 |
|||||
|
|
|
||
| Defendant Questionnaire | ACDI-Corrections Version II | DVI-Juvenile |
|
|
The validity of the Domestic Violence Inventory (DVI) was investigated in a sample of 7,941 domestic violence offenders. The DVI has six scales for measuring offender risk of violence (lethality), substance (alcohol and drugs) abuse, controlling behaviors, emotional and mental health problems. Reliability analyses showed that all DVI scales had alpha reliability coefficients of between .88 and .93. DVI scales successfully discriminated between two groups: offenders with 2 or more domestic violence arrests scored significantly higher than offenders who had 1 or no such arrests. The Violence Scale identified 98% of the offenders who admitted to being violent. The Control Scale identified 95% of the offenders who admitted to controlling others. The Alcohol and Drugs scales identified offenders who had been treated for alcohol and drug problems, 99% and 100%, respectively. DVI classification of offender risk was shown to be within 2% of predicted risk range percentile scores for all DVI scales.
Reliability and Validity Study
The enactment of the Violence Against Women Act in 1994 set a precedent that domestic violence would no longer be tolerated and that such violence must stop. The placement of violent offenders in programs that are aimed at long-term solutions to reducing the violence would target offender accountability and, in turn, promote victim safety. These were the concerns addressed by this legislation. Assessment tests can screen violence potential and other offender problems to gain an understanding of offender needs and to select appropriate interventions.
It is clear that domestic violence can lead to more serious degrees of violence, including homicide (Buzawa & Buzawa, 1996). Furthermore, in many domestic violence cases the violence occurred over several months leading up to incidents involving the police (US Department of Justice, 1994). Even mild forms of domestic violence can lead to more serious forms of violence if left unchecked. Identification of violence and other offender problems is an important step toward tailoring effective interventions to help stop domestic violence. Whereas, there are many self-report tests to assess violence in the field of domestic violence (Tolman and Bennett, 1990), few incorporate additional measures to assess other offender problems that co-exist with violence and contribute to domestic violence as well.
Aggressiveness, violence, controlling behaviors, substance abuse and stress coping abilities are factors that have been demonstrated to be relevant to domestic violence. Brown, Werk, Caplan and Seraganian (1999) found that as the severity of the substance abuse increased, so too did the dangerousness and frequency of abusive behaviors. Brown, et al. (1999) also found that dual-problem men had more personality, psychosocial and hostility problems and a history of past arrests. Stets (1988) reported that men acknowledged that control of the women’s behavior was a goal of their violence. Rice (1997) found that psychopathology was a good predictor of reassault for violent offenders in general. These studies suggest that assessment of domestic violence offenders should entail more than just a measure of violence. Furthermore, a test that is multidimensional lends itself to recidivism prediction.
Placement of offenders in appropriate interventions is important. Offender programs that incorporate cognitive and behavioral intervention reduce recidivism by an average of 15% (Andrew, et al., 1989). Certain programs for high-risk offenders reduced recidivism by as much as 25% (Andrews & Bonta, 1994). Moreover, when offenders are properly targeted and matched to the appropriate program, recidivism is reduced by an average of 25% to 50% (Carey, 1997). Rooney and Hanson (2001) reported that men who dropped out after starting treatment tended to have unstable lifestyles (e.g., substance abuse problems, criminal history, and unemployment) and to have inflicted more severe abuse than those who completed treatment. Multidimensional assessment tests can provide important predictor variables for the prediction of recidivism. Few domestic violence tests attempt to determine the risk of reassault.
Researchers are recognizing the importance of identifying offenders who are at risk for reassault (Jones and Gondolf, 2001). According to Jones and Gondolf (2001), researchers have focused primarily on personality characteristics and prior behavior, and have suggested possible "risk markers" for reassault (see Hamberger and Hastings, 1990, Holtsworth-Munroe and Stuart, 1994, and Rice, 1997). Yet, in everyday assessment settings, practitioners do not have time to administer and score multiple tests nor do they have the wherewithal to calculate predictions of reassault from pieced-together data. A multidimensional test can provide them with relevant risk and needs assessment on a number of dimensions easily, efficiently and timely. For this purpose, a reliable and valid test is essential.
The present study investigated the validity of the Domestic Violence Inventory (DVI). The DVI is a multidimensional test that was developed to meet the needs of judicial court screening and assessment. DVI scales measure violence (lethality) tendencies (Violence Scale), controlling attitudes and behaviors (Control Scale), alcohol and drug abuse severity (Alcohol & Drugs Scales) and emotional or mental health problems (Stress Coping Abilities Scale). In addition, there is the Truthfulness Scale to measure offender truthfulness, denial and minimization while completing the test. There is some evidence that batterers substantially underreport or minimize violence incidents (Heckert and Gondolf (2000). Truthfulness Scale scores are used for truth-correcting other scale scores.
This study sought to validate the DVI in a sample of domestic violence offenders who were processed as part of standard offender evaluation procedures in court and community service programs. The data for this study was obtained from the agencies that used the DVI in their assessment programs. Two methods for validating the DVI were performed in this study. The first method (discriminant validity) compared scale scores between two offender groups. Group 1 consisted of offenders who had one or no domestic violence arrest. Group 2 consisted of offenders who had two or more domestic violence arrests. It was hypothesized that multiple offenders (Group 2) would score significantly higher than offenders who had 0 or 1 arrest (Group 1). Multiple offenders would be expected to score higher on the Violence Scale because having a second domestic violence arrest is indicative of a serious violence problem.
The second validation method (predictive validity) examined the accuracy at which the DVI identified violence prone and controlling offenders, problem drinkers and problem drug abusers. In the DVI, alcohol and drug treatment information is obtained from the offenders’ responses to test items. Undoubtedly, there are some offenders who have an alcohol or drug problem but have not been in treatment. Nevertheless, offenders who have been in treatment would be expected to score in the corresponding scale’s problem range. For treatment information the following test items were used, "I go to Alcoholics Anonymous or Rational Recovery meetings because of my drinking." "How would you describe your desire to get alcohol treatment? Want help/may need help/undecided/no desire." "I go to Narcotics Anonymous or Cocaine Anonymous meetings because of my drug use." "How would you describe your desire to get drug treatment? Want help/may need help/undecided/no desire."
In regards to violence and control, offenders direct admissions of problems were used as the criteria. The violence test item used was, "How would you describe your domestic violence? A serious problem/a moderate problem/a mild problem/no problem." The control test item was, "In many relationships one person dominates and the other person submits to their control. I usually dominate and control."
For the predictive validity analyses, offenders were separated into two groups, those who had treatment or admitted problems and those who did not have treatment or did not admit to problems. Then, offender scores on the relevant DVI scales were compared. It was predicted that offenders with an alcohol and/or drug treatment history would score in the problem risk range (70th percentile and above) on the Alcohol and/or Drugs Scales. Similarly, offenders who admitted being violent or controlling others would score in the problem risk range on the Violence and Control Scales, respectively. Non-problem was defined in terms of low risk scores (39th percentile and below). The percentage of offenders who had been in treatment or admitted problems and also scored in the 70th percentile range and above was considered a correct identification of problems. High percentages of offenders with treatment or admit problems and elevated problem risk scores would indicate the scales were valid.
Subjects
There were 7,941 domestic violence offenders tested with the DVI. Data for this study was provided by the court service providers, probation departments and professional community service agencies that use the DVI. Test data were collected during the year 2001. There were 6,565 males (82.7%) and 1,376 females (17.3%). The ages of the participants primarily ranged from 19 through 60 as follows: 19 & Under (6%); 20-29 (36%); 30-39 (35%); 40-49 (19%); 50-59 (4%) and 60 & Over (1%). The average age of the males was 32.7 (SD 9.03) and the average age of the females was 30.9 (SD 8.80). The demographic composition of the participants was as follows. Race/Ethnicity: Caucasian (63%); Black (19%), Hispanic (13%) and Other (5%). Education: Eighth grade or less (7%); Some high school (27%); High school graduate/GED (47%); Some college (14%) and College graduate (4%). Marital Status: Single (44%); Married (36%); Divorced (12%); Separated (7%) and Widowed (1%).
The participants’ criminal histories were obtained from their DVI answer sheets. The participants reported this information although the staff were to verify the information provided. Nearly 80 percent of the participants or 6,325 offenders reported having one or no domestic violence arrest. 5,099 or 77.7 percent of the males had one or no arrest. 1,226 or 89.1 percent of the females had one or no arrest. These offenders were designated as Group 1. Fourteen percent of the participants had two domestic violence arrests, four percent had three arrests and three percent had four or more domestic violence arrests. The offenders with two or more domestic violence arrests (multiple offenders) were designated as Group 2. There were 1,616 offenders or 20.3 percent of the participants in Group 2. 1,466 or 22.3 percent of the males were multiple offenders and 150 or 10.9 percent of the females were multiple offenders.
Twenty-seven percent of the participants had been arrested for assault. Nearly one-fourth (22%) of the participants had one alcohol arrest, 10 percent had two arrests and 13 percent had three or more arrests. Eleven percent of the participants had one drug arrest, three percent had two arrests and 2.2 percent had three or more arrests.
Procedure
Participants completed the DVI as part of the normal routine for domestic violence offender evaluations in court service and community service programs. These practitioners administered the DVI to offenders when they first entered their programs. Probation departments used the DVI to select appropriate levels of supervision and interventions.
The DVI contains six measures or scales. These scales are briefly described as follows. The Truthfulness Scale measures the truthfulness, denial and minimization of the respondent while taking the DVI. The Alcohol Scale measures severity of alcohol use or abuse. The Drugs Scale measures severity of drug use or abuse. The Control Scale measures controlling behaviors that affect self and others. The Violence Scale measures offender proneness to commit violence. The Stress Coping Abilities Scale measures ability to cope with stress.
The inter-item reliability (alpha) coefficients for the six DVI scales are presented in Table 1. All scales were highly reliable. All of the alpha reliability coefficients for all DVI scales were at or above 0.88. These results demonstrate that the DVI is a reliable test for domestic violence offender assessment.
DVI Scales
Number of Items
Alpha
Truthfulness Scale 20
.88
Alcohol Scale 22
.93
Control Scale 19
.88
Drugs Scale 22
.91
Violence Scale 28
.90
Stress Coping Abilities 40
.93
Because there were many more offenders in Group 1 (one or no domestic violence arrest) than Group 2 (multiple offenders), a random sample from Group 1 was taken to equal the number of Group 2 offenders (N = 1,616). In the comparisons of DVI scale scores using these equal number groups, Group 2 scored significantly higher than Group 1 offenders on the Alcohol Scale, Control Scale, Drugs Scale, Violence Scale and Stress Coping Abilities Scale. Higher scores on these DVI scales are associated with more severe problems. These discriminant validity results are presented in Table 2. The scale scores presented in the table are derived from test items alone by adding the points assigned to each test item. These scores do not include points for court-related history or truth-correction. The scores can be compared between the groups because the scores are not inflated by court history.
Table 2. T-test Comparisons between Group 1 (1 or no arrest) and
Group 2 (2 or more domestic violence arrests).
DVI Scales
Group 1
Group 2
T-value
Mean
SD
Max
Mean
SD
Max
Truthfulness Scale
9.09
5.45
21
7.88
5.21
21
t = 6.44* Alcohol Scale
5.97
8.65
38
10.84
11.74
38
t = 13.41* Control Scale
3.47
3.95
19
5.01
4.65
19
t = 10.20* Drugs Scale
3.25
6.66
39
5.09
7.88
38
t = 7.14* Violence Scale
9.03
8.93
47
13.68
10.42
45
t = 13.63* Stress Coping Abilities
111.00
42.11
235
99.49
38.77
237
t = 8.08* * Significant at the p < .001 level.
Note: The Stress Coping Abilities Scale is reversed in that the higher the score the better one copes with stress.
Table 2 shows that scale scores for Group 2 were significantly higher than scores for Group 1 on all DVI scales except the Truthfulness Scale. As expected, multiple offenders scored significantly higher on the Alcohol Scale, Control Scale, Drugs Scale, Violence Scale and Stress Coping Abilities Scale than did offenders with one or no arrest. The Truthfulness Scale shows that Group 1 scored significantly higher than Group 2. Truthfulness Scale results suggest that multiple offenders did not minimize their problems or fake good when tested as much as offenders with one or no arrest. Multiple offenders may have largely stopped attempting to minimize their problems in court related settings. Having a history of arrests lessens the likelihood that a multiple offender will deny problems. Whereas offenders with one or no arrest, who are unfamiliar with court settings, consequences or assessment, may try to fake good in order to lessen the impact of their situation.
The Alcohol, Control, Drugs, Violence and Stress Coping Abilities Scales results support the discriminant validity of the DVI. The offenders who were believed to have more severe problems (multiple offenders) scored significantly higher on these scales than offenders with one or no arrest. The Stress Coping Abilities Scale results indicate that offenders who have multiple domestic violence arrests demonstrate poorer stress coping skills than do offenders with one or no arrest. It is generally accepted that stress exacerbates emotional and mental health symptomatology.
Correlation coefficients between offenders’ DVI scale scores and arrests records are presented in Table 3. Again, the scale scores did not include points for court history or truth-correction and the equal number groups was used. The most notable of these correlation results includes the finding that the Violence Scale was not very well correlated with domestic violence or assault arrests. This finding suggests that there was a tendency for multiple domestic violence offenders to minimize their violence, and that offenders with one or no arrest can be nearly as violent as are multiple offenders. Because multiple offenders scored significantly higher than offenders with 1 or 0 arrest, it would be expected that the correlation between violence assessment and arrests be much higher. It is likely that number of arrests is not a good indicator of a domestic violence offender’s potential for violence. An individual’s assessment of his or her own violence should be taken into account rather than their arrest history.
A similar correlation was found between drug arrests and the Drugs Scale. Again, these offenders may have been unwilling to admit to having drug problems in court assessment settings. The Alcohol Scale was correlated with alcohol arrests.
Table 3. Correlations between Arrest Records and DVI Scales
Alcohol
ScaleControl
ScaleDrugs
ScaleViolence
ScaleStress Coping
Domestic violence arrests
.18
.16
.14
.20
.13
Assault arrests
.08
.10
.09
.11
.11
Alcohol arrests
.40
.19
.21
.19
.11
Drug arrests
.12
.06
.23
.05
.07
Total arrests
.28
.20
.25
.22
.14
Truthfulness Scale
-.33
-.54
-.27
-.54
-.39
Alcohol Scale
.40
.46
.43
.30
Control Scale
.37
.83
.49
Drugs Scale
.39
.29
Violence Scale
.53
Further analyses of the correlation results indicate that there were significant correlations among the DVI scales. The between scales correlation coefficients demonstrate that offenders who score high on one scale tend to score high on other DVI scales. These finding suggest that domestic violence offenders have more problems than just those associated with violence. They have alcohol, drugs, control and emotional problems as well.
There was a very high correlation between the Violence Scale and the Control Scale, r = 0.83. It is apparent that violence and controlling behavior are closely related characteristics in domestic violence offenders. Either control begets violence or control and violence co-occur in instances of domestic violence.
Predictive validity results for the correct identification of problems (violence tendencies, control, drinking and drug abuse problems) are presented in Table 4. Table 4 shows the percentage of offenders that had or admitted to having problems and who scored in the problem risk range on the selected DVI scales in comparison to offenders who scored in the low risk range. For the Alcohol and Drugs Scales problem behavior means the offender had alcohol or drug treatment. For the Violence Scale the offender admitted to having a serious or moderate domestic violence problem. For the Control Scale the offender admitted to dominating and controlling others. In these analyses the total number of participants (7,941) were included.
For the Alcohol Scale comparisons between problem risk and low risk offenders, there were 1,382 offenders who reported having been in alcohol treatment. These offenders were considered problem drinkers. Of these 1,382 offenders, 1,370 individuals, or 99.1 percent, had Alcohol Scale scores at or above the 70th percentile. The Alcohol Scale correctly identified nearly all (99%) of the offenders classified as problem drinkers. These results validate the DVI Alcohol Scale.
The DVI Drugs Scale also identified offenders who have had drug problems. There were 1,337 offenders who reported having been in drug treatment. All 1,337 individuals, or 100 percent, had Drugs Scale scores at or above the 70th percentile. These results support the validity of the DVI Drugs Scale.
For Violence Scale comparisons there were 1,202 offenders who admitted having serious or moderate domestic violence problems. Of these 1,202 offenders, 1,178 individuals or 98 percent had Violence Scale scores in the problem range (70th percentile and above). These results validate the Violence Scale. Control Scale comparisons found that for the 942 offenders who admitted to dominating and controlling others, 893 or 94.8 percent had Control Scale scores in the problem range. This result supports the validity of the Control Scale.
Table 4. Predictive Validity of the DVI
DVI Scales
Correct Identification of Problem Behavior
Alcohol 99.1%
Drugs 100%
Violence 98.0%
Control 94.8%
For ease in interpreting domestic violence offender risk, DVI scale scores were divided into four risk ranges: low risk (zero to 39th percentile), medium risk (40 to 69th percentile), problem risk (70 to 89th percentile), and severe problem risk (90 to 100th percentile). By definition the expected percentages of offenders scoring in each risk range (for each scale) is: low risk (39%), medium risk (30%), problem risk (20%), and severe problem risk (11%). Scores at or above the 70th percentile would identify offenders as having problems.
The above predictive validity results lend support for using these particular percentages. The 70th percentile cut off for problem identification correctly classified 95 percent or more of problem offenders. The low risk level of 39 percent avoids putting a large percentage of offenders into a "moderate" range.
Risk range percentile scores were derived by adding points for test items, truth correction points and criminal history points, if applicable, then converting to percentages. These results are presented in Table 5. Risk range percentile scores represent degree of severity. Analysis of the DVI risk range percentile scores involved comparing the offender’s obtained risk range percentile scores to predicted risk range percentages as defined above. These percentages are shown in parentheses in the top row of Table 5. The actual percentage of offenders falling in each of the four risk ranges, based on their risk range percentile scores, was compared to these predicted percentages. The differences between predicted and obtained are shown in parentheses.
Table 5. Accuracy of DVI Risk Range Percentile Scores
Scales
Low Risk
(39%)Medium Risk (30%)
Problem Risk (20%)
Severe Problem (11%)
Truthfulness Scale 39.5
(0.5) 30.9
(0.9) 19.4
(0.6) 10.2
(0.8) Alcohol Scale 38.9
(0.1) 30.5
(0.5) 20.0
(0.0) 10.6
(0.4) Control Scale 37.1
(1.9) 30.7
(0.7) 22.1
(2.1) 11.1
(0.1) Drugs Scale 40.6
(1.6) 30.5
(0.5) 18.6
(1.4) 10.3
(0.7) Violence Scale 38.0
(1.0) 30.1
(0.1) 20.7
(0.7) 11.1
(0.1) Stress Coping Abilities 39.1
(0.1) 30.0
(0.0) 20.0
(0.0) 10.9
(0.1) As shown in Table 5, the objectively obtained percentages of participants falling in each risk range were very close to the expected percentages for each risk category. All of the obtained risk range percentages were within 2.1 percentage points of the expected percentages and many (20 of 24 possible) were within one percentage point. Only one obtained percentage was more than two percent from the expected percentage and that was the Control Scale problem risk classification, which was 2.1 percent higher than predicted. These results demonstrate that risk range percentile scores are accurate.
Gender differences between male and female scale scores are shown in Table 6. The offender groups were studied separately. The total number of participants (7,941) was included in these analyses. These results demonstrated that males scored significantly higher than females on the Truthfulness, Alcohol and Drugs Scales. The Stress Coping Abilities Scale showed that females scored significantly higher than males. The differences between males and females on the Violence Scale were not significant for either Group 1 or Group 2. The Control Scale differences were significant for the multiple offenders (Group 2) but not for the offenders in Group 1. On all DVI scales the maximum scale scores of the females were very close to those of the males.
Males tend to have more substance abuse problems than females and they tend to minimize their problems more. Females have more emotional or stress coping problems than males. The multiple offenders showed that females had more control problems than males. However, males and females did not differ in terms of violence. The lack of significant differences between males and females on the Violence Scale was unexpected. However, the females in this study were domestic violence offenders. It appears that violence does not differ across genders when both are domestic violence offenders.
Table 6. Comparisons between Males and Females
Group 1
DVI Scales
Males
Females
T-value
Mean
SD
Max
Mean
SD
Max
Truthfulness Scale
9.21
5.47
21
8.43
5.52
21
t = 4.48* Alcohol Scale
6.60
9.22
38
4.78
8.19
38
t = 6.82* Control Scale
3.38
3.88
19
3.70
4.02
19
t = 2.56** Drugs Scale
3.34
6.47
39
2.92
6.63
37
t = 2.05** Violence Scale
9.16
8.90
47
9.24
9.37
43
t = 0.27 Stress Coping Abilities
112.12
41.57
237
104.25
42.96
235
t = 5.92* Group 2
DVI Scales
Males
Females
T-value
Mean
SD
Max
Mean
SD
Max
Truthfulness Scale
8.03
5.25
21
6.45
4.61
21
t = 3.94* Alcohol Scale
11.16
11.88
38
7.67
9.81
38
t = 4.07* Control Scale
4.99
4.68
19
5.23
4.40
19
t = 0.59 Drugs Scale
5.26
8.01
39
3.37
6.23
37
t = 3.43* Violence Scale
13.63
10.46
47
14.17
10.04
43
t = 0.60 Stress Coping Abilities
100.19
38.83
237
92.73
37.63
235
t = 2.25* * Significant at the p < .001 level. ** Significant at the p < .05 level.
Note: The Stress Coping Abilities Scale is reversed in that the higher the score the better one copes with stress.
Conclusions This study demonstrated that the DVI is a reliable and valid assessment test for domestic violence offenders. Reliability results showed that all six DVI scales were highly reliable. Reliability is necessary in domestic violence screening tests for accurate measurement of offender risk.
Discriminant validity analyses demonstrated that multiple offenders (had prior domestic violence arrests) scored significantly higher than offenders with one or no arrest. Predictive validity analyses demonstrated that the DVI identified domestic violence offenders who had violence, control and substance abuse problems. The Violence Scale identified offenders who admitted having domestic violence problems. The Control Scale correctly identified offenders who admitted dominating and controlling others. The Alcohol and Drugs Scales correctly identified offenders who were in or desired treatment for alcohol and drugs, respectively. Furthermore, obtained risk range percentages on all DVI scales very closely approximated predicted percentages. These results support the validity of the DVI.
Correlation results show that problems with violence, substance abuse, control and stress coping co-exist in domestic violence offenders. These results support the Brown, et al. (1999) finding that as the severity of substance abuse increased the severity of violence potential also increased. This study also supports the Stets (1988) finding that control and violence co-exist in domestic violence offenders. Identification of these problems and prompt intervention that target these problems can reduce a domestic violence offender’s risk of recidivism or future violence.
One of the most important decisions regarding a domestic violence offender is what supervision level and/or intervention program is appropriate for the offender. The DVI can be used to tailor intervention (levels of supervision and treatment) to each domestic violence offender based upon their assessment results. Low scale scores are associated with low levels of supervision as well as intervention and treatment, whereas high scale scores relate to more intense intervention/treatment recommendations and levels of supervision. Placing domestic violence offenders in appropriate treatment can enhance the likelihood that an offender will complete treatment, benefit from program participation and change their violent behavior.
This study supports the reliability and validity of the DVI. The next level of research is the development of a predictor equation for domestic violence recidivism. Many offender risk (criminal history) and needs (DVI scale scores) predictor variables are contained in the DVI.
References Andrews, D.A., Zinger, I., Hoge, R., Bonta, J. Gendreau, P., and Cullen, F. (1989). Does correctional treatment work? A clinically relevant and psychologically informed meta-analysis. Paper presented at the Research Seminar of National Associations Active in Criminal Justice, Ottawa, Canada.
Andrews, D., and Bonta, J. (1994). The Psychology of Criminal Conduct. Cincinnati, OH: Anderson Press.
Brown, T.G., Werk, A., Caplan, T., and Seraganian, P. (1999). Violent substance abusers in domestic violence treatment. Violence and Victims, 14, 179-190.
Buzawa, E.S. & Buzawa, C.G. (1996). Domestic violence: The criminal justice response (2nd edition). Thousand Oaks, CA: Sage Publications.
Carey, M. (1997). Cog probation. American Probation and Parole Association, Perspectives, Spring, 27-42.
Current or Recently Completed Research in Domestic Violence and Child Abuse. Washington, DC: U.S. Department of Justice, National Institute of Justice, 1994.
Hamberger, L.K., and Hastings, J. (1990). Recidivism following spouse abuse abatement counseling: Treatment program implications. Violence and Victims, 5, 157-170.
Heckert, D.A., and Gondolf, E.W. (2000). Assessing assault self-reports by batterer program participants and their partners. Journal of Family Violence, 15, 181-197.
Holtsworth-Munroe, A., and Stuart, G. (1994). Typologies of male batterers: Three subtypes and differences among them. Psychological Bulletin, 116, 476-497.
Jones, A.S., and Gondolf, E.W. (2001). Time-varying risk factors for reassault among batterer program participants. Journal of Family Violence, 16, 345-359.
Rice, M. (1997). Violent offender research and implications for the criminal justice system. American Psychologist, 52, 414-423.
Rooney, J. and Hanson, R. K. (2001). Predicting attrition from treatment programs for abusive men. Journal of Family Violence, 16, 131-149.
Stets, J.E. (1988). Domestic violence and control. New York: Springer-Verlag.
Tollman, R.M., and Bennett, L.W. (1990). A review of quantitative research on men who batter. Journal of Interpersonal Violence, 5, 87-118.
Violence Against Women Grants Office. Stalking and Domestic Violence. Washington, DC: US Department of Justice, Office of Justice Programs, 1998.
Donald D. Davignon, Ph.D.
Senior Research Analyst
COURTESY LINKS
Behavior Data Systems www.bdsltd.com New Client Orders
* * *
