On Behalf of Women Offenders

Women's Place in the Science of Evidence-Based Practice

The argument on behalf of women offenders was made at least four decades ago at the National Conference on Corrections convened by then President Richard Nixon in response to the 1971 Attica Prison riots. Among many speakers, Dr. Edith Flynn delivered the only address on women offenders. In “The Special Problems of Female Prisoners,” Dr. Flynn noted that female prisoners were largely ignored. She supported her assertion with reference to the then-recent President's Commission on Law Enforcement and Administration of Justice (1967), noting that not a “single paragraph or statistic on the female offender could be found in any of the material” (Flynn, 1971:113). She asserted further that the prevailing theories of criminal behavior were inapplicable to women and that the resulting lack of information had adverse implications for managing and treating female offenders. In the intervening years, research has put forward a clearer picture of how women become involved in the justice system and what their treatment needs are when they get there. However, there is clear reason to lament the arduously slow pace in which emerging evidence is impacting policies and front-line practices and services for women (Belknap, 2007; Belknap and Holsinger, 2006; Blanchette and Brown, 2006; Bloom, Owen, and Covington, 2003; Chesney-Lind, 2000; Holtfreter, Reisig, and Morash, 2004; Messina, Grella, Cartier, and Torres, 2010; Reisig, Holtfreter, and Morash, 2006; VanDieten, 2011; VanVoorhis, 2009).

Attempts to fill the knowledge gap observed by Dr. Flynn were addressed initially by surveys of correctional programs (Glick and Neto, 1977) and of women offenders (U.S. Government and Accounting Office [GAO], 1979).Several classic qualitative studies followed over the ensuing decades (Arnold, 1990; Bloom, 1996; Chesney-Lind and Rodrigues, 1983; Chesney-Lind and Shelden, 1992; Daly, 1992, 1994; Gilfus, 1992; Holsinger, 2000; Owen, 1998; Richie, 1996; Smart, 1976). Over time, these studies portrayed different pictures of women's and men's entry (pathways) to crime, one that, for women, implicated abuse and trauma, poverty, unhealthy relationships, mental illness, substance abuse, and parental concerns.

Little attention was devoted to showing how these differences might impact correctional programs and services for women. An earlier report by Barbara Bloom and James Austin sought to highlight innovative strategies and programs (Austin, Bloom, and Donahue, 1992), but even by the end of that decade, many feminist scholars simply observed that there was an appalling lack of research on which to build correctional approaches for women (Chesney-Lind, 1997, 2000; Girls Incorporated, 1996; Holtfreter et al., 2004; Morash, Bynum, and Koons, 1998; Van Voorhis and Presser, 2001).

At the same time, however, correctional treatment was coming back into favor after decades of policies favoring incapacitation and deterrence and that change was research driven (see Cullen, 2005). Even so, the research fueling the policy transition was conducted largely on boys and men. For example, two highly influential meta-analyses of correctional programs effectively proved to policy makers that certain types of treatment programs substantially reduced future offending. However, both concluded with warnings that women and girls were underrepresented in the research (Andrews, Zinger, et al., 1990; Lipsey, 1992).The meta-analysis conducted by Donald A. Andrews and his associates at Carlton University generated a series of “Principles of Effective Intervention”(see also Gendreau, 1996) that fueled the development of the now predominant correctional treatment paradigm, variously referred to as “the Canadian Model,” the Risk Needs Responsivity Model (RNR), the “What Works” Model, and the General Personality and Cognitive Social Learning Model (GPCSL).Through the remainder of this address, I will refer to this approach as the RNR model.

A parallel body of research developed classification and assessment systems on men. As with other research, the custody classification assessments used to assign inmates to prisons were validated initially on men or samples that made up such a small proportion of women that the resulting findings were attributable to men (Brennan, 1998; Morash et al., 1998; Van Voorhis and Presser, 2001). A national survey of state prison correctional classification directors found that 36 states had not validated their classification systems on female inmates but generalized findings to them, just the same (Van Voorhis and Presser, 2001). Of course, the flawed policy of generalizing male-based research to women resulted in erroneous assignments of women to custody levels. Specifically, these involved problems with overclassification or with assigning women to higher custody levels than warranted on the bases of their actual behavior. In fact, comparative studies found that maximum-custody women incurred serious misconducts at roughly the rate of medium-custody men. Moreover, women's aggressive behaviors while incarcerated occurred much less frequently than men's and involved much less serious forms of aggression (Hardyman and Van Voorhis, 2004).

A second type of correctional assessment, dynamic risk/needs assessments, were designed to classify community correctional offenders into low, medium, and high levels of community risk on the basis of needs known to predict future offending. Because the assessments identified an array of predictive needs, they also served as a valuable tool for triaging offenders into programs most likely to turn them away from lives of crime. The early construction validation studies for these assessments also were based largely on male offender samples (e.g., see Blanchette and Brown, 2006; Brennan, 1998; Holtfreter et al., 2004; Van Voorhis, Wright, Salisbury, and Bauman, 2010) and validated on women much later than their initial construction (e.g., see Andrews, Dowden, and Rettinger, 2001; Lowenkamp, Holsinger, and Latessa, 2001; Manchak, Skeem, Douglas, and Siranosian, 2009; Smith, Cullen, and Latessa, 2009). For the most part, the revalidation studies found these assessments to be valid for women. However, by the time researchers addressed this problem with external validity, it was too late to include the needs that gender-responsive scholars found most relevant to women offenders. Thus, the programs were not targeted to many problems that brought women into crime (Belknap and Holsinger, 2006; Bloom et al., 2003; Hannah-Moffat, 2009; Van Voorhis et al., 2010). With no assessments to identify these problems, women were less likely to be triaged to gender-specific services such as protection from abusive partners, childcare services, access to reliable transportation, low self-efficacy, trauma and abuse, parenting programs, healthy relationships, and realistic employment opportunities that allowed for self-support (Bloom et al., 2003).

These oversights were not lost on federal policy makers and practitioners. Several initiatives, mostly in the area of disseminating research and innovative ideas, encouraged awareness of women offenders (see Buell, Modley, and Van Voorhis, 2011).National conferences on the subject began in 1985 with The First National Adult and Juvenile Female Offender Conference. The National Institute of Corrections (NIC) was especially instrumental in providing publications and technical assistance, beginning with the 1993 publication of A Guide to Programming for Women in Prison (Education Development Center, Inc., 1993), the creation of a curriculum in Sentencing Women Offenders: A Training Curriculum for Judges (Cicero and DeCostanzo, 2000), and the development of the Federal Center for Children of Prisoners. In 1994, the U.S. Congress passed the Violence Against Women Act, establishing the Office on Violence Against Women within the U.S. Department of Justice. Relevant federal responses also have occurred under the rubric of the 2003 Prison Rape Elimination Act. Later work funded by the NIC produced an award-winning publication Gender-Responsive Strategies: Research, Practice, and Guiding Principles for Women Offenders (Bloom et al., 2003) along with several tools to assist agencies’ efforts to design programs and services that are more amenable to women. These programs included the Women Risk/Needs Assessment (uc.edu/womenoffenders), the Women Offender Case Management Model (Van Dieten, 2008), and the Gender-Informed Practice Assessment (GIPA) (Center for Effective Public Policy, 2010). In 2010, the Bureau of Justice Assistance, in partnership with NIC, established the National Resource Center on Justice Involved Women, “to provide guidance and support to criminal justice professionals and to promote evidence-based, gender-responsive policies and practices” (cjinvolvedwomen.org).¹

I was privileged to work on several of these projects along with teams of extremely talented and committed graduate students, government officials, scholars, administrators, practitioners, and activists. Beginning in the late 1990s, the University of Cincinnati secured a cooperative agreement with NIC to construct a public domain women's risk/needs assessment (WRNA). Along with the research, my staff and I operated in the role of embedded researchers (Petersilia, 2008), as research partners with the agencies participating in our research and adopting the assessments.

Having spent most of my career studying male offenders, including a good deal of research in the RNR approach, I had not until then directly experienced how difficult it could be to study women. These challenges were apparent on a political and a professional level, impacting me as a scholar, consultant, advisor, and teacher. I can state unequivocally that it is professionally and financially much more difficult to study women than men. I experienced these difficulties firsthand and believe that they have been documented accurately by others (see Chesney-Lind, 2012).However, I was least prepared for my first-hand introduction to the abysmal state of science as it accounts for, or more accurately, fails to account for, the lives of women. On reflection, how that science has unfolded in a culture where “male is norm” (Tavris, 1992) was discouraging to observe, and the costs of the “male is norm” scientific model are substantial. Thus, I also was embedded in the process of an emerging body of evidence that ran contrary to the prevailing evidence of the day. There is a story to that, and I believe that it is important to tell it.

If I may be permitted a moment of personification, social science has not been an innocent party to the slow pace at which the needed policies and practices for women offenders unfolded. After 12 years of studying women offenders, I now find myself with a less than sanguine view of science than I had at the outset. There are many social and political challenges to developing policy and programs relevant to women, but I wish to discuss the challenges that science, of all things, imposed on this task. As will be observed, several of these scientific issues were not unique to corrections but reflected the scientific culture of our times. Other challenges emerged from the recent science of correctional treatment itself. In the pages that follow, I discuss the challenges impacting the gender-responsive movement in corrections. I will, however, conclude on a more optimistic note with an overview of the emerging evidence, a body of research that, although still not as plentiful as that regarding men, is nevertheless achieving consistency across studies and showing a promising path to improving approaches for women (VanDieten, 2011).

It is now almost 40 years since Professor Flynn (1971) reminded the National Conference on Corrections that the field had produced no research on women offenders and that, as a consequence, women served by the male model of corrections were not receiving appropriate programs and services. The science needed to correct this situation emerged too slowly. Moreover, new evidence-based treatment models for women are even now mostly in a dissemination stage, far from full implementation. True, some correctional and pretrial agencies adopted evidence-based, gender-responsive assessments and programs, but many of these efforts have experienced fit-full starts and stops. I continue to agree with Professor Flynn. As an overview, science was a factor in the following key ways:

• 1
  As far as women and minorities are concerned, many endeavors of science, including medicine, education, and mental health, fall far short of formulating scientifically representative samples. Many such studies then develop conclusions that generalize findings inappropriately to women and minorities. As far as women and minorities are concerned, scientific problems with external validity (a concept taught early in most research methods courses) are pervasive.
• 2
  The recent policy mandates for evidence-based practice and the commensurate elevation of meta-analysis as the “gold standard” have had the effect of blaming women for their invisibility. The perceived failure to produce the multitude of studies needed to support a meta-analysis of interventions for women offenders runs the strong risk of stifling innovation and causing some to downplay the emerging evidence on women that is available.
• 3
  After surviving many years of its own struggles to achieve legitimacy as a correctional policy, some proponents of the popular Canadian RNR treatment paradigm were resistant to evidence suggesting that their treatment targets and modalities may need to be expanded and modified to accommodate women offenders better. One would think that the RNR model, like other scientific products, would evolve as science does; however, women were the subject of the emerging science, and the science involving women advances very slowly.
• 4
  When all is said and done, there is an emerging body of evidence on women offenders. This literature, although not sufficient in numbers to support meta-analytic study, is remarkably consistent across studies and linked to favorable outcomes for women. Taken as a whole, the emerging science also forms a coherent model for women offenders that departs from some but not all of the principles underlying the RNR model.

But reaching the current stage of progress (number 4 in the preceding list) required that arguments “on behalf of women offenders” sustain several identifiable “scientific” challenges.

First, the problem observed by Professor Flynn four decades ago (1971) was not unique to corrections, but it was embedded in the wider scientific culture, impacting women in the general population as well as those encountering the criminal justice system. Sadly, inattention to women was apparent in medical trials, validations of educational exams used to determine college entrance and receipt of scholarships, and validations of mental health assessments, to name a few.

The historical exclusion of women from vital clinical trials ultimately led to the National Institute of Health Revitalization Act of 1993, which required the inclusion of women and members of minority groups in all National Institutes of Health (NIH)-supported biomedical and behavioral research except in instances where a clear and compelling reason was established that to do so would be inappropriate (e.g., the study of a sex-specific illness).² The guidelines further stipulated that childbearing potential and the added cost of including women and minorities were no longer acceptable justifications for not including women in equal numbers to men in clinical trials. Up until that point, exclusion of women from medical research was, according to some, an unintended consequence of protecting vulnerable populations, including pregnant women, and premenopausal women who were capable of becoming pregnant (Goldenberg, 2003; Killien et al.,2000). For others, the exclusion was the outcome of a naive assumption that findings observed from studies on male subjects could be generalized to women without modification, a startling “leap of faith: in an otherwise rigorous research enterprise” (NIH, 1999: 10, quoted in Bloom et al., 2003). So strong was the “male is norm” filter that it successfully trumped one of the core lessons in any graduate research methods class—external validity.

Notwithstanding the 1993 guidelines, which had no enforcement provisions, subsequent forums and publications demonstrated an ongoing failure to recruit sufficient numbers of women in clinical trials. Even fewer studies disaggregated findings by gender, where true gender-specific findings would be observed (Geller, Adams, and Carnes, 2006; NIH, 1999; Ramasubbu, Gurm, and Litaker, 2001; Vidaver, Lafleur, Tong, Bradshaw, and Marts, 2000).³ Among the costs incurred by generalizing findings from male samples to females are(a) a mistaken understanding of the role of aspirin in preventing women's strokes and heart attacks (Ridker et al., 2005), (b) a limited understanding of heart disease in women (Chen, Woods, and Puntillo, 2005; Dey et al., 2008; Rathore, Wang, and Krumholtz, 2002), and (c) a host of issues with pharmaceutical dosages (Keiser, 2005; Vidaver et al., 2000).

Similar practices disparage early validations of college entrance examinations, including the Scholastic Aptitude Test (SAT), the National Merit Examination, and the Graduate Record Examination (GRE). Replicated studies conducted during the 1980s and 1990s consistently found that educational tests used for vital college entrance decisions performed differently for men and women, including on the National Merit Exam (National Association for College Admission Counseling [NACAC], 2008), the SAT (Bridgeman and Wendler, 1991; Clark and Gandy, 1984; Leonard and Jiang, 1999; Silverstein, 2000; Wainer and Steinberg, 1992), and the GRE (House, Gupta, and Xiao, 1997; Sternberg and Williams, 1997). The core problem was not that women scored lower than men on such examinations but that such tests tended to underpredict the ultimate performance of women and to overpredict the performance of men. In large competitive schools that placed primary reliance on the examination results, women were observed to have lower entrance rates than men (Leonard and Jiang, 1999).

Use of the disparate tests in awarding scholarships and National Merit Awards was particularly egregious and was found to have a discriminatory effect that in one case resulted in a change to state policy (Sharif v. New York State Education Department, 1989) and, in the case of the National Merit Exam, a large out-of-court settlement. Reportedly, the gender prediction gap on these exams was known to insiders for more than a quarter of a century (Leonard and Jiang, 1999) and finally led to the inclusion of a writing sample in 2005 that presumably improved the prediction for female students, but not everyone is convinced (NACAC, 2008).

I first learned of the external validity problems associated with some cognitive, personality, and mental health assessments from Carol Gilligan. I had the good fortune to be sent to Harvard University by my dissertation advisor, Marguerite Warren, to learn how to classify probationers according to Lawrence Kohlberg's Stages of Moral Judgment (Kohlberg, Colby, Gibbs, Speicher-Dubin, and Candee, 1979). Gilligan, a faculty member, addressed my fellow workshop participants and me after a long day of lessons on the Moral Development scoring protocol. She explained to us and a group of Harvard researchers and instructors, who clearly were less than happy with her, that the Stages of Moral Judgment had been formulated on the study of the lives of boys and men and then erroneously generalized to girls and women. After the fact, females were assessed on the protocol, only to find that many clustered around stage 3 on the stage-based typology. Stage 3 is a stage reserved for humans who base moral decisions on a concern for reciprocity in close relationships. One could develop to higher stages of moral development, stages reserved for those who valued the importance of maintaining social systems or universal principles of moral action, but women seldom did. Gilligan later rectified the problem by studying samples of women and observing that “the stage 3 problem” was a function of the “male is norm” assumption and the failure to account for the fact that women are relational and factor relationships into most decision-making regardless of “maturity” (Gilligan, 1982; Taylor, Gilligan, and Sullivan, 1995).

Strong professional guidelines recommend the use of mental health assessments only on populations “whose validity and reliability has been established for use with members of the population tested” (American Psychological Association [APA], 2010). However, one can now deviate from these in cases where the author expresses appropriate reservations. Concerns have been raised for the Mf (Masculinity-Femininity) scale of the Minnesota Multiphasic Personality Inventory-2 (Lewin and Wild, 1991), the Psychopathy Checklist-Revised (Baker and Mason, 2010), tests of worker satisfaction (Hesse-Biber, Nagy, and Yaiser, 2004), and intelligence tests (Hyde, 1990). The absence of females from psychological research is similar to that observed in medicine, education, and criminal justice with concerns raised for psychotherapy in general (APA Divisions 17 and 35, 2004; Levrant and Silverstein, 2005) as well as for specific specialty areas such as school psychology (Holverstott et al., 2002), mental retardation (Porter, Christian, and Poling, 2003), psychopharmacology (Poling et al., 2009), and organizational psychology (Jarema, Snycerski, Bagge, Austin, and Poling, 1999).

In sum, women's issues do not become the focus of policy and innovation because the science that would foster such change devotes limited attention to them, and what is not observed is not attended to. This obvious knowledge gap underscores the poignant titles chosen for some recent scholarship, for example The Mismeasure of Woman (Tavris, 1992), The Invisible Woman (Belknap, 2007), and Half the Human Experience (Hyde, 2006).

The second challenge occurred within the past decade when public-sector funding placed a premium on those practices and policies that showed evidence of achieving effective outcomes. The evidence-based practice mantra refers to the use of research and science, particularly controlled studies, to identify the best practices in a field. It has been voiced by policy makers ranging from agency heads to presidents of the United States. However, the evidence-based mandate places women and minorities, who have been understudied, at a distinct disadvantage.

The movement to evidence-based practice began in medicine in the early 1990s and then moved to other fields such as psychotherapy (Task Force on Promotion and Dissemination of Psychological Procedures, 1995) and more slowly to corrections (Cullen and Gendreau, 2001; MacKenzie, 2000). It forms the foundation for many public, performance-based budgeting systems, holds a prominent place in the recent health care law, and factors heavily into funding of social policy and research. Evidence-based practice also is the basis for recent federally sponsored Web sites designed to help correctional professionals select programs that show evidence of decreasing offender recidivism, including Blueprints for Violence Prevention (colorado.edu/cspv), the Office of Juvenile Justice and Delinquency Prevention Model Programs Guide (ojjdp.gov), and the National Registry of Evidence-based Programs and Practices (nrepp.samsha.gov). Most recently, The Office of Justice Programs opened a similar site (crimesolutions.gov). Many individuals, myself included, believed that the evidence-based mandate was past due, especially in the field of corrections with its less-than-illustrious tradition of eclectic and creative interventions that could not possibly have produced favorable outcomes, for example cake decorating, drums and candles, horseback riding, wagon trains, and plastic surgery (Van Voorhis, Cullen, and Applegate, 1995).

The keys to the “evidence-based” movement in corrections are several influential meta-analyses, a methodologically rigorous strategy for synthesizing findings across numerous controlled studies (Glass, McGaw, and Smith, 1981). Such studies produce “effect sizes” for each modality studied, and the “effect size” statistic was noted to produce far more stable findings than former methods of synthesizing research (e.g., votecounting).

Several meta-analyses of correctional treatment programs were conducted during the 1990s, but two were especially noteworthy. One was a study of 154 evaluations of correctional programs (Andrews et al., 1990b) that generated a series of guidelines referred to as the Principles of Effective Intervention (see also Andrews, Dowden, and Gendreau, 1999; Gendreau, 1996). The second reviewed 443 delinquency prevention and intervention programs (Lipsey, 1992). Both showed policy makers that rehabilitation models substantially reduced future offending. Other meta-analyses established treatment-relevant predictors of recidivism (Andrews, Bonta, et al., 1990; Gendreau, Little, and Goggin, 1996). Meta-analyses also convincingly countered naïve assumptions that the crime problem could be solved by such approaches as boot camps (MacKenzie, Wilson, and Kider, 2001) or other “punishing smarter” ideas (Andrews et al., 1990b; Gendreau et al., 1996; Gendreau, Goggin, and Cullen, 1999; Langen and Levin, 2002).

Even so, the power of evidence, especially evidence put forward by the meta-analyses, looped around full circle to fault the gender-responsive movement for the invisibility of women in key policy and programmatic research. Evidence came to drive policy, but for women, there was no evidence; as noted, the invisibility of women in key research was pretty much a fact of science. Indeed, only 2.4% of the experimental studies examined in Mark Lipsey's (1992) meta-analysis sampled only girls, and 5.9% sampled primarily girls. The meta-analysis conducted by Andrews and his associates (1990) concluded with the admonition that gender effects required a more detailed analysis. Even, Lipsey's larger, most recent analysis reported that only 4.0% of the studies sampled mostly female studies versus 87.0% accounting for all male or mostly male samples (Lipsey, 2009: 132). The authors acknowledged their concerns for the limited research on women, but their findings nevertheless formed the foundations of today's approach to correctional treatment, treatment models that are offered to both males and females.

The founders of the meta-analysis technique warned of such problems when they noted that findings are highly dependent on the criteria for selecting studies from the total universe of available studies (i.e., selection bias) (Glass et al., 1981; Smith, 1980). Although the authors of the correctional meta-analyses certainly did not seem to commit selection bias, their results had the same effect because the requisite studies on women were not available. In a review of psychological, educational, and behavioral treatments, Lipsey and Wilson (1993: 1200) presented the problem in thoughtful terms:

Meta analysis is only possible for treatment approaches that have generated a corpus of research sufficient in quantity and comparability for systematic analysis within a statistical framework. Such a body of studies, in turn, is only likely to be produced for widely used and well-developed approaches growing out of established theory or practice, or for promising innovations. Thus the treatment approaches represented in meta analysis and reviewed in this article represent rather mature instances that are sufficiently well developed and credible to attract practitioners and sufficiently promising (or controversial) to attract a critical mass of research.

Simply put, meta-analysis and evidence-based practice is not the friend of underrepresented groups attempting to secure knowledge of optimal medical, therapeutic, or other treatments (Sue and Zane, 2005), and it should not purport to be. For their part, the Canadian authors of the RNR model sought to rectify the under-representation of women by conducting meta-analyses on necessarily smaller programmatic databases of women offenders (e.g., Dowden and Andrews, 1999). Validations of the risk/needs assessment accompanying the RNR model, the Level of Service Inventory (Andrews and Bonta, 1995), were conducted on samples of women offenders, and the sample sizes of these studies increased over time (e.g., Andrews and Bonta, 1995; Andrews, Bonta, and Wormith, 2004; Coulson, Ilacqua, Nutbrown, Giulekas, and Cudjoe, 1996; Lowenkamp et al., 2001; McConnell, 1996; Rettinger, 1998; Simourd and Andrews, 1994; Smith et al., 2009). Evidence, in the case of these studies, conformed to a pattern of repeated tests of the RNR programs and assessments and to proud assertions that the favorable findings refuted critics of the RNR model, including feminist scholars and other proponents of alternative gender-responsive approaches (see Andrews and Bonta, 2010). The studies that supported RNR for girls and women did not test the gender-responsive models. As such, there was no basis for any conclusions that gender-responsive approaches were flawed. Only two of these authors (see Blanchette and Brown, 2006; Smith et al., 2009) acknowledged the logical error of refuting gender-responsive proponents without testing directly the gender-responsive treatment targets and programs.

Not everyone would say there is anything wrong with this state of science. For example, in response to the well-established ethnic disparities in mental health research, the U.S. Surgeon General (2001) issued the guideline that minority mental health clients should be given treatments supported by the “best available evidence.” A similar argument has been made on behalf of delinquent girls, as when Hubbard and Matthews (2008) thoroughly reviewed the impressive empirical support for the RNR model along with more limited research relevant to the arguments put forward by gender-responsive scholars. They noted many ways in which the RNR model could be modified for girls and struggled to find areas of consensus between the two approaches. But at several points, particularly with regard to risk/need factors and assessments, the authors lamented the paucity of research on the recommendations of gender-responsive scholars and somewhat regretfully recommended approaches that were well within the confines of the RNR model.

Of course, if left on the table, “the best available evidence” argument minimizes the urgency to conduct more appropriate research and risks inattention to emerging research. Moreover, it is likely the case that the “best available evidence” is not a picture of the assessment and treatment models we would have if we had started with girls and women. Critics of the RNR approach note that, whereas it was evidence based, it was nevertheless formulated on the basis of research on male populations and only later found to be effective with women (Bloom et al., 2003). Several feminist critics faulted the over-reliance on the meta-analysis to the dismissal of qualitative studies that comprised most of the evidence supporting gender-responsive approaches to corrections (see Chesney-Lind, 2000; Hannah-Moffat and Shaw, 2000: Kendall, 2004). More scholars faulted the LSI-R for neglecting to include gender-specific factors (Blanchette and Brown, 2006; Funk, 1999; Holtfreter and Morash, 2003; Reisig et al., 2006; Van Voorhis et al., 2010).The consistent response of a least two Canadian authors underscores the point of this section (Andrews and Bonta, 2010: 514):

With all due respect, it is time for those who feel they are entitled to offer programs inconsistent with GPCSL and RNR perspectives to show some social responsibility. They must begin to program and evaluate in a “smarter” manner. To our knowledge, the evidence base in support of their approaches flirts with nil.

Third, on the strength of the evidence and with a good deal of dissemination and technical support from the NIC, the RNR model has been implemented widely and represents considerable investment on the part of adopting jurisdictions. In some circles, the RNR model is used synonymously with “evidence-based practice.” Implementation of the RNR model incurred many political struggles in its own right and was in many ways a dramatic improvement over previous approaches. However, with strong roots in the notion of a general theory of crime, the proponents of RNR resisted suggestions that the model could be improved for women. Thus, although improvements, modifications, and progress, would seem to be inevitable to science and social policy, the authors and proponents of RNR resisted suggestions for change and evidence supportive of change. And they, after all, seemed to be in the driver's seat.

The seminal meta-analysis (Andrews et al., 1990b; Lipsey, 1992) showed that only appropriate treatments could reduce recidivism by as much as 30%. The characteristics of an appropriate program were outlined in the Principles of Effective Intervention. Accordingly, appropriate programs (a) targeted intensive services to high-risk offenders (the risk principle); (b) targeted programs to needs (risk factors) related to future offending;(c) were consistent with cognitive-behavioral, social learning, and radical behavioral approaches to treatment (general responsivity principle); and (d) were responsive to personal and interpersonal considerations that posed barriers to successful treatment in appropriate programs (e.g., learning styles, intelligence, motivation, gender, ethnicity, and personality).

In time, the RNR paradigm drew in others who wrote consistent cognitive behavioral and social learning curricula (e.g., see Bush and Bilodeau, 1993; Bush, Taymans, and Glick, 1998; Goldstein, Glick, and Gibbs, 1998; Ross and Fabiano, 1985; Taymans and Parese, 1998) and alternative dynamic risk-needs assessments such as the Northpointe COMPAS (Brennan, Dieterich, and Oliver, 2006), and later the Ohio Risk Assessment System (Latessa, Smith, Lemke, Makarios, and Lowenkamp, 2009). From my own experience, RNR was admirably clear and translated well into treatment protocols. It soon brought psychologically informed treatment modalities into correctional environments for the first time in decades.

However, the RNR model did not come on the scene without its detractors. Evidence notwithstanding, many still clung to antirehabilitation themes first prompted by Martinson (1974; Farabee, 2005; Gaes, Flanagan, Motiuk, and Stewart, 1999; Logan and Gaes, 1993). The main proponents of RNR were frequently answering their detractors (Andrews and Bonta, 2010; Cullen, Smith, Lowenkamp, and Latessa, 2009; Gendreau et al., 2002b; Latessa, Cullen, and Gendreau, 2002) and occasionally expressed impatience with the number of times key findings had to be published and republished (see Andrews and Bonta, 2010). Additionally, high-quality implementation and program fidelity proved to be extraordinarily difficult (Bonta, Bogue, Crowley, and Motiuk, 2001; Gendreau, Goggin, and Smith, 2001; Latessa et al., 2002; Taxman and Bouffard, 2003) and clearly attenuated program outcomes when it was not present (Lipsey, 2009; Lowenkamp, 2004; Lowenkamp and Latessa, 2002; Lowenkamp, Latessa, and Smith, 2006).⁴ That there was ongoing concern for whether the RNR model would become widely implemented and have “staying power” was perhaps expressed most passionately in one of the last publications Don Andrews (Andrews and Bonta, 2010: 506) participated in prior to his death:

[M]any agencies are struggling with the implementation of RNR. And need help now or the movement is going to suffer and the community will be exposed to more crime. Quite frankly, for some of us, it is difficult to bear the thought of prevention and corrections returning to that “nothing works—we can't predict—we can't influence” position of anticipatory failure. Never again do we want perspectives on offenders that negate human diversity, dismiss human agency, and indeed destroy hope.

One does not have to read far in the Psychology of Conduct (Andrews and Bonta, 2010) to observe that the authors were especially exasperated by the feminist scholars who objected to the idea of developing an entire treatment paradigm on the basis of research men and boys (e.g., Belknap, 2007; Belknap and Holsinger, 2006; Bloom et al., 2003; Chesney-Lind, 2000; Reisig et al., 2006; Van Voorhis et al., 2010). Several feminist critics also rejected the notion of risk and risk/need assessment (Hannah-Moffat, 2004, 2009; Smart, 1982). As discussed, other authors noted that the gender-neutral risk needs assessments such as the LSI-R and the Northpointe COMPAS did not include the risk/need factors most appropriate to women offenders (Covington, 1998; Farr, 2000; Hardyman and Van Voorhis, 2004; Holtfreter and Morash, 2003; Morash et al., 1998; Van Voorhis and Presser, 2001; Van Voorhis, Salisbury, Wright, and Bauman, 2008; Van Voorhis et al., 2010). In response, the Canadian authors made it clear and provided evidence that the RNR model was consistent with general theories of crime and was applicable across types of crimes, offenders, and social conditions; gender was not relevant to its effectiveness (Andrews and Bonta, 1995, 2010; Andrews et al., 1990; Bonta, 1995; Dowden and Andrews, 1999).

In time, support was shown for a hybrid model, one that would modify the RNR model to fit the needs of women offenders better. At a policy level, this approach also had strong backing from the NIC. For the proponents of the hybrid model, the needs principle came under a good deal of scrutiny. There was wide consensus that if the goal were to decrease recidivism, then the risk factors for recidivism should be targeted. However, scholars looking to modify the RNR model raised questions about what should be targeted (see Blanchette and Brown, 2006; Holsinger and Van Voorhis, 2005; Holtfreter and Morash, 2003; Reisig et al., 2006; Salisbury et al., 2009a; Van Voorhis et al., 2010; Wright, Salisbury, and Van Voorhis, 2007).

By following two key meta-analyses (Andrews, Bonta, et al., 1990; Andrews, Zinger, et al., 1990), agencies steeped in the RNR tradition were encouraged to prioritize program resources to a limited but potent set of risk factors, referred to as the “Big 4” (criminal history, criminal thinking, personality attributes, and criminal peers) or the Central 8 (the Big 4 plus family/marital, education/employment, substance abuse, and leisure/recreation).All were key factors on the LSI-R (Andrews and Bonta, 1995) and later the LS/CMI (Andrews et al., 2004) and referred to as “criminogenic needs.” RNR proponents also faulted forensic psychology and other models of correctional interventions for undue attention to mental illness, self-esteem, and poverty (Andrews and Bonta, 2010). In contrast, the gender-responsive hybrid model focused empirical research on risk/need factors suggested by feminist and pathways research (Gehring, 2011; Jones, 2011; Van Voorhis et al., 2010) and hypothesized that many of the traditional, gender-neutral risk factors for future offending would be shared by men and women but that another set of needs would be unique to women offenders. Still others might be predictive of offense-related outcomes for both women and men but work differently (e.g., antisocial associates) (see Salisbury and Van Voorhis, 2009).

In the United States, research on the utility of a hybrid gender-responsive classification and risk/needs assessments began in 2000 with a cooperative agreement awarded to the University of Cincinnati (UC) by the NIC. Not surprisingly, at the outset of this research, there was little in the way of relevant correctional research to guide the project. This was my first insider's view of the invisibility of women and the first time in my career that I had to design a study without what I viewed to be a solid quantitative knowledge base. Women were truly “correctional after thoughts” (Ross and Fabiano, 1985).

We relied heavily on the now-classic qualitative studies conducted prior to that point (Arnold, 1990; Bloom, 1996; Chesney-Lind and Rodrigues, 1983; Chesney-Lind and Shelden, 1992; Daly, 1992, 1994; Owen, 1998; Richie, 1996; Smart, 1976). However, these were not completely suited to the research question at hand. That is, in a policy context that placed priority on the treatment of criminogenic needs/risk factors, it was not clear at the time whether needs such as past victimization, abuse and trauma, mental health, parental stress, unhealthy relationships, poverty, and self-efficacy were risk factors or extremely prevalent and unfortunate conditions that were, nevertheless, not bringing women back into the system. We conducted focus groups with women and staff in varied correctional settings, seeking their reactions to questions such as (a) “what areas of difficulty are likely to bring women/you back into the system,” (b) “what do you see as their/your most difficult challenges,” (c) “do you perceive women offenders to be dangerous,” and (d) “do current assessments tap needs most pertinent to you/women offenders?”

Afterward, we set out to test the following gender-responsive needs to determine whether they were predictive of offense-related outcomes: anger, family conflict, relationship dysfunction, child abuse, adult abuse, mental health history, depression (symptoms), psychosis (symptoms), and parental stress. In keeping with emerging research on positive psychology (Seligman, 2002; Sorbello, Eccleston, Ward, and Jones, 2002; van Wormer, 2001), which was finding many advocates among feminist criminologists (Blanchette and Brown, 2006; Bloom et al., 2003; Morash et al., 1998; Schram and Morash, 2002), we also incorporated strengths such as family support, parental involvement, self-efficacy, and educational assets.

We held, perhaps naively, to the belief that the UC/NIC findings would simply suggest an evolution to the Canadian RNR model. As such, one of our goals was to construct a “trailer” assessment that could serve as an addendum to gender-neutral risk/needs assessments such as the LSI-R, the Northpointe COMPAS, and the LS/CMI. In time, the NIC/UC construction validation found the traditional gender-neutral, dynamic risk/need factors and assessments to be predictive, however the addition of the gender-responsive risk/need factors significantly improved the overall predictive validity of the gender-neutral risk/needs assessments for women offenders. This lent support to our assumption that a modification of the basic RNR model would better accommodate women.

Reactions to the UC/NIC validation studies varied. The WRNA Trailer was appended to the Northpointe COMPAS after Northpointe researchers conducted a pilot validation test on a prison sample in the State of California (Brennan et al., 2008). These findings were confirmed on a revalidation prison sample 2 years later (Van Voorhis, 2011; Van Voorhis and Groot, 2010).

Canadian scholars continued to maintain that gender-responsive needs were noncriminogenic and unrelated to future offending (i.e., not risk factors) (Andrews and Bonta, 2010). They stopped far short of advocating that nothing be done to address these areas, however. Instead, the RNR model addresses gender and ethnicity through the specific responsivity and the normative principle of effective intervention. That is, specific responsivity factors are considered needs and conditions that pose a barrier to the treatment of the “Big 4” or “Central 8.” The normative principle maintains that some treatments must be performed in the name of respect for the humane, just, and ethical principles. Translated, the rationale for treating a need such as mental health is to assure that conditions are stable enough to treat the needs identified as criminogenic, especially antisocial attitudes, associates, and personality characteristics. Our position was much different. Because our research found factors such as mental health, parenting issues, trauma, and other gender-responsive needs to be risk factors for future offending, their treatment was not simply a means to another end but a priority. Our research specifically questioned the notion that the “Big 4” should be a given a higher priority over other risk factors for women (Van Voorhis et al., 2010).

Andrews and Bonta's 2010 review of the UC/NIC findings pushed back. Their account of the research selectively reported findings, omitting three of the nine validation study results because they involved gender-neutral assessments other than the LSI-R. They then took findings out of context, indicating that the WRNA Trailer was intended to be an alternative (rather than a supplement) to the gender-neutral LSI-R. Our findings that the block of gender-responsive risk factors made statistically significant contributions to the LSI-R and other gender-neutral assessments seemed to be misunderstood by these authors. They concluded, as some editors also have done, by faulting the UC/NIC study for focusing exclusively on women and not including a comparison group of men. That is right, decades of research on men, generalized to women, and one of the first studies to focus on women is faulted for not having a male comparison group.⁵

Fourth, an emerging body of evidence supports the various components of the gender-responsive approach. Consistent with other attempts to conduct research on women offenders (first point in this address), many of these studies are poorly funded relative to similar research on men. Additionally, currently it is likely that not enough studies exist to support a large meta-analysis, so the gender-responsive approach could be faulted for not withstanding such rigorous scrutiny (point 2). However, an emerging body of research supports the risk factors cited in feminist writings and many of the programs designed to address them. Moreover, there is consistency across studies. In other words, the evidence is far from “nil.”

A modest push for empirical evidence on women offenders in correctional settings came about in the early 2000s. Recognizing the paucity of research on women offenders, the NIC funded a broad review of strategies deemed appropriate to women offenders—Gender Responsive Strategies: Research, Practice and Guiding Principles for Women Offenders (Bloom et al., 2003). One component of this expansive review took the researchers outside of the criminal justice and correctional literature to studies conducted in medicine, mental health, child welfare, domestic violence, substance abuse, education, welfare, and families. The authors addressed several concerns including sexual safety, hiring practices, and sentencing practices, but their treatment recommendations focused on multimodal (wrap-around) services. They recommended a focus on the confluence of mental health, substance abuse and trauma, and interventions to improve women's socioeconomic conditions. Treatment modalities also should incorporate culturally sensitive and relational approaches that maintain women's connections to community, family, children, and other relationships. The authors recommended correctional assessments that would provide a better picture of women's needs and decrease overclassification.

At the same time, many studies compared the needs of male and female offenders. A lengthy review of these studies is beyond the scope of this address; however, the studies generally noted higher rates of mental illness, abuse, and trauma among women than men (see Blanchette and Brown, 2006; Hubbard and Pratt, 2002; Langan and Pelissier, 2001; Messina, Grella, Burdon, and Prendergast, 2007; Salisbury and Van Voorhis, 2009). Another suggestion that the picture of women's risk might be qualitatively different than men's risk appeared on the LSI-R itself. Several authors found the LSI-R predictive for males and females (Kroner and Mills, 2001; Manchek et al., 2009; Smith et al., 2009). However, a comparison of needs scores showed the differences between males and females. For example, women scored significantly higher than men on the emotional personal (mental health) (Holsinger et al., 2003; Manchak et al., 2009; Mihailides, Jude, and Van den Bosshe, 2005; Palmer and Hollin, 2007; Raynor, 2007), family/marital (Holsinger, Lowenkamp, and Latessa, 2003), and financial domains (Heilbrun et al., 2008; Holsinger et al., 2003; Manchak et al., 2009; Mihailides et al., 2005; Raynor, 2007). Women scored significantly lower than men on criminal history (Heilbrun et al., 2008; Holsinger et al., 2003; Manchak et al., 2009; Mihailides et al., 2005; Raynor, 2007), use of leisure time, criminal thinking (Holsinger et al., 2003; Manchak et al., 2009), companions, and substance abuse (Holsinger et al., 2003). Male-to-female comparisons on other measures of the same gender-neutral risk/need factors as those noted on the LSI-R show a similar pattern of findings (e.g., see Bell, 2012; Gehring, 2011). Most of these studies did not compare the predictive merits of each of the LSI-R need domains. However, in one study, financial issues were potent predictors for women, whereas criminal history financial needs and substance abuse were predictive for men (Manchak et al., 2009).

The UC/NIC research took this inquiry a step further to determine whether gender-responsive needs noted in the feminist literature were predictive of future offending and serious prison misconducts. The research generally found the traditional gender-neutral dynamic risk/need factors and assessments to be predictive of recidivism and prison misconducts, but the addition of the gender-responsive risk/need factors improved the overall predictive validity of the gender-neutral risk/needs assessments for women offenders. In addition to findings noted in the previous section with regard to the significant incremental validity of the block of gender-responsive factors, the predictive merits of specific gender-responsive factors identified several important treatment targets. These targets varied somewhat across types of correctional settings (probation, prerelease, and prison) but generally implicated mental health issues, financial problems, parental stress, unsafe housing, and self-efficacy in community settings. Abuse variables seemed to lead to mental health and substance abuse problems in a pathway that ultimately led to recidivism (Salisbury and Van Voorhis, 2009), a pathway that also is observed in other studies (e.g., McClellan, Farabee, and Crouch, 1997; Messina et al., 2007). Risk factors predisposing women to more serious forms of misconduct in prison settings included mental health problems, child abuse, and dysfunctional relationship dynamics. A revalidation study is underway with larger samples and will be completed in the months ahead; however, three samples have been analyzed (see Brushette, Van Voorhis, and Bauman, 2011; Van Voorhis and Groot, 2010), and the results are consistent with the previous construction validation research. Consistent findings also were noted in a study reexamining the Youth Assessment and Screening Instrument (Jones, 2011).

The search for evidence-based programs to address gender-responsive needs also shows a slowly improving empirical picture of “what works” for women offenders. For example, a key risk factor for women's recidivism, especially in community settings, is parental stress exhibited by women who have little financial and emotional support in raising their children and who experience difficulties with child management (Van Voorhis et al., 2010). The Visiting Nurses Program, which is a fairly well-known intervention for at-risk mothers, provides support addressing child health and child management. Experimental research found favorable outcomes for both the children and their mothers who had lower post program offense rates than mothers in a comparison group (Olds et al., 2004). Behavioral child management programs have long showed favorable effects on at-risk children, but we are beginning to learn that they have important outcomes for parents as well (Piquero, Farrington, Welsh, Tremblay, and Jennings, 2009). Another parenting program with promising outcomes is the Female Offender Treatment and Employment Program (FOTEP), which is a residential reentry program for women that offered intensive case management to women and focused on employment and substance abuse. The parenting focus was on reunification with dependent children. Findings showed a reduction in recidivism for FOTEP participants (Grella, 2009).

One gender-responsive principle noted in Gender Responsive Strategies: Research, Practice and Guiding Principles for Women Offenders (Bloom et al., 2003) advocated for wrap-around services. Multimodal services are recommended for most offender populations (see Lipsey, 2009), but two program models tailor the notion to women offenders. Moving On (Van Dieten and MacKenna, 2001) teaches women to access and mobilize varied community resources. Consistent with the emerging profiles of women offenders, Moving On also works with women to enhance strengths, build healthy relationships, and target self-defeating thoughts. The program uses a cognitive-behavioral treatment modality. A matched comparison group study was completed recently among probationers in Iowa, which found significant reductions in recidivism (Gehring, Van Voorhis, and Bell, 2010). A second program, Women Offender Case Management Model (WOCMM) (Van Dieten, 2008), works with correctional practitioners to develop comprehensive case management strategies for women. The development of a network of community services and partnerships is a requirement of WOCMM program sites. The program also trains case managers to address gender-responsive risk/need factors and use strengths-based and relationship-focused approaches. This program also was evaluated within the past year, with favorable decreases in recidivism (Orbis Partners, Inc., 2010; see also Morash, 2010).

Advocating for an approach to substance abuse that recognizes its co-occurrence with mental health and trauma, Stephanie Covington developed a women's substance abuse program Helping Women Recover: A Program for Treating Substance Abuse (2008). The program builds from four perspectives on women's addiction: These accommodate the importance of women's pathways to crime, relationship issues, and addictions co-occurring with mental health issues and trauma. Attention is given to self-efficacy and the impact of sexism and trauma on perceptions of the self and the self in relationship with others. Program modules also discuss families of origin, healthy support systems, sexuality, body image, and spirituality. A second program Beyond Trauma (Covington, 2003) provides information on trauma and its effects and then moves to the development of coping skills. Both programs use cognitive-behavioral elements and exercises but also incorporate psychoeducation, mindfulness, guided imagery, and expressive art techniques. A recent randomized experimental study of both programs administered sequentially found a significantly lower return to prison rates for women in the two gender-responsive programs than those in the standard therapeutic model (Messina et al., 2010). The effects on intermediate outcomes pertaining to psychological well-being also have been favorable (Covington, Burke, Keaton, and Norcott, 2008; Messina et al., 2010).

Two additional programs for addressing abuse and trauma, Seeking Safety (Najavits, 2002) and Dialectical Behavioral Therapy (DBT) (Linehan, 1993), were not developed specifically for offender populations. As such, numerous studies exist, but all speak to favorable intermediate outcomes such as reductions in suicide attempts and drug use and improvements in treatment retention, mental health, and post-traumatic stress disorder (PTSD) systems. Seeking Safety is a cognitive-behavioral program for co-occurring disorders of trauma/PTSD and substance abuse. Evaluation research shows favorable intermediate outcomes, but it was not possible to locate any evaluations of the program's impact on offense-related outcomes (Najavits, Gallop, and Weiss, 2006; Najavits, Weiss, Shaw, and Muenz, 1998). DBT is also a cognitive-behavioral approach involving skills training, motivational enhancement, and coping skills. The impact of DBT has been tested in several treatment settings and found to have many positive intermediate outcomes (for a summary of evaluation findings, see Dimeff, Koerner, and Linehan, 2002).

Another substance abuse program for women, Forever Free, targeted gender-responsive risk factors such as self-efficacy, healthy relationships, abuse and trauma, and parenting. Forever Free included a voluntary aftercare program. The services were multimodal, and the evaluation results showed that the program reduced drug use and recidivism significantly (Hall, Prendergast, Wellish, Patten, and Cao, 2004; Prendergast et al., 2002).

In sum, a promising picture of studies seems to support cognitive-behavioral programs targeted to gender-responsive risk/need factors. Of course, given funding patterns and other roadblocks to implementing women's programs and services (a paper in its own right), it is likely to be some time before enough research on women is available to support a large meta-analysis on women offenders. Indeed, it took decades to amass the male-based, evaluation databases. A more favorable approach to positioning women's programs in an evidence-based policy climate would be to continue to conduct controlled studies of women-specific models and make greater use of the Web-based evidence compilation platforms prepared by the Substance Abuse and Mental Health Services Administration, Office of Justice Programs, and Office of Juvenile Justice and Delinquency Prevention Model Programs. These provide program-specific information with references to the specific programs and program components. The needs of women offenders are too great to wait (years) for a large meta-analysis to produce information on maybe six program characteristics (given the realities of meta-analysis). Moreover, use of “best available,” male-based evidence is not advisable, given the emerging state of knowledge.

Conclusion: Women's Place in the Science of Evidence-based Practice

It is certainly possible that meta-analysis and other forms of study counting will continue to be the gold standard for determining what research is translated into policy. Moreover, limited scholarly interest may continue to collide with weak funding to keep women understudied. In such a climate, however, gender-responsive approaches to correctional treatment are doomed to be caught in the senseless cycle of no research, no evidence, and ongoing adherence to “best available evidence” where the “male is norm.” In contrast, if we can build on available evidence, then a picture is beginning to emerge that outlines an evidence-based approach for women.

From my vantage point, the evidence presented continues to converge on a hybrid model that modifies the prevailing principles of effective intervention for women. However, in the case of some principles, such as the needs principle (defined previously), extensive modification seems to be warranted. I believe also that the principles of effective intervention continue to form a meaningful organizational structure for presenting an evidence-based model for women, but that model differs from the current RNR model in several key ways.

First, the evidence suggests that the risk principle should continue to apply to women but do so with important qualifications. The risk effect (an interaction between risk and intensive treatment) has been found in evaluations of two intensive gender-responsive programs (Gehring et al., 2010; Orbis Partners, Inc., 2010) and one evaluation of gender-neutral halfway houses across the State of Ohio (Lovins et al., 2007). That is, even with women, high-risk offenders have better treatment outcomes in intensive programs than low-risk offenders. Moreover, what too often gets ignored in policy formulations of the risk principle is the fate of low-risk offenders who have worse outcomes even in state-of-the-art, “evidence-based” programs than they might have had if we had not intervened or brought them further into the justice system. By definition, low-risk offenders have many prosocial influences in their lives. These women may need less intensive interventions for fewer needs, but they also will benefit, where possible, from ongoing contact with the prosocial influences in their lives (Salisbury et al., 2009).

The evidence does not support the argument that risk management and risk assessment are inappropriate for women offenders (Blanchette and Brown, 2006). Underlying this argument is the assertion that women are not dangerous and, therefore, should not be classified by levels of risk (Hannah-Moffitt, 2004, 2009; Smart, 1982). In our research, however, 12-month recidivism in community samples ranged from 21% in a probation sample to 44% in a parole sample. Among high-risk groups, these rates are much higher. This is sufficient to support interventions for high-risk women and accurate, assessment-based indications of who they are.

Just the same, an appropriate risk management policy for women should reconceptualize notions of maximum custody and high risk. The high-risk/high-custody female offender is not the same as the high-risk/high-custody male offender, and this is seldom reflected in correctional policy. Most validations of risk and custody assessments find that even in high-risk groups, women reoffend, commit serious misconducts, and return to prison at considerably lower rates than men (Hardyman and Van Voorhis, 2004; Wright, Van Voorhis, Salisbury, and Bauman, 2009).⁶ A simple comparison of high-risk males and females on their rate of offense-related outcomes would, in most cases, reveal this distinction to policy makers and administrators. Women's rates are typically much lower than men's. These comparisons should perhaps be made before impractical investments are devoted to overly secure and austere prison structures located far from children and other supportive family members (Wright et al., 2009). Supervision policies for high-risk females in the community also should reflect differences between males and females (Salisbury, Van Voorhis, Wright, et al., 2009b). Furthermore, legal analyses of this issue are finding that equal treatment for males and females is not required if the data show they are not, in fact, equal (Raeder, 2012).

The scholarship specific to women offenders placed the need principle of the principles of effective intervention under greatest scrutiny (Blanchette and Brown, 2006) and found it to be incomplete and in need of considerable modification (Blanchette, 2009; Buell et al., 2011; Salisbury, Van Voorhis, Wright, et al., 2009b; Van Voorhis et al., 2010; Wright et al., 2009). The commonsense notion still holds that to reduce criminal behavior, we must address the risk factors for criminal behavior. However, recent research has identified a new set of gender-responsive risk/need factors. The programs designed to address these gender-responsive needs seem to be working. Empirical observations of the influences of trauma, mental illness, parental stress, poverty, and unhealthy relationships also suggest a merger of the criminogenic focus of correctional policy with a public health focus (Butler and Engle, 2011). Evidence supports this shift, and the shift advocates well for policies and approaches that bring other social service agencies (e.g., substance abuse, labor, education, mental health, child services, and welfare) to the table. In fact, partnerships among such agencies are observed in several prison reentry programs and in several pretrial, “preentry” programs (e.g., Buell et al., 2011).

As for the general responsivity principle, the emerging gender-responsive programs tend to be cognitive-behavioral programs and, therefore, fit (if somewhat uneasily⁷) within the behavioral, social learning, and cognitive-behavioral modalities noted by the general responsivity principle. To guide gender-responsive programming more carefully in this regard, Blanchette and Brown (2006: 126) reformulated the general responsivity principle:

A gender-informed responsivity principle states that in general, optimal treatment response will be achieved when treatment providers deliver structured behavioral interventions [grounded in feminist philosophies as well as social learning theory] in an empathic and empowering manner [strength-based model] while simultaneously adopting a firm but fair approach.

Finally, the suggestion that women can be accommodated through little more than adherence to the specific responsivity principle or the normative principle (just being decent) (Andrews and Bonta, 2010) is difficult to support on the basis of emerging evidence. Responsivity factors are needs (e.g., parenting) that should be accommodated (e.g., childcare) for the purpose of freeing a client to address a criminogenic need (e.g., antisocial thinking) more effectively. In this example, the treatment of antisocial thinking is the priority and child care is a means to that end. If we referred instead to parental stress as a risk factor, then it assumes greater priority and is addressed with the hope of decreasing the mother's future offending. In addition to child care, then, we might recommend consideration of such interventions as child management programs, parental support groups, and various forms of financial assistance. Thus, the qualitative difference between a responsivity factor and a risk factor has important implications for policy and programmatic strategy. Whether in the realm of criminal justice or public health, a policy of addressing a risk factor is a statement that our goal is to stabilize or ameliorate that condition and that doing so will produce the desired outcome of changing offender behavior. It becomes a stronger priority; it is not a means to an end of treating Big 4 or Central 8 risk factors. Thus, it is not enough to address old treatment targets with a focus that might be more amenable to women, for example, relational therapists or pink walls. For many women offenders, serious attention and priority must be given to a modified list of risk factors.

In closing, most innovative approaches for women offenders have been implemented only within the past decade. Moreover, these changes have occurred on a small scale, largely because the research needed to support such innovation was unavailable in corrections and other fields. More startling, scientific enterprises habitually generalized findings pertinent to men to women, and this practice resulted in substantial costs to women. Additional costs were observed when evidence-based guidelines imposed “best available practices” and faulted the critics of “male is norm” practices for the fact that women are understudied. I do not refrain from issuing the frequent call for more research, but I also wanted to make a case for highlighting the evidence that is present. There is the ongoing risk that women's invisibility to science could extend to a denial of the evidence that is beginning to amass. The evidence on behalf of women offenders is not nil, and policy makers should not be encouraged to ignore it.

Footnotes

• 1

  Some maintain that even these advances would not have occurred without dramatic increases in the number of women incarcerated (Buell et al., 2011). Largely a result of policies promoting mandatory sentencing for drug offenders and reductions in funding for mental health services (see Austin, Bruce, Carroll, McCall, and Richards, 2001; Mauer, Potler,and Wolf, 1999), growth in the size of women's prison populations far outpaced growth in the size of men's prison populations (Bureau of Justice Statistics, 1999). The most recent figures show a decline in state and prison populations (Guerino, Harrison, and Sabol, 2011; Pew Center on the States, 2010); however, the national imprisonment rate declined for men and remained unchanged for women (Guerino et al., 2011).

• 2

  By 1995, the National Institute of Health Revitalization Act of 1993 had been adopted by other federal agencies, including the Agency for Health Research and Quality and the Center for Disease Control and Prevention.

• 3

  Clear evidence of the problem emerged in a study accounting for only those federally funded trials that could have been started after the NIH 1993 guidelines took effect. The authors found 30% of the later studies failed to assemble samples that were composed of at least 30% women or more. This figure increased to 44% when drug trials were examined. Furthermore, 87% of the trials failed to disaggregate findings by sex or include sex as a covariate. None of these acknowledged concerns for generalizability (Geller et al., 2006).

• 4

  Program fidelity will continue to be a major challenge for both the RNR and the new gender-responsive models.

• 5

  Actually, although we cannot deny the irony, we had similar concerns of our own. There are some risk/need factors such as mental health and trauma that we would not want to dismiss out of hand for men, particularly returning veterans and others. We also knew that we had different methods of measuring mental health and abuse, and wanted assurances that they were not risk factors for men as well as women. As a result, we initiated two small studies with male control groups. Both found gender-responsive predictors for women and not men (Bell, 2012; Gehring, 2011). The risk factors found in the male samples were supportive of the RNR model, which is not to suggest that treatments for mental health, abuse, and trauma should be denied to anyone, but they are most appropriate to women. Moreover, treatment priorities for women should be considerably different from those suggested by the RNR model.

• 6

  In prison settings, this comparison should not include minor infractions such as insubordination. These actually tend to be higher for women than men, reflecting poor staff skills in managing women offenders and a tendency to revert to excessive issuance of misconducts to do so (Hardyman and Van Voorhis, 2004). A comparison of serious or aggressive misconducts typically finds much lower rates for women than men.

• 7

  Many of the women's programs also add guided imagery, mindfulness, and other approaches to the cognitive modality.

Court Case Cited

Sharif v. New York State Education Department, 709 F. Supp. 345, 362 (S.D.N.Y. 1989).

Statutes Cited

National Institute of Health Revitalization Act, P. L. 103–43, 103rd Congress (1993).

Prison Rape Elimination Act, P. L. 108–79, 108th Congress (2003).

Violence against Women Act, H.R. 3355, 103rd Congress (1994).

Patricia Van Voorhis, is a Professor of Criminal Justice at the University of Cincinnati. She received her doctorate degree in Criminal Justice from the State University of New York at Albany. Dr. Van Voorhis has published extensively in the leading criminology and criminal justice journals and is the author of Psychological Classification of the Adult, Male Prison Inmate, and lead author of Correctional Counseling and Rehabilitation, currently in its 7^th edition. She has provided expertise to federal, state and local agencies on topics pertaining to correctional effectiveness, program implementation, evaluation techniques, women offenders, risk assessment and correctional classification. She has directed numerous federal and state-funded research projects on the same topics. She recently concluded a federally funded, multisite study of the risk factors for female recidivism.