Individuals convicted of committing domestic violence are often court mandated to attend a batterer intervention program (BIP). These programs aim to change attitudes and beliefs related to domestic violence and are expected to minimize the future incidence of violence in families (Adams 2000). The majority of the programs mandate that offenders attend Duluth-Model BIPs (Pence & Paymar, 1993 as cited in Babcock et al. 2004). According to the Duluth Model, violence results from a patriarchal culture, which fosters male privilege and entitlement. Advocates of the Duluth Model argue that a patriarchic society leads men to exert power and control, manifested as violence, over intimate partners (Dutton and Corvo 2006). Thus, the goal of Duluth-Model BIPs is to challenge men’s perceived right to exert power over women, with the ultimate goal of ending violence used to control women. In line with this belief, BIPs that employ the Duluth Model are often punitive in nature, using confrontation to address issues of power and control in ways that often ignore other causes of violent behavior. Importantly, a meta-analysis of BIPs found Duluth-Model BIPs to be minimally effective and no more successful at reducing violence than cognitive-behavioral therapy (Babcock et al. 2004; see also Gondolf 2000, 2002, 2004).

In light of such limitations, researchers have questioned the utility of the Duluth Model and claim that this ‘one-size fits all model’ is insufficient to address the diverse group of individuals who commit domestic violence. Instead, critics argue that BIPs built around the Duluth Model actually serve as intensive probation monitoring systems for domestic violence offenders with no legitimate, or at best a limited ability, to reduce violence (e.g., Dunford 2000). As we will discuss in detail below, primary limitations of the Duluth Model include its failures to address external risk factors that cause violence (e.g., violence in the family of origin, stress), its inability to foster a therapeutic relationship with clients, its lack of instruction in important skills such as anger management, and its ineffectiveness in treating both men and women. These oversights may explain the particularly high dropout rates that characterize this traditional BIP (e.g., Davis et al. 2000; Dutton and Corvo 2006).

Limitations of the Duluth Model

The primary critique of the Duluth Model is that it addresses issues of power and control while ignoring other external risk factors associated with domestic violence, including perpetrators’ personal experience with abuse, stress, and anger (Dutton and Corvo 2006). Violence in the family of origin is associated with later use of violence (e.g., Bronfenbrenner 2005). Further, increased life stress, such as experiencing periods of poverty, is linked to higher rates of intimate partner violence (e.g., Carlson et al. 2000). Men who use domestic violence experience higher levels of anger and hostility than do nonviolent men (e.g., Stuart and Holtzworth-Munroe 2005). Ignoring these external causes of violence likely contributes to the lack of positive treatment outcomes found for Duluth-Model BIPs (e.g., Babcock et al. 2004).

Advocates for alternatives to the Duluth Model also take issue with the way the model’s tenets hinder the establishment of a therapeutic relationship between offenders and rehabilitation professionals (e.g., Dutton and Corvo 2006). Duluth-Model BIPs rely heavily on criticizing and shaming offenders’ violent acts, which inhibits opportunities for honest, vulnerable disclosure. The use of these punitive strategies is particularly problematic given the clientele of BIPs. For example, many individuals who commit violent acts come from violent homes, where they likely experienced shame from their families’ use of criticism and abusive punishment (Dutton 2006). In fact, individuals who engage in violent behaviors often suffer from symptoms of post-traumatic stress disorder or other types of psychological disorders (Dutton 1995). Whereas Duluth-Model BIPs focus on holding perpetrators accountable for their offenses, these programs neglect to address the abuse or trauma the perpetrators themselves have experienced. Ignoring violence in BIP clients’ families of origin contributes to ineffectiveness in attempting to end a likely cycle of violence by treating potential victims of violence only as perpetrators. Given that the therapeutic relationship is arguably one of the most valuable components of successful therapy (e.g., Horvath and Symonds 1991), it is essential for an effective BIP to create a safe environment that invites honest disclosure. Unfortunately, Duluth-Model BIPs are not technically categorized as therapy programs and, as such, are not required to follow confidentiality practices that contribute to such therapeutic disclosures (Rosenbaum and Leisring 2001).

It is equally important for BIPs to teach specific skills commonly omitted from Duluth-Model BIPs, such as stress and anger management, which reduce the use of violent behaviors. The development of self-control promotes the inhibition of hostile impulses (Gottfredson and Hirschi 1990). Thus, teaching anger management is necessary, particularly for individuals who use violence to resolve a conflict (Pence 2002). Indeed, violent men who go through counseling show lower levels of indirect hostility, resentment, and irritability whereas violent men who do not undergo treatment are more likely to commit assault, to demonstrate indirect hostilities, and to feel resentment (Barnett et al. 1991). In sum, addressing external factors that contribute to violence by teaching skill sets, such as anger and stress management, and by establishing therapeutic relationships is an essential element of an effective BIP.

In addition to ignoring external risk factors associated with violence, Duluth-Model BIPs also ignore the reality of gender equity in rates of violence perpetration. That is, because the Duluth Model assumes that violence stems from patriarchal beliefs, Duluth-Model BIPs target male offenders, a notable limitation given that men are not the only clientele of BIPs. Domestic violence is perpetrated at comparable rates by men and women in the United States (e.g., Bates et al. 2014; Schafer et al. 1998; Straus and Gelles 1990). In fact, a meta-analysis, which included 60,000 individuals, found that women are actually violent more frequently than are men, particularly later in life (Archer 2000).Footnote 1 Moreover, changes in the criminal justice system have resulted in an increase in the number of women arrested for committing acts of domestic violence (Hamberger and Arnold 1989). Specifically, mandatory arrest policies have resulted in dual-arrest practices—both parties are arrested in domestic violence cases—and thus, in some states female arrests constitute up to 30 % of all domestic violence arrests (Hirschel et al. 2007).

Given these patterns and changes, it is essential to address gender-specific issues within the therapeutic environment. For example, female perpetrators of domestic violence tend to be influenced more by lower socioeconomic status and their experience of significant life stressors; therefore, women may benefit from practical support for these stressors (i.e., assistance with housing, employment, parenting) in an effort to reduce violence (Dowd et al. 2005). It is also important for treatment evaluation programs to include both men and women in their samples. Unfortunately, to date, most studies have focused exclusively on men (e.g., Craig et al. 2006; Eckhardt et al. 2008). As a result, it is unclear whether treatment evaluation programs are equally successful for both men and women.

A New Approach

In response to the limitations of the Duluth-Model BIPs, domestic violence treatment providers developed Resolution Counseling Intervention Programs (RCIPs), psychoeducational treatment groups for violent men and women. RCIPs are founded on counseling principles that foster rapport between therapists and clients. Similar to other BIPs, RCIPs address topics such as taking accountability for violent actions, learning about the impact of violence on families, creating safety plans, and understanding definitions of violence. These topics are addressed in a supportive environment with trained counseling professionals who intentionally avoid shaming clients. Departing from the tenets of the Duluth Model, which focus primarily on issues of power and control, RCIPs teach individuals the skills they need to maintain healthy relationships, including anger management, conflict resolution, respectful communication, healthy relationship practices, and coping with violence in the family of origin. RCIPs’ basis in counseling principles and techniques allows facilitators to cater sessions to meet the specific vulnerabilities of clients in different support groups. Tailoring violence intervention to the needs of RCIPs’ clients offers more diversity in treatment than the ‘one-size-fits-all’ Duluth Model.

Overview of the Current Study

The goal of the current study was to test the effectiveness of RCIPs in reducing both psychological and physical violence. Given the unique nature of RCIPs, we hypothesized that changes would occur in variables related to violence such as stress, desire for change, accountability, safety planning, and anger management (i.e., variables that RCIPs target specifically). In addition to these measures, we were also interested in whether constructs that are targeted in traditional programs, such as taking accountability for violence and reducing controlling behaviors, would change even if they were not the sole focus of RCIPs. We hypothesized that the program would be effective in both reducing levels of violence and changing attitudes and behaviors that contribute to violence.

Method

Participants

One hundred and forty-nine participants (35 women; 110 men; four unidentified) from Austin, TX completed a violence intervention programFootnote 2. The sample ranged in age from 21- to 61-years-old (M = 35.06 years) and was ethnically diverse (44.8 % self-identified as White; 23.9 % as Black or African American; 21.6 % as Hispanic; 1.5 % as Asian, 8.2 % as “Other”). The majority of the sample had graduated from high school (28.1 %) or had some college experience (34.2 %; 14.5 % of the sample had not completed high school; 6.8 % were currently in college; 26.4 % had at least a college degree). Of the 149 participants, 37.0 % were living with their partners at the outset of the treatment program and study. Approximately one-third (35.6 %) of the sample was married. The number of unmarried participants living with partners (28.7 %) was less than the number of married participants who reported living with their spouses (51.9 %). Additionally, 68.6 % of participants reported having children; 48.0 % of the clients with children had their children living with them at least part of the time. At the end of the study, participants reported slightly lower rates of marriage (33.1 %) and slightly higher rates of living with their partners (married or unmarried; 50.0 %) than those reported at the beginning of the study.

Procedure

Pre-Treatment Survey

The initial survey was completed during the program’s intake process. It is common during BIP intake for clients to sit in a room as they wait to speak with the staff member who will assign them to a group and tell them the date of the first group session. During this waiting period, participants were given the option to complete the initial survey. Participants were assured that their responses were anonymous and would not be viewed by the RCIP staff. Willing participants completed a questionnaire that included measures of psychological violence, physical violence, accountability for violent behavior, controlling behaviors, safety planning, anger management, perceived stress, and desire for change (detailed below). Upon completion of the survey, participants’ surveys were placed in sealed envelopes and picked up by a member of the research team. Program staff never had access to clients’ responses.

Treatment Program

The treatment groups consisted of same-gendered participants and met once per week for two-hour sessions. The program duration was either 21 or 30 weeks, depending on the appointment of the court (which typically differed as a function of the severity of the violence and the estimated likelihood that individuals would become repeat offenders). Groups were led by a licensed counseling professional (counselors were predominantly women) and consisted of up to 12 offenders. The small group size allowed for highly individualized and group-needs-based treatment. The completion rate of the RCIP was approximately 58.9 % (60.8 % for men and 57.0 % for women). The current program has a comparable rate of attrition (41.1 %) to other therapeutic and psychoeducational programs (39.8 %) but a lower rate of attrition in comparison to Duluth-Model intervention programs (55.4 %; Babcock et al. 2004).

Post-Treatment Survey

The completion survey was administered upon participants’ graduation from the program. At clients’ final counseling sessions, staff members administered a survey almost identical to the pre-treatment survey. This survey included all of the measures that participants had completed at the program’s outset, with instructions for participants to complete each question in the survey as it applied to them “TODAY” (i.e., as they were graduating from the program). Additionally, the survey included questions assessing clients’ experiences in the program. Again, participants were assured that all answers would remain confidential and anonymous. Participants placed their surveys in a sealed envelope that was then picked up by a member of the research team.

Participant Compensation

Participants were allowed to attend one group session for free or were offered the monetary equivalent of a free session ($20) for each survey completed. Participants received either $40 total in cash (n = 106), their first and last session for free (n = 31), or $20 in cash and one free session (n = 12).

Measures

Perceived Stress (Pre- and Post-Treatment)

Stress is associated with the use of interpersonal violence (Cano and Vivian 2001). One advantage of using a counseling-based program is that counselors teach clients how to better manage their stress, which should in turn reduce violence. In order to assess the amount of stress participants perceived in their lives, we used the 10-item Perceived Stress Scale (PSS-10; Cohen and Williamson 1988). The PSS was designed to measure how unpredictable, uncontrollable, and overwhelming respondents find the accumulation of stressors in their lives; these three issues are key components to the appraisal of stress (Cohen et al. 1986). Participants were asked to report, on a frequency scale from 1 (never) to 5 (very often), how often in the last month they had experienced a number of specific perceptions. Five items assessed positive appraisals (e.g., “How often have you felt that things were going your way?”), and five items assessed negative appraisals (e.g., “How often have you been upset because of something that happened unexpectedly?”). All positive items were reversed coded and then responses were summed to create a perceived stress total (pre-treatment α = .85, n = 132; post-treatment α = .82, n = 132).

Desire for Change (Pre- and Post-Treatment)

In order for an offender intervention program to successfully alter behavior, participants must have a desire to change their behaviors (Williamson et al. 2003). As such, we included seven of the original twelve items of the Anger Readiness to Change Questionnaire (ARCQ; Williamson et al. 2003) to measure individuals’ desire for change. Adapted from the Readiness to Change Questionnaire (RCQ; Heather et al. 1993) used in alcohol research, the ARCQ assesses three different stages of change: pre-contemplation, contemplation, and action. Participants were asked to report how much they agreed with change-related statements on a scale ranging from 1 (strongly disagree) to 5 (strongly agree). The scale included five items that indicated participants wanted to see change in their violent behavior (e.g., “I am at the stage where I should think about managing my anger.”) and two items that indicated participants were not ready for change (e.g., “There is no need for me to think about changing how I deal with anger.”). Items that indicated reluctance to change were reversed scored and then all items were summed to create a total desire for change score (pre-treatment α = .76, n = 138; post-treatment α = .71, n = 132).

Accountability (Pre- and Post-Treatment)

Offenders’ explanations for their emotions and behaviors are related to the likelihood that they will be open to receiving help for violence issues and ultimately become less violent (Holtzworth-Munroe 1988). We created a measure to assess the level of accountability individuals assumed for the act(s) that led to them being in the RCIP. Specifically, participants were asked to report on a scale from 1 (not true) to 7 (very true) the degree to which they agreed with eight items that explained why they had entered the RCIP. Of these items, four items were rationales that indicated taking responsibility for violent acts (e.g., “I have myself to blame for what I did.”), and four items were rationales that indicated a deflection of responsibility (e.g., “I am here because I was provoked.”). Deflecting responsibility items were reverse scored and then the items were summed to create a single composite of accountability (pre-treatment α = .62, n = 135; post-treatment α = .71, n = 125).

Safety Planning (Pre- and Post-Treatment)

In order to avoid violent situations, it is necessary to be able to recognize escalating tension and react in a way that will decrease the probability that violence will occur (Bryant 1994). We thus created a measure of safety planning to determine participants’ potential for defusing situations that could become violent. Participants were asked to report whether or not they used each of five healthy strategies to de-escalate potentially violent situations (i.e., “stress reduction/relaxation activities,” “focus on positive thoughts,” “call someone who is supportive to talk it out,” “validate partner’s feelings,” and “physical exercise”). Items were summed to create a total safety planning score (pre-treatment α = .72, n = 134; post-treatment α = .72, n = 125).

Anger Management (Pre- and Post-Treatment)

Individuals who commit violent acts typically display more anger when discussing areas of disagreement than do non-violent individuals (Jacobson et al. 1994). To measure the extent to which individuals were both experiencing and controlling their emotions and anger-related behaviors, we used a shortened version of the State Trait Anger Expression Inventory (STAXI; Spielberger et al. 1988). Participants were asked to report how frequently they acted on the emotions they experience on a scale ranging from 1 (never) to 7 (always). We were interested in eight items that measure anger management (e.g., “I control my temper.”) in order to assess change in the ability to inhibit anger expression from pre- to post-treatment. These items were summed to create a score for anger management (pre-treatment α = .86, n = 132; post-treatment α = .84, n = 122).

Controlling Behaviors (Pre- and Post-Treatment)

Violence can reflect an attempt to control another individual (i.e., intimate terrorism; Johnson 1995). We used Stets’ (1993) measure of control in romantic relationships in order to assess controlling behaviors. Participants were asked to rate, on a frequency scale ranging from 1 (never) to 5 (very often), how often they used 10 controlling behaviors (e.g., “I regulate who s/he sees.”). The 10 items were summed to create a measure of controlling behaviors in the relationship (pre-treatment α = .85, n = 134; post-treatment α = .88, n = 125).

Violence (Pre- and Post-Treatment)

In order to assess violence, we used the physical and psychological violence subscales of the Revised Conflict Tactics Scale (CTS-2; Straus et al. 1996), the most widely used self-report measure of violent behavior. Participants reported the frequency with which they engaged in each of 12 physically violent and eight psychologically violent behaviors during a typical month. If the behavior never occurred participants would write a “0”; if the behavior occurred once in a typical month, participants would write a “1”; if the behavior occurred twice in a typical month, participants would write a “2”; etc. Participants were also able to indicate if the behavior had ever occurred, but occurred less frequently than once in a typical month by marking the item with an “X.” (As noted above, at the start of the post-treatment questionnaire there were instructions asking participants to answer the questions thinking about their attitudes and behaviors “TODAY,” (emphasis in instructions) after completion of the intervention program; as such, at post-treatment, participants were reporting about a typical month upon completion of treatment requirements for 21–30 weeks). Participants who reported having engaged in each violent behavior, but doing so less than once a month (i.e., marked an “X”) were assigned a frequency score of 0. Monthly frequencies were summed to create scales of both physical violence (e.g., “pushing or shoving my partner”; pre-treatment α = .67, n = 129; post-treatment α = .79, n = 123); and psychological violence (e.g., “shouting or yelling at my partner”; pre-treatment α = .81, n = 129; post-treatment α = .75, n = 119). In short, these scores represent the number of times participants had committed violent acts in a typical month.

Social Desirability (Pre- and Post-Treatment)

Although we went to great lengths to assure participants that the RCIP staff would not see respondents’ data, individuals who commit domestic violence offenses often underreport abusive behaviors (e.g., Browning and Dutton 1986). To assess and control for socially desirable responding (e.g., Craig et al. 2006), we used a shortened version of the Marlowe Crowne Social Desirability Scale (MCSDS; Crowne and Marlowe 1960). This measure consists of 14 items, half of which ask whether or not the participant behaved in ways that were culturally accepted, but likely untrue (e.g., “Before voting, I thoroughly investigate the qualifications of all the candidates.”) and the other half of which were likely true, but not socially acceptable (e.g., “I like to gossip at times.”). If a person responded ‘false’ to an item that is likely true, or ‘true’ to an item that is likely false, they were considered to be responding in a socially desirable, but dishonest, way to that item. Each socially desirable response was given a score of 1 and then summed to create a total score of social desirability (pre-treatment α = .65, n = 131; post-treatment α = .83, n = 128).

Results

Descriptive statistics and correlations for all study variables at pre-treatment assessment are displayed in Table 1 and at post-treatment assessment in Table 2. Desirable responding was associated with all variables of interest at pre-treatment, with the exception of desire for change. The pattern was similar for post-treatment assessments. Further, as predicted by prior research, all variables of interest, with the exception of controlling behaviors, were correlated with either psychological or physical violence (or both) at the pre-treatment assessment. At post-treatment, desire for change and accountability were the only two measures that were not associated with psychological or physical violence. It should also be noted that physical and psychological violence were highly correlated with each other at both pre- and post-treatment assessments.

Table 1 Correlations between study variables pre-treatment (N = 149)
Full size table
Table 2 Correlations between study variables post-treatment (N = 149)
Full size table

Missing Data

In order to maximize statistical power to detect changes in our key constructs over time, and in light of known tendencies for individuals to underreport extreme forms of violence (Emery 2010), we employed multiple imputation analyses in SPSS to adjust for the nontrivial (but typical in studies of violence; Babcock et al. 2004) amount of missing data in the current study (Bodner 2008; Rubin 1987). Specifically, we generated 30 imputed datasets using all study variables; we specified constraints for number generation based on scale maximums and minimums. Findings reported are averaged model estimates produced from these 30 imputed datasets. Importantly, multiple imputations have been show to successfully address issues of item non-response in samples of violent domestic partners (Emery 2010).

Analytic Technique

A series of repeated measures ANOVAS were used to determine pre-test to post-test differences in each of the variables of interest. Given that social desirability was significantly correlated with most measures of interest, participants’ centered Marlowe-Crowne score was entered as a covariate in all analyses to control for socially desirable responding (Craig et al. 2006). We tested for gender interactions for each model and none were found. Thus, the gender interaction term was dropped from all final models reported below. Effect sizes were calculated using Cohen’s d. To protect against inflated effect sizes due to the inherent correlation between pre- and post-test measures, we used standard deviations as the metric of pooled variance in our calculations (Dunlop et al. 1996). Again, all findings reported reflect average estimates across 30 imputed datasets.

Findings

The results of the repeated measures ANOVAs yielded a main effect for time (i.e., pre- to post-treatment) in a number of behaviors and cognitions. There was a significant increase in taking accountability for violent behaviors from pre- to post-treatment (Ms (SDs) = 29.11 (8.61) and 37.51 (9.33), respectively; F (1147) = 76.00, p < .001; d = .94). There was not a statistically significant reduction in clients’ use of controlling behaviors from pre- to post- treatment (Ms (SDs) = 17.46 (6.72) and 16.69 (6.24), respectively; F (1147) = 1.27, p = .32; d = .12). There was, however, a significant improvement in clients’ intentions to use a variety of safety-planning strategies (Ms (SDs) = 6.14 (1.98) and 6.99 (1.92), respectively; F (1147) = 18.64, p < .001; d = .44). Clients were also significantly more likely to manage their anger from pre- to post- treatment (Ms (SDs) = 40.37 (8.27) and 42.31 (7.55), respectively; F (1147) = 8.63, p = .01; d = .25). There was also a significant decrease in participants’ perceived stress levels from pre- to post- treatment (Ms (SDs) = 26.39 (6.80) and 24.56 (6.80), respectively; F (1147) = 10.21, p < .01; d = .27). Finally, clients’ desire for change in their violent behaviors significantly improved from pre- to post-treatment (Ms (SDs) = 33.25 (8.35) and 35.26 (7.65), respectively; F (1147) = 9.58, p < .01; d = .25).

Completion of the RCIP was not only associated with a number of positive outcomes linked to attenuated violent behavior, but also with reduced levels of violent behaviors. Specifically, clients’ reports of engaging in psychological violence decreased from pre- to post-treatment (Ms (SDs) = 10.23 (13.32) and 7.79 (10.90), respectively; F (1147) = 5.01, p = .04; d = .20). Clients’ reports of engaging in physical violence also significantly decreased from pre- to post-treatment (Ms (SDs) = 3.41 (8.05) and 1.32 (3.08) respectively; F (1147) = 9.11, p < .01; d = .34).Footnote 4

Discussion

We used a pre-/post-treatment design to investigate the effectiveness of a counseling-based, psychoeducational Batterer Intervention Program (BIP). In contrast to traditional Duluth-Model BIPs, the intervention program that we evaluated employed non-punitive, therapeutic principles to teach non-violent relationship skills to both male and female domestic violence offenders in same-gendered groups. We hypothesized that this program would successfully decrease future violent acts and improve psychological correlates of nonviolent behavior because it teaches offenders to take accountability for their violent acts, to explore alternative safety plans, to process histories of violence in their families of origin, and to effectively manage anger and stress. A sample of individuals who attended a Resolution Counseling Intervention Program (RCIP) completed surveys prior to and after completing the BIP; these surveys included measures of physical and psychological violence as well as measures of controlling behaviors, accountability for violent behavior, anger management, perceived stress, safety planning, and desire for change. Our results generally supported our hypotheses; the program reduced self-reported psychologically and physically violent behaviors and created positive changes in associated constructs.

Unlike more traditional Duluth-Model BIPs, RCIPs do not employ a punitive approach; yet, this non-punitive approach facilitates offenders taking accountability for their violent behaviors. When clients were treated as equals in the therapeutic process, participants did take significantly more accountability for their violent actions by the conclusion of the treatment program. Interestingly, there was not a statistically significant reduction in clients’ self-reported use of controlling behaviors. Our findings provide preliminary evidence that taking accountability, however, a significant correlate of violent behavior (Johnson 1995), decreases even when using a ‘softer’ approach to intervention. Future research should attempt to replicate this finding with a larger sample within a similar therapeutic treatment program.

Clients of the RCIP also showed significant improvements in key skills known to be associated with reductions in violence. Specifically, RCIP participants used a greater variety of de-escalation strategies when situations that could potentially become violent occurred. RCIP clients were also better able to manage their anger. These anger management techniques are particularly useful for individuals who resort to violence as a means to resolve a situation because their anger has gotten out of control (i.e., situational violence; Pence 2002). Finally, participation in the RCIP was associated with lower levels of perceived stress at the conclusion of the program. Notably, safety planning, anger management, and stress management skills are typically ignored in Duluth-Model BIPs. RCIP advocates argue that fostering these skills should increase clients’ feelings of self-efficacy when managing potentially violent situations. In that sense, teaching these skills positively empowers clients to create change in their lives in contrast to the negative empowerment that results from shaming offenders.

It should also be noted that RCIP participants showed an increase in their desire to change their violent behaviors. Desire for change is associated with successful program treatment, and thus, is often used as a screening criterion for whether or not a person would benefit from attending a BIP (Williamson et al. 2003). However, our research demonstrates that desire for change is not a static construct, but instead improves over the course of effective treatment. Batterers’ level of desire for change is associated with positive strides in fostering empathy and communication as well as in reducing abusive behavior (Scott and Wolfe 2003). As such, enhancing a desire for change over the course of a BIP is one avenue by which to effectively implement violence prevention strategies and skills. Although the goals of Duluth-Model BIPs are to facilitate recognition in offenders of power and violence problems, the positive impact on desire for change in RCIP participants demonstrates that counseling-based programs also successfully achieve these aims.

Finally, reports of psychological and physical violence decreased over the course of the RCIP. The changes observed in the current study, however, were small in magnitude. In other words, and consistent with a meta-analysis of similar treatment programs, it is likely that the RCIP only minimally reduced clients’ recidivism above the impact of involvement with the criminal justice system (Babcock et al. 2004). Still, the magnitude of the changes in physical violence (d = .34) and psychological violence (d = .20) observed in the current study are comparable to those found in other intervention program evaluations (average Cohen’s d values range from .03 to .35; Babcock et al. 2004). Further, any decrease in violent behaviors in domestic offenders is meaningful.

Limitations

The currents study’s findings and implications must be considered in light of its limitations and strengths. One notable limitation of the study is the reliance on self-reports from the violent offenders as opposed to partner reports or criminal records. This reliance on self-reports means that the measures of violence and other undesirable behaviors are susceptible to underreporting (Craig et al. 2006). Indeed, in our sample, the number of violent acts reported both pre- and post-treatment were low considering it was a presumably violent population. To address this limitation, we controlled for desirable responding in all of our analyses (e.g., Hanson and Brussière 1998). It is also important to note, however, that self-reports of violent behavior have statistically-equivalent predictive validity for future violent offenses as do more objective measures of violent risk factors (i.e., psychopathology batteries, risk appraisals, statistical correlates of recidivism; Kroner and Loza 2001). Further, desirable responding is likely most impactful when clients enter the program because self-disclosure is associated with the length of time spent in therapy and with the development of a therapeutic relationship (Farber 2003). Intuitively, it seems clients’ concern for hiding information would be heightened prior to building a trusting therapeutic relationship with the RCIP staff. As such, desirable responding should have hindered our ability to detect changes from pre- to post-treatment, making the results reported a conservative test of the program’s effectiveness.

Another limitation of the study design is the lack of a control group or alternative treatment group with which to compare the RCIP program. Although unlikely, particularly for female clients, it could be that a program using the Duluth Model would have found similar results for reduced violence and its related constructs. Therefore, we cannot conclude whether RCIPs are more effective than punitive programs or than no program at all; rather, we can conclude that the RCIP, and its specific techniques, was successful at reducing violence as well as behaviors and attitudes associated with violence, including those specifically targeted in Duluth-Model BIPs (e.g., accountability).

It is also a limitation that, because the data were collected by program staff (and not the researchers), we do not have records of how many participants were approached to complete the pre-treatment survey, how many pre-treatment survey participants dropped out of the program, or how many pre-treatment participants refused to participate in the post-treatment assessment. Although we recognize this as a limitation, we felt that having program staff (with whom clients had established relationships and trust) made participants more comfortable participating and providing honest responses. In our opinion, this strength outweighed the cost of the control we lost during data collection.

Finally, there was a significant drop in participation from pre-treatment (n = 414) to post-treatment (n = 149). The decrease in the number of surveys collected could have occurred for several reasons, including offenders dropping out of the program or choosing not to complete the post-treatment survey. It is important to note that of the participants who did not complete the program, some were no longer required to do so due to reversals of or alterations to their court sentences. Indeed, about 10 % of clients who go through this RCIP undergo changes in their sentences after being assigned to complete the treatment program. Importantly, this reduction in participants is not unusual for domestic violence studies (Babcock et al. 2004). Further, participants who did versus did not complete post-treatment measures did not differ in their use of physical violence, psychological violence, or in their desire to change prior to treatment.

The limitations of this study, however, are counterbalanced by a number of strengths, which have been elaborated on previously. Primarily, the study sample was not limited to men, but rather included both male and female domestic violence offenders. Unlike many studies of BIPs that only use a male sample (typically because of programs’ all-male populations; e.g., Eckhardt et al. 2008), we can generalize the findings of program effectiveness to both genders. It should be noted that we tested for gender interactions for each construct. No interactions with gender were found, suggesting that the RCIP curriculum may be equally successful for both men and women.

Future Directions

Although this study was an important first step in determining the effectiveness of a counseling-based, psychoeducational BIP, several additional research questions still need to be addressed. Specifically, future research should determine for which typology of domestic violent offenders non-punitive programs are most effective. There is a consensus that individuals resort to violence for different reasons (e.g., Holtzworth-Munroe and Stuart 1994; Johnson 1995). For example, some individuals lack the skills necessary to properly manage anger, resulting in situations that escalate (i.e., situational violence; Johnson 1995). Others may have a personality disorder that leads to violent behavior (Holtzworth-Munroe and Stuart 1994). Still, there are some who, consistent with the Duluth Model, use violence as a means of exerting power or control over other individuals (i.e., intimate terrorists; Johnson 1995). Although it is possible that RCIPs’ reliance on therapeutic techniques allows their counseling professionals to cater to clients’ ‘typologies’ and equally benefit individuals who resort to violence for different reasons, this idea should be tested by future research.

Conclusion

Given that courts often mandate domestically violent offenders to attend BIPs, evaluating the effectiveness of programs’ diverse treatment strategies is essential. Our findings suggest that counseling-based skill training decreases the use of psychological and physical violence and improves behaviors and attitudes related to reductions in violent behavior (e.g., increasing accountability and desire for change, and improving skills associated with nonviolence). Thus, and although more research is certainly needed, our results highlight the value of BIPs recognizing the multi-faceted causes of violence and teaching individuals, both men and women, the necessary skills to succeed in relationships.