were no differences detected because the admission of using controlling tactics was already low at intake - a potential floor effect. This low endorsement could be due to denial and minimization of abuse and/or to social desirability response bias (at pre-test, 3 of the 4 ICB subscales showed significant correlations with at least one of the BIDR subscales). Furthermore, less than a quarter of the sample had complete data for the frequency questions of the ICB at both pre- and post-test. The small sample size may have also prevented detection of treatment-related differences that existed.

For these reasons, it is not reasonable to conclude that the ICB has no potential utility in this treatment context. There is a possibility that the reliability and validity of the instrument could be enhanced through the development of better administration techniques. For example, perhaps the use of more specific definitions of the various behaviours could be included so that offenders would be better able to identify abusive behaviours in which they have engaged. Alternatively, instructions might be amended to encourage positive responding to items that clearly refer to anti-social behaviours. For example, the instructions might be revised to make offenders feel more comfortable about endorsing the items or increasing the perception that the information is confidential.

With respect to the measures that did reveal treatment-related change, the relationships between pre-test/background information and the post-test results suggested encouraging associations. Correlations between post-test attitudes toward wife beating and pre-test reports of the use of reasoning tactics, expression of anger and greater amount of childhood abuse were in line with expectations regarding how abusers deal with conflict and the intergenerational transmission of abuse. Relationships between anger at post-test and pretest measures also seemed understandable. For instance, one possible explanation for the relationship between the pre-test "emotional" measure and post-test "anger in" subscale may be that the tactics depicted in the ICB match the types of manipulative behaviours that one might expect to see from a person who has difficulty expressing anger in a direct manner.

The relationship between prior abuse perpetrated and the post-test anger (MAI) subscales is also reasonable, given the type of items comprising the MAI. It was not surprising to find that men who have perpetrated more abusive acts were also more prone to anger and less able to resolve feelings of anger. The relationship between substance abuse and post-test MAI "anger in" scores is also comprehensible, given that both substance abuse and the MAI reflect internalizing tendencies. Substance abuse was also significantly related to pre- to post-program difference scores on IBWB and MAI subscales but further investigation revealed that the effect was due to dissimilar pre-test scores by abusers and non-abusers rather than to differential ability to benefit from treatment.

The anger/violence measures were less successful at predicting early program termination and therapist ratings than they were at reflecting change; only two significant relationships emerged. The relationship between the "justify" subscale of the IBWB (which has a low correlation with social desirability) and early termination suggests that men who know that wife beating is not justified are more resistant to treatment and therefore more likely to drop out.

However, subsequent analyses revealed that this finding may also have been an artifact of the sampling procedure. Specifically, those in institutional programs were more likely to see abuse as justified and were more likely to complete treatment. Similarly, lower scores on the "anger out" scale of the MAI were associated with program location (institutional) and program completion.

These data imply that the associations between the "justify" and "anger out" subscales and early program termination are mediated by program location rather than differential proclivity to treatment completion.

The results reported here can only be considered preliminary in that there was a substantial amount of missing data. Moreover, not all sites were equally successful in collecting and forwarding information (see Table 1A, Appendix B). Additionally, validity data for the measures used in the current study are relatively absent. Therapists’ failure to provide ratings of treatment performance precludes inference regarding the relationship between changes on self-report scales and actual changes in behaviour (i.e., reduced intrafamilial abuse).

Increased confidence might be placed in these findings as greater numbers of federal offenders participate in these programs, and more data are aggregated across sites. Nevertheless, the current results are somewhat promising. The continued use of these four measures in evaluation studies of treatment programs for men identified as at risk for family violence is recommended. Two of the three anger/violence measures demonstrated the ability to detect treatment-related change. A substantial proportion of missing post-test data precludes conclusions regarding the third (ICB) measure. Nevertheless, the ICB was useful as a correlate variable despite its failure to pick up pre- to post-program differences.

Use of a social desirability measure is strongly recommended as it may prevent overly optimistic appraisals of the impact of programming. The significant influence of BIDR social desirability measures is supported in research with general samples of offenders (Kroner & Weekes, 1996), and with family violent offenders in particular (Dutton & Hemphill, 1992).

Directions for future research include assessing the durability of attitudinal and/or behavioural changes (although the high rate of attrition from longitudinal studies poses a problem in this regard). Also, it is important to develop methods other than offender self-report to measure whether the treatment has contributed to less abusive behaviour by the participant. An expansion of intermediate measures of treatment gain and outcome is recommended. Fortunately, it is now correctional policy for program providers to employ structured progress evaluations as offenders proceed through treatment (Commissioner’s Directive #730). Multi-method assessment strategies are also advocated, using collateral information found in offender files. Examples include institutional incident reports, private family visit grants and reports, community assessments, and parole reports.

For longitudinal research, post-treatment (and post-release, for incarcerated offenders) indices are paramount for a sound program evaluation. Police records, parole officer reports, and consultation with partners of abusers (Rondeau et al., 1994) are suggested outcome measures. Finally, to enable conclusions regarding the effectiveness of family violence programming, a no-treatment (e.g., waiting list) control group is recommended.

References

Alksnis, C. & Robinson, D. (1995). Childhood victimization and violent behaviour among adult offenders. Ottawa: Correctional Service Canada.

Alksnis, C. & Taylor, J. (1995). The impact of experiencing and witnessing family violence during childhood: Outcomes on child and adult behaviour. Ottawa: Correctional Service Canada.

Correctional Service Canada (1994). Conceptual Model: Family Violence Programming within a Correctional Setting. Ottawa: Correctional Service Canada.

Daniels, J., W., & Murphy, C., M. (1997). Stages and process of change in batterers’ treatment. Cognitive and Behavioral Practice, 4, 123-145.

Dutton, D. G., & Hart, S.D. (1992). Evidence for long-term, specific effects of childhood abuse and neglect on criminal behavior in men. International Journal of Offender Therapy and Comparative Criminology, 36, 129-137.

Dutton, D. G., & Hemphill, K. J. (1992). Patterns of socially desirable responding among perpetrators and victims of wife assault. Violence and Victims, 7 (1), 29-39.

Kroner, D. G., & Weekes, J., R. (1996). Balanced inventory of desirable responding: factor structure, reliability, and validity with an offender sample. Personality and Individual Differences, 21 (3), 323-333.

Malinosky-Rummell, R., & Hansen, D. J. (1993). Long-term consequences of childhood physical abuse. Psychological Bulletin, 114, 68-79.

McCord, J. (1983). A forty year perspective on effects of child abuse and neglect. Child Abuse & Neglect, 7, 265-270.

Meredith, C, & Burns, N. (1990). Evaluation of Batterer’s Treatment Programs. Project report submitted to Ontario Ministry of Correctional Services. Ottawa: Abt Associates of Canada.

Oliver, J. E. (1993). Intergenerational transmission of child abuse: Rates, research, and clinical implications. American Journal of Psychiatry, 150, 1315-1324.

Paulhus, D. L. (1991). Measurement and control of response bias. In J. P. Robinson, P. R. Shaver & L. S. Wrightsmen (Eds.). Measures of Personality and Social Psychological Attitudes. San Diego: Academic Press.

Preston, D. L., & Murphy, S. (1997). Motivating treatment-resistant clients in therapy. Forum on Corrections Research, 9 (2), 39-43.

Robinson, D. & Taylor, J. (1995). The Incidence of Family Violence Perpetrated by Federal Offenders: A File Review Study. Ottawa: Correctional Service Canada.

Rondeau, G., Brochu, S., Lemire, G., Brodeur, N., & Schneeberger, P. (1994). Review of the Literature on Treatment Programs for Men who are Violent within their Familial Relations. Montreal: Centre international de criminologie comparee.

Saunders, D. G., Lynch, A. B., Grayson, M., & Linz, D. (1987). The Inventory of Beliefs about Wife Beating: The construction and validation of a measure of beliefs and attitudes. Violence and Victims, 2, 39-55.

Siegel, J. M. (1986). The Multidimenstional Anger Inventory. Journal of Personality and Social Psychology, 51, 191-200.

Vanderberg, S., & Knoll, A. (1996). Review of Family Violence Program Development: Treatment Demonstration Projects. Ottawa: Correctional Service Canada.

Walker, G. (1990). Family Violence and the Women’s Movement. Toronto: University of Toronto Press.

Widom, C. S. (1989a). Child abuse, neglect, and violent criminal behavior. Criminology, 27, 251-271.

Widom, C. S. (1989b). Does violence beget violence? A critical examination of the literature. Psychological Bulletin, 106, 3-28.

Widom, C. S. (1991). Childhood victimization: Risk factor for delinquency. In M. E. Colten & S. Gore (Eds.) Adolescent Stress: Causes and Consequences (pp.201-221). New York: Aldine de Gruyter.

Appendix A CLIENT INFORMATION FORM

(Adapted from the John Howard Society Abuse History Interview)

RESEARCH # ____________ DATE (D,M,Y) ____________ FPS # ___________

Program Location: 1 Institution 2 Community

Previous enrollment in this program? 1 Yes 2 No

If yes, specify date client started in program: (D,M,Y) _____________________

BASIC INFORMATION

1. Client Name:

Last ___________
First ___________
Middle __________

2. Birthdate:

Day ____ Month _______ Year _____

3. Language:

1. English
2. French
3. Other

4. Aboriginal Origins:

1 Yes 2 No

5. Current Marital Status (circle all applicable)

1. Single (never married)
2. Common-law
3. Married
4. Divorced
5. Separated
6. Widowed
7. Other ______________

6. Previous Marital Status (circle all applicable)

1. Ever Common-law
2. Ever Married
3. Ever Divorced
4. Ever Separated
5. Other __________

7. Do you have any children?

If Yes,_____ how many_____ number of children living with you at your home (or usual residence if incarcerated)

8. Length of time with current partner:________________

EDUCATION/EMPLOYMENT

1. Highest grade completed: _________

Any post-secondary:
1 Yes 2 No

2. Current employment status:

1. full time
2. part-time
3. unemployed
4. disable
5. retired
6. incarcerated

3. Occupation (when working):______________________

FINANCIAL STATUS

(Non-Applicable for Institutional Programs)

On social assistance?

1 Yes 2 No
Personal annual income: $_______

HISTORY OF ASSAULT

1. Have you ever been charged for assaulting your partner, past partner, or children?

1 Yes 2 No

2. Have you ever spent time in prison for assault against your partner, past partner, or children?

1 Yes 2 No

3. Have you ever received treatment for violent behaviour toward family members before?

1 Yes 2 No

SOCIAL HISTORY

The next section of the interview has to do with what things were like when you were growing up, so I’m going to ask you some questions about your family and your childhood.

FAMILY OF ORIGIN

1. Did your parents ever separate or divorce?

1 Yes 2 No

2. Were you or any of your brothers/sisters ever removed from your parents’ care?

1 Yes 2 No

3a. Did your father or step-father ever physically abuse his wife or your mother (that you know of/witnessed)?

1 Yes 2 No

3b What kinds of things did you see?

1. hitting
2. pushing
3. kicking
4. slapping
5. grabbing
6. choking
7. use of weapon
8. other ____________

4a Did your father or step-father ever emotionally abuse his wife or your mother (that you know of/witnessed)?

1 Yes 2 No

4b What kinds of things did you see?

1. yelling
2. swearing
3. threatening
4. insulting
5. blaming/inducing guilt
6. jealousy
7. isolation
8. property damage
9. other ________________

5a Did your father or step-father ever sexually abuse his wife or your mother (that you know of/witnessed)?

1 Yes 2 No

5b What kinds of things did you see?

1. pressuring sex
2. forcing sex
3. sexual touching/fondling
4. derogatory sexual comments
5. other __________________

6a Were you ever physically abused as a child?

1 Yes 2 No

By whom:

1. Mother
2. Father
3. Brother
4. Sister
5. Other Relative
6. Neighbour/Family Friend
7. Stranger
8. Institutional Staff

6b What kinds of behaviours were involved?

1. hitting
2. pushing
3. kicking
4. slapping
5. grabbing
6. choking
7. use of weapon
8. other ____________

7a Were you ever emotionally abused as a child?

1 Yes 2 No

By whom:

1. Mother
2. Father
3. Brother
4. Sister
5. Other Relative
6. Neighbour/Family Friend
7. Stranger
8. Institutional Staff

7b What kinds of behaviors were involved?

1. yelling
2. swearing
3. threatening
4. insulting
5. blaming/inducing guilt
6. jealousy
7. isolation
8. other ________________

8a Were you ever sexually abused as a child?

By whom:

1. Mother
2. Father
3. Brother
4. Sister
5. Other Relative
6. Neighbour/Family Friend
7. Stranger

8b What kinds of behaviors were involved?

1. pressuring sex
2. forcing sex
3. sexual touching/fondling
4. derogatory sexual comments
5. other ___________________

9. Did you ever witness the abuse of a (step) brother or (step) sister?

1 Yes 2 No

Type:

1. Physical
2. Emotional
3. Sexual

CURRENT PROBLEMS

Now I’d like to go over some other problems areas, to see where your current problems are, so let me know if these things are a problem or concern for you, right now. (Note: Questions 1,2, and 3 are skipped for offenders enrolled in an institutional programs)

1. Do you have any financial problems right now?

1 Yes 2 No

2. Are you having employment problems (frequently unemployed or changing jobs)?

1 Yes 2 No

3. Do you have any housing problems (e.g., do you move a lot, ever evicted

1 Yes 2 No

4. Are you having any difficulties with education (achievement problems, unable to access resources)?

1 Yes 2 No
Specify _________________

5. Are your children currently placed in care?

1 Yes 2 No

6. Is alcohol use currently a problem for you?

1 Yes 2 No

7. Is your drug use causing any problems?

1 Yes 2 No
Specify _________________

8. Do you currently have suicidal thoughts?

1 Yes 2 No
Specify _________________

9. Are you being treated for a psychiatric disorder?

1 Yes 2 No
Specify _________________

10. Are you currently involved in a legal conflict (other than divorce)?

1 Yes 2 No
Specify _________________

11. Are you currently being physically abused?

By whom:

1. Mother
2. Father
3. Brother
4. Sister
5. Other Relative
6. Neighbour/Family Friend
7. Stranger
8. Spouse/Partner
9. Offenders or Correctional Staff

12. Are you currently being emotionally abused?

1 Yes 2 No

By whom:

1. Mother
2. Father
3. Brother
4. Sister
5. Other Relative
6. Neighbour/Family Friend
7. Stranger
8. Spouse/Partner
9. Offenders or Correctional Staff

HISTORY OF PERPETRATION OF FAMILY VIOLENCE

I’m going to ask you some questions about the abuse that you may have perpetrated in your relationship with ____________ (partner’s name, if known; focus on the current or most recent relationship).

1. What types of abuse were committed (define each type with examples):

1. Yes, Abuse (if Yes, check type below)
2. No Abuse

1. physical (hit, slap, punch, kick, push, use of weapon)
2. emotional (name calling, insults, guilt, degradation, intimidation)
3. sexual (forced sexual activity of any type)
4. property damage (hurt pets, destroy furniture, belongings)
5. isolation (restrict use of car/phone/money/visits to and from family and friends)
6. financial

2. What types of abuse were committed with previous partners (if any):

1. Yes, Abuse (if Yes, check type below)
2. No Abuse

1. physical (hit, slap, punch, kick, push, use of weapon)
2. emotional (name calling, insults, guilt, degradation, intimidation)
3. sexual (forced sexual activity of any type)
4. property damage (hurt pets, destroy furniture, belongings)
5. isolation (restrict use of car/phone /money/visits to and from family and friends)
6. financial

Did any of this abuse continue after you became separated from this partner?

1 Yes 2 No

3. What types of abuse were committed with your children or step-children:

1. Yes, Abuse ___ (if Yes, check type below)
2. No Abuse

1. physical (hit, slap, punch, kick, push, use of weapon)
2. emotional (name calling, insults, guilt, degradation, intimidation)
3. sexual (forced sexual activity of any type)
4. property damage (hurt pets, destroy furniture, belongings)
5. isolation (restrict use of car/phone/money/visits to and from family and friends)
6. financial

Alcohol Problem Rating:

Based on all available sources of information, rate the extent to which this client exhibited alcohol problems at treatment intake:

1. No alcohol problems
2. Alcohol causes some interference in functioning
3. Alcohol causes serious interference in functioning

Appendix B

i) Index of Controlling Behaviour

Below is a list of things you may have done in your relationships. First, circle the number showing how many times you did these things in the last six months. Then circle "YES" or "NO" to show if you have ever done these things.

Use this scale as a guide to answer the following questions.

Never	Once	Twice	3-5 Times	6-10 Times	11-20 Times	More than 20 Times	Not Applicable
0	1	2	3	4	5	6	N/A

 

		Have you done this in the last six months?								Have you ever done it?
1.	Discussed an issue calmly that your partner raised?	0	1	2	3	4	5	6	N/A	Yes	No
2.	Got information to back up your side of things?	0	1	2	3	4	5	6	N/A	Yes	No
3.	Brought in, or tried to bring in, someone to help settle things?	0	1	2	3	4	5	6	N/A	Yes	No
4.	Insulted or swore at her?	0	1	2	3	4	5	6	N/A	Yes	No
5.	Criticized her clothes or her physical appearance?	0	1	2	3	4	5	6	N/A	Yes	No
6.	Criticized her care of the children?	0	1	2	3	4	5	6	N/A	Yes	No
7.	Yelled and screamed at her?	0	1	2	3	4	5	6	N/A	Yes	No
8.	Sulked or refused to talk about an issue?	0	1	2	3	4	5	6	N/A	Yes	No
9.	Stomped out of a room or house or yard?	0	1	2	3	4	5	6	N/A	Yes	No
10.	Demanded a strict account of how your partner spends money?	0	1	2	3	4	5	6	N/A	Yes	No
11.	Made a major financial decision without consulting her?	0	1	2	3	4	5	6	N/A	Yes	No
12.	Accused her of having an affair?	0	1	2	3	4	5	6	N/A	Yes	No
13.	Discouraged her contact with friends or family members?	0	1	2	3	4	5	6	N/A	Yes	No
14.	Did not allow her to go out of the house when she wanted to go?	0	1	2	3	4	5	6	N/A	Yes	No
15.	Restricted her use of the car or phone?	0	1	2	3	4	5	6	N/A	Yes	No
16.	Embarrassed her in front of others?	0	1	2	3	4	5	6	N/A	Yes	No
17.	Drove the car recklessly to frighten her?	0	1	2	3	4	5	6	N/A	Yes	No
18.	Interrupted her sleeping to bother her?	0	1	2	3	4	5	6	N/A	Yes	No
19.	Threatened to take the children away from her?	0	1	2	3	4	5	6	N/A	Yes	No
20.	Blamed her for your problems?	0	1	2	3	4	5	6	N/A	Yes	No
21.	Threatened to hurt yourself, or her, if she left you?	0	1	2	3	4	5	6	N/A	Yes	No
22.	Withheld affection from her?	0	1	2	3	4	5	6	N/A	Yes	No
23.	Withheld sex from her?	0	1	2	3	4	5	6	N/A	Yes	No
24.	Were insensitive to her sexual needs and desires?	0	1	2	3	4	5	6	N/A	Yes	No
25.	Pressured her for sex against her will?	0	1	2	3	4	5	6	N/A	Yes	No
26.	Hurt her sexually, or made her have intercourse against her will?	0	1	2	3	4	5	6	N/A	Yes	No
27.	Intentionally harmed a pet to upset her?	0	1	2	3	4	5	6	N/A	Yes	No
28.	Threw something in your partner’s presence?	0	1	2	3	4	5	6	N/A	Yes	No
29.	Threw something at your partner?	0	1	2	3	4	5	6	N/A	Yes	No
30.	Pushed, grabbed, or shoved her?	0	1	2	3	4	5	6	N/A	Yes	No
31.	Slapped her?	0	1	2	3	4	5	6	N/A	Yes	No
32.	Kicked, bit, or hit her with a fist?	0	1	2	3	4	5	6	N/A	Yes	No
33.	Hit or tried to hit her with something?	0	1	2	3	4	5	6	N/A	Yes	No
34.	Beat her up?	0	1	2	3	4	5	6	N/A	Yes	No
35.	Choked her?	0	1	2	3	4	5	6	N/A	Yes	No
36.	Threatened to harm her with a knife or gun?	0	1	2	3	4	5	6	N/A	Yes	No
37.	Used a knife or fired a gun in her presence to upset her?	0	1	2	3	4	5	6	N/A	Yes	No
38.	Threatened to harm her with a knife or gun?	0	1	2	3	4	5	6	N/A	Yes	No
39.	Your partner was unable to perform regular activities due to injuries caused by you?	0	1	2	3	4	5	6	N/A	Yes	No
40.	The police were called because of a conflict with your partner?	0	1	2	3	4	5	6	N/A	Yes	No
41.	Your partner had to leave home for her own safety?	0	1	2	3	4	5	6	N/A	Yes	No
42.	Your partner has gotten support from a woman’s group or shelter?	0	1	2	3	4	5	6	N/A	Yes	No

ii) Multidimensional Anger Inventory

INSTRUCTIONS: Everybody gets angry from time to time. A number of statements that people have used to describe the times that they get angry are included below. Read each statement and circle the number to the right of the statement that best describes you. There are no right or wrong answers.

Completely Undescriptive	Mostly Undescriptive	Partly Descriptive	Mostly Descriptive	Completely Descriptive
0	1	2	3	4

 

1.	I tend to get angry more frequently than most people.	0	1	2	3	4
2.	Other people seem to get angrier than I do in similar circumstances.	0	1	2	3	4
3.	I harbor grudges that I don’t tell anyone about.	0	1	2	3	4
4.	I try to get even when I’m angry with someone.	0	1	2	3	4
5.	.I am secretly quite critical of others.	0	1	2	3	4
6.	It is easy to make me angry.	0	1	2	3	4
7.	When I am angry with someone, I let that person know.	0	1	2	3	4
8.	I have met many people who are supposed to be experts who are no better than I am.	0	1	2	3	4
9.	Something makes me angry almost every day.	0	1	2	3	4
10.	I often feel angrier than I think I should.	0	1	2	3	4
11.	I feel guilty about expressing my anger.	0	1	2	3	4
12.	When I am angry with someone, I take it out on whoever is around.	0	1	2	3	4
13.	Some of my friends have habits that annoy and bother me very much.	0	1	2	3	4
14.	I am surprised at how often I feel angry.	0	1	2	3	4
15.	Once I let people know I’m angry, I can put it out of my mind.	0	1	2	3	4
16.	People talk about me behind my back.	0	1	2	3	4
17.	At times, I feel angry for no specific reason.	0	1	2	3	4
18.	I can make myself angry about something in the past just by thinking about it.	0	1	2	3	4
19.	Even after I have expressed my anger, I have trouble forgetting about it.	0	1	2	3	4
20.	When I hide my anger from others, I think about it for a long time.	0	1	2	3	4
21.	People can bother me just by being around.	0	1	2	3	4
22.	When I get angry, I stay angry for hours.	0	1	2	3	4
23.	When I hide my anger from others, I forget about it pretty quickly.	0	1	2	3	4
24.	I try to talk over problems with people without letting them know I am angry.	0	1	2	3	4
25.	When I get angry, I calm down faster than most people.	0	1	2	3	4
26.	I get so angry, I feel like I might lose control.	0	1	2	3	4
27.	If I let people see the way I feel, I’d be considered a hard person to get along with.	0	1	2	3	4
28.	I am on my guard with people who are friendlier than I expected.	0	1	2	3	4
29.	It’s difficult for me to let people know I’m angry.	0	1	2	3	4
30.	I get angry when: someone lets me down people are unfair something blocks my plans I am delayed someone embarrasses me I have to take orders from someone less capable than I I have to work with incompetent people I do something stupid I am not given credit for something I have done.	0	1	2	3	4

iii) Inventory of Beliefs about Wife Beating

For each statement, please circle the number that best matches how much you would agree or disagree with the statement.

Strongly Agree	Agree	Slightly Agree	Neither Agree Nor Disagree	Slightly Disagree	Disagree	Strongly Disagree
1	2	3	4	5	6	7

 

1.	A husband has no right to hit his wife even if she breaks agreements she has made with him.	1	2	3	4	5	6	7
2.	Even when a wife’s behaviour challenges her husband’s manhood, he’s not justified in hitting her.	1	2	3	4	5	6	7
3.	A wife doesn’t deserve to be hit even if she keeps reminding her husband of his weak points.	1	2	3	4	5	6	7
4.	Even when women lie to their husbands they do not deserve to be hit.	1	2	3	4	5	6	7
5.	A sexually unfaithful wife deserves to be hit.	1	2	3	4	5	6	7
6.	Sometimes it is OK for a man to hit a wife.	1	2	3	4	5	6	7
7.	It would do some wives some good to be hit by their husbands.	1	2	3	4	5	6	7
8.	Occasional violence by a husband toward his wife can help maintain the marriage.	1	2	3	4	5	6	7
9.	There is no excuse for a man hitting his wife.	1	2	3	4	5	6	7
10.	Wives who are hit are responsible for what happened because they intended it to happen.	1	2	3	4	5	6	7
11.	A woman who constantly refuses to have sex with her husband is asking to be hit.	1	2	3	4	5	6	7
12.	Wives who are hit are responsible for what happened because they should have foreseen it would happen.	1	2	3	4	5	6	7
13.	Battered wives try to get their partners to hit them as a way of getting attention.	1	2	3	4	5	6	7
14.	When a wife is hit, it is caused by her behaviour in the weeks before it happened.	1	2	3	4	5	6	7
15.	Most wives secretly desire to be hit by their husbands.	1	2	3	4	5	6	7
16.	Wives try to get hit by their husbands to get sympathy from others.	1	2	3	4	5	6	7
17.	Episodes of a man hitting his wife are the wife’s fault.	1	2	3	4	5	6	7
18.	Wives could avoid being hit by their husbands if they knew when to stop talking.	1	2	3	4	5	6	7
19.	If I heard a woman being hit by her husband, it would be best that I do nothing.	1	2	3	4	5	6	7
20.	If I heard a woman being hit by her husband, I would call the police.	1	2	3	4	5	6	7
21.	Women feel pain and no pleasure when hit by their husbands.	1	2	3	4	5	6	7
22.	Cases of wife assault are the fault of the husband.	1	2	3	4	5	6	7
23.	If a wife is assaulted by her husband, she should divorce him immediately.	1	2	3	4	5	6	7
24.	The best way to deal with wife assault is to arrest the husband.	1	2	3	4	5	6	7
25.	A man who has assaulted his wife should go to jail.	1	2	3	4	5	6	7
26.	A wife should move out of the house if her husband assaults her.	1	2	3	4	5	6	7
27.	Husbands who hit their wives are responsible for what happened because they intended to do it.	1	2	3	4	5	6	7
28.	Wife assault should be given a high priority as a social problem by government agencies.	1	2	3	4	5	6	7
29.	Social agencies should do more to help assaulted women.	1	2	3	4	5	6	7
30.	Women should be protected by law if their husbands hit them.	1	2	3	4	5	6	7
31.	Husbands who assault their wives should be responsible for the abuse because they should have foreseen that it would happen.	1	2	3	4	5	6	7

Appendix C

Table 1A Missing Pre-test Data for Each Demonstration Site

	ALL	B.C.	TORONTO	EDMONTON	OTTAWA	NFLD	QUEBEC
Original Sample Size	555	180	63	71	133	82	26
Sample size after removing those with no pretest data on any of the 4 measures of interest8	336	132	63	36	47	32	26
BIDR
Data for one or more subscales	186 (55.4%)	68 (51.5%)	56 (88.9%)	24 (66.7%)	13 (27.7%)	19 (59.4%)	6 (23.1%)
No pretest data on BIDR	150 (44.6%)	64 (48.5%)	7 (11.1%)	12 (33.3%)	34 (72.3%)	13 (40.6%)	20 (76.9%)
IBWB
Data for one or more subscales	275 (81.8%)	131 (99.2%)	61 (96.8%)	21 (58.3%)	15 (31.9%)	21 (65.5%)	26 (100%)
No pretest data on IBWB	61 (18.2%)	1 (0.8%)	2 (3.2%)	15 (41.7%)	32 (68.1%)	11 (34.4%)
ICB (7-point scale9)
Data for one or more subscales	164 (48.8%)	28 (21.2%)	63 (100%)	18 (50%)	12 (25.5%)	17 (53.1%)	26 (100%)
No pretest data on ICB	172 (51.2%)	104 (78.8%)		18 (50%)	35 (74.5%)	15 (46.9%)
MAI
Data for one or more subscales	273 (81.3%)	70 (53.0%)	62 (98.4%)	36 (100%)	47 (100%)	32 (100%)	26 (100%)
No pretest data on MAI	63 (18.8%)	62 (47.0%)	1 (1.6%)
Complete data on all 4 measures	65 (20.3%)	12 (9.5%)	39 (62.9%)	8 (22.2%)	2 (4.4%)	2 (8.3%)	2 (7.7%)

Table 2A Raw Alphas Calculated at Pre-test

	ALL (n=336)	B.C. (n=132)	TORONTO (n=63)	EDMONTON (n=36)	OTTAWA (n=71)	NFLD (n=32)	QUEBEC (n=26)
BIDR TOTAL	.81 (91)		.79 (47)	.76 (20)	.78 (10)	.83 (14)	.86 (3)
SDE	.58 (103)		.54 (49)	.53 (23)	.63 (12)	.37 (19)	.63 (6)
IM	.85 (99)		.84 (54)	.84 (20)	.87 (11)	.87 (14)	.73 (3)
IBWB TOTAL	.89 (251)	.87 (120)	.87 (52)	.85 (21)	.83 (13)	.86 (21)	.77 (24)
Justify	.83 (266)	.76 (127)	.75 (58)	.80 (21)	.78 (13)	.79 (21)	.80 (26)
Gain	.87 (267)	.83 (128)	.74 (58)	.77 (21)	.83 (14)	.78 (21)	.85 (26)
Help	.54 (267)	.62 (127)	.75 (57)	.44 (21)	.38 (15)	.60 (21)	-.005 (26)
Punish	.73 (271)	.74 (130)	.79 (59)	.55 (21)	.79 (15)	.63 (21)	.53 (25)
Responsible	.59 (267)	.61 (127)	.63 (58)	.69 (21)	.56 (15)	.75 (21)	.02 (25)
ICB -- 7 pt. scale TOTAL	.88 (107)	.86 (13)	.84 (57)	.86 (4)	.84 (4)	.95 (4)	.88 (25)
Reason	.68 (151)	.44 (24)	.73 (57)	.59 (16)	.50 (9)	.62 (19)	.73 (26)
Emotional	.94 (144)	.92 (22)	.78 (60)	.77 (17)	.89 (6)	.98 (13)	.94 (26)
Sexual	.80 (156)	.69 (23)	.56 (63)	.77 (16)	.57 (10)	.87 (18)	-- (no var)
Physical	.89 (156)	.87 (25)	.77 (61)	.70 (16)	.67 (11)	.93 (17)	.94 (26)
MAI TOTAL	.92 (218)	.90 (59)	.92 (52)	.87 (25)	.93 (39)	.92 (20)	.87 (23)
Arousal	.90 (260)	.87 (69)	.92 (62)	.84 (25)	.90 (47)	.88 (31)	.90 (26)
Range	.86 (243)	.85 (65)	.87 (61)	.79 (25)	.90 (45)	.78 (21)	.82 (26)
Hostile	.75 (246)	.66 (66)	.69 (62)	.53 (25)	.82 (46)	.63 (21)	.85 (26)
Anger in	.81 (257)	.76 (69)	.84 (60)	.67 (25)	.89 (45)	.76 (32)	.73 (24)
Anger out	.11 (260)	-.17 (69)	-.46 (62)	.49 (25)	.52 (47)	.09 (31)	.46 (26)

Table 3A Pre-test Scores for each Demonstration Site

	ALL	B.C. (1)	TORONTO (2)	EDMONTON (3)	OTTAWA (4)	NFLD (5)	QUEBEC (6)
BIDR
Self-Deceptive Enhance (20 items)/20	6.8 (173)	6.2 (63)	9.6 (49)	3.3 (24)	5.8 (12)	6.7 (19)	5.2 (6)
Impression Management (20 items)/20	6.0 (165)	4.8 (62)	8.4 (54)	3.4 (21)	4.5 (11)	6.8 (14)	7.0 (3)
IBWB (higher = endorses wife beating)
TOTAL (31 items) /217	71.7 (251)	69.6 (120)	65.6 (52)	68.3 (21)	57.3 (13)	57.1 (21)	118.9 (24)
Justify (12 items) /84	23.4 (266)	21.3 (127)	21.1 (58)	21.4 (21)	16.6 (13)	17.5 (21)	48.8 (26)
Gain (7 items) /49	15.1 (267)	13.3 (128)	13.3 (58)	15.1 (21)	11.4 (14)	11.4 (21)	34.4 (25)
Help (5 items) /35	11.5 (267)	11.6 (127)	11.1 (57)	11.0 (21)	11.4 (15)	10.0 (21)	13.9 (26)
Punish (5 items) /35	16.8 (271)	18.4 (130)	15.2 (59)	15.8 (21)	14.7 (15)	14.7 (21)	16.3 (25)
Responsible (4 items) /28	11.4 (267)	12.2 (127)	11.0 (58)	10.8 (21)	9.9 (15)	9.2 (21)	12.0 (25)
ICB (higher = offender used controlling tactics more often) TOTAL
Reason (7-pt. scale) -- 3 item/18	5.7 (151)	7.0 (24)	4.9 (57)	7.2 (16)	7.4 (9)	6.7 (15)	3.8 (26)
Emotional (7-pt.) -- 18 items/108	8.6 (144)	11.1 (22)	4.1 (60)	14.6 (17)	13.8 (6)	17.3 (10)	4.7 (26)
Sexual (7-pt.) -- 5 items/30	1.8 (156)	2.3 (23)	0.49 (63)	3.8 (16)	3.7 (10)	3.0 (15)	0 (26)
Physical (7-pt.) -- 11 items/66	1.4 (156)	2.0 (25)	0.62 (61)	2.6 (16)	0.73 (11)	2.9 (14)	0.88 (26)
Reason (ever done? Y/N) -- 3 item/3	2.3 (273)	2.4 (135)	2.0 (58)	1.9 (19)	2.4 (14)	1.9 (21)	2.6 (26)
Emotional (ever done?) -- 18 items /18	6.4 (244)	6.5 (112)	4.6 (57)	7.2 (18)	4.6 (12)	7.2 (19)	9.4 (26)
Sexual (ever done?) -- 5 items/5	1.3 (272)	1.4 (135)	1.0 (59)	1.3 (18)	1.4 (14)	1.2 (20)	1.2 (26)
Physical (ever done?) -- 11 items/11	2.4 (276)	2.3 (134)	2.0 (62)	3.3 (19)	1.6 (14)	2.8 (21)	3.7 (26)
MAI (higher = more angry)
Anger Arousal --8 items /40	16.1 (271)	14.4 (69)	13.8 (62)	18.5 (36)	18.3 (47)	18.0 (31)	16.8 (26)
Range of situations -- 7 items/35	19.6 (254)	18.9 (65)	17.1 (61)	21.5 (36)	21.3 (45)	20.0 (21)	21.0 (26)
Hostile outlook -- 4 items /20	10.2 (257)	9.9 (66)	8.4 (62)	11.3 (36)	11.5 (46)	10.3 (21)	11.3 (26)
Anger in -- 5 items /25	11.3 (266)	11.0 (69)	8.8 (60)	13.6 (36)	12.8 (45)	12.3 (32)	10.4 (26)
Anger out -- 2 items /10	7.0 (271)	5.9 (69)	7.3 (62)	7.2 (36)	7.1 (47)	7.8 (31)	7.8 (26)

Table 4A Post-test Scores for each Demonstration Site

	ALL	B.C. (1)	TOR. (2)	EDM. (3)	OTT. (4)	NFLD (5)	QUEB (6)
BIDR
Self-Deceptive Enhance (20 items) /20	6.7 (90)	6.2 (56)	7.7 (22)	6.0 (1)	0 (0)	7.5 (8)	0 (0)
Impression Management (20 items) /20	5.9 (86)	5.1 (53)	8.0 (23)	4.0 (1)	0 (0)	5.7 (6)	0 (0)
IBWB (higher = endorses wife beating)
TOTAL (31 items) /217	68.0 (164)	64.2 (111)	57.9 (23)	59.6 (7)	49.0 (1)	57.6 (5)	114.8 (17)
Justify (12 items) /84	22.0 (174)	19.7 (117)	18.7 (25)	16.7 (7)	12.0 (1)	17.4 (7)	47.4 (17)
Gain (7 items) /49	14.4 (174)	12.0 (118)	10.8 (25)	14.9 (7)	7.0 (1)	12.2 (6)	37.5 (17)
Help (5 items) /35	10.7 (172)	10.9 (115)	10.1 (25)	9.7 (7)	8.0 (1)	8.6 (7)	11.9 (17)
Punish (5 items) /35	16.0 (176)	16.7 (117)	14.8 (26)	13.9 (7)	19.0 (1)	13.8 (8)	14.0 (17)
Responsible (4 items) /28	10.7 (173)	11.1 (114)	10.7 (26)	9.3 (7)	11.0 (1)	9.5 (8)	9.2 (17)
ICB (higher = offender used controlling tactics more often) TOTAL
Reason (7-pt. scale) -- 3 item /18	5.06 (81)	7.0 (25)	5.3 (26)	5.0 (6)	3.0 (2)	5.8 (6)	1.6 (16)
Emotional (7-pt.) -- 18 items /108	4.0 (76)	4.5 (20)	2.5 (26)	11.6 (7)	2.0 (2)	12.2 (5)	0.25 (16)
Sexual (7-pt.) -- 5 items /30	0.73 (78)	0.22 (22)	0.08 (26)	1.7 (6)	2.0 (2)	6.0 (6)	0 (16)
Physical (7-pt.) -- 11 items /66	0.33 (80)	0.26 (23)	0 (27)	0.86 (7)	0 (2)	2.8 (5)	0 (16)
Reason (ever done? Y/N) -- 3 item /3	2.2 (115)	2.4 (56)	2.0 (25)	1.3 (7)	3.0 (1)	2.1 (10)	2.0 (16)
Emotional (ever done?) -- 18 items /18	6.4 (111)	5.5 (57)	6.6 (23)	7.5 (6)	4.0 (1)	5.6 (8)	9.5 (16)
Sexual (ever done?) -- 5 items /5	1.1 (117)	0.77 (57)	1.6 (26)	1.3 (7)	1.0 (1)	1.9 (10)	0.94 (16)
Physical (ever done?) -- 11 items /11	2.1 (118)	1.6 (58)	2.0 (27)	3.0 (7)	0 (0)	3.1 (9)	3.5 (16)
MAI (higher = more angry)
Anger Arousal --8 items /40	14.7 (121)	13.8 (59)	15.1 (27)	15.1 (7)	11.5 (2)	16.7 (10)	15.8 (16)
Range of situations -- 7 items /35	18.4 (120)	18.3 (59)	18.6 (21)	18.4 (7)	15.5 (2)	19.5 (8)	18.1 (17)
Hostile outlook -- 4 items /20	9.9 (119)	10.5 (59)	8.7 (27)	9.0 (7)	7.5 (2)	10.1 (8)	10.3 (16)
Anger in -- 5 items /25	9.9 (118)	10.1 (57)	9.2 (26)	11.3 (7)	6.5 (2)	10.6 (10)	9.8 (16)
Anger out -- 2 items /10	6.6 (122)	5.8 (59)	7.3 (27)	7.7 (7)	5.0 (2)	6.9 (10)	7.6 (16)

Table 5A Differences Between Pre-test and Post-test

	PRE (n)	SD (pre)	POST (n)	SD (post)	DIFF SCORE (n)	T	PROB
BIDR
Self-Deceptive Enhance (20 items) /20	6.8 (173)	4.0	6.7 (90)	3.6	-0.16 (77)	-0.44	0.65
Impression Management (20 items) /20	6.0 (165)	4.3	5.9 (86)	4.2	-0.05 (73)	-0.17	0.86
IBWB
TOTAL (31 items) /217	71.7 (251)	26.7	68.0 (164)	26.2	-6.87 (148)	-4.34**	0.0001
Justify (12 items) /84	23.4 (266)	12.9	22.0 (174)	12.7	-2.48 (165)	-2.94**	0.0038
Gain (7 items) /49	15.1 (267)	9.9	14.4 (174)	10.0	-1.10 (164)	-2.20*	0.0290
Help (5 items) /35	11.5 (267)	5.1	10.7 (172)	5.2	-0.74 (163)	-2.04*	0.0431
Punish (5 items) /35	16.8 (271)	6.4	16.0 (176)	6.3	-1.61 (168)	-3.54***	0.0005
Responsible (4 items) /28	11.4 (267)	4.8	10.7 (173)	5.0	-1.29 (163)	-3.20**	0.0016
ICB TOTAL
Reason (7-pt. scale) -- 3 item /18	5.7 (151)	4.7	5.06 (81)	4.4	0.32 (64)	0.61	0.54
Emotional (7-pt.) -- 18 items /108	8.6 (144)	13.9	4.0 (76)	7.8	-0.64 (59)	-0.52	0.60
Sexual (7-pt.) -- 5 items /30	1.8 (156)	3.5	0.73 (78)	2.4	-0.28 (61)	-0.92	0.36
Physical (7-pt.) -- 11 items /66	1.4 (156)	3.5	0.33 (80)	1.5	-0.27 (63)	-1.62	0.11
Reason (ever done? Y/N) -- 3 item /3	2.3 (273)	0.8	2.2 (115)	0.9	0.04 (103)	0.40	0.69
Emotional (ever done?) -- 18 items /18	6.4 (244)	4.1	6.4 (111)	4.6	0.54 (103)	1.48	0.14
Sexual (ever done?) -- 5 items /5	1.3 (272)	1.4	1.1 (117)	1.4	-0.15 (107)	-1.13	0.26
Physical (ever done?) -- 11 items /11	2.4 (276)	2.4	2.1 (118)	2.5	-0.32 (110)	-1.73	0.09
MAI
Anger Arousal --8 items /40	16.1 (271)	6.7	14.7 (121)	5.5	-0.95 (120)	-1.70	0.09
Range of situations -- 7 items /35	19.6 (254)	6.3	18.4 (120)	5.5	-1.28 (111)	-2.12*	0.03
Hostile outlook -- 4 items /20	10.2 (257)	3.7	9.9 (119)	3.7	-0.41 (111)	-1.24	0.22
Anger in -- 5 items /25	11.3 (266)	4.7	9.9 (118)	3.7	-1.19 (116)	-2.95**	0.0039
Anger out -- 2 items /10	7.0 (271)	1.8	6.6 (122)	1.8	-0.01 (121)	-0.09	0.93

Appendix D

CASE CLOSURE SUMMARY

Program Location (Circle): 1 Institution 2 Community

Program Start Date (D,M,Y) _____ _____ _____

Case Closure Date _____ _____ _____

Number of individual sessions attended _____

Number of group sessions attended _____

Reason for Closure: {circle 1 or 2}

1. Regular Closure - Program Completed
2. Early Closure (if early closure, circle as many reasons as apply and check principle reason)

1. Reincarcerated
2. Community supervision terminated
3. Reason Unknown - Client did not return
4. Client doubted program could help
5. Satisfied with gains made
6. Disagreement with therapist
7. Referred elsewhere
8. Client uncooperative
9. Client disruptive of group
10. Client involved in family violent behaviour
11. Client placed in institutional segregation
12. Client transferred to new institution
13. Client participating in alternative institutional program
14. Client released from institution
15. Other (specify) ________________________________________________________

Early case closure was initiated by:

1. Therapist
2. Client
3. Both therapist and client
4. Institutional staff
5. Criminal justice system:

Has there been any change in the client’s material status during the program:

1. No change
2. Separated
3. Divorced
4. Widowed
5. New Partner
6. Multiple changes (Please provide details)

Outcome Ratings

Please rate the extent to which the following statements describe this client’s involvement in the program using the 5 point scales provided.

0 Not At All Descriptive
1 Not Descriptive
2 Uncertain
3 Descriptive
4 Very Descriptive

1	The client’s lack of interest motivation was involved in the reason for case closure.	0 1 2 3 4
2	This client exhibited a high degree of participation in most of the sessions he attended.	0 1 2 3 4
3	This client understood the material presented in the sessions.	0 1 2 3 4
4	This client recognized that his behaviour is abusive.	0 1 2 3 4
5	This client confronted others in the group about their behaviour.	0 1 2 3 4
6	This client attributes responsibility for his abusive behaviour to other people, or to factors such as addiction.	0 1 2 3 4
7	This client is aware of his pattern of abusive behaviour.	0 1 2 3 4
8	This client is able to empathize with those he has abused.	0 1 2 3 4
9	This client demonstrates an awareness of the attitudes and behaviours (e.g. power and control) which promote violence toward women and children.	0 1 2 3 4
10	This client has altered attitudes and behaviours which promote violence toward women and children.	0 1 2 3 4
11	This client is working toward equality within the relationship (i.e., share responsibilities within the family, etc.)	0 1 2 3 4