Legal and Criminological Psychology

Legal and Criminological Psychology (2012), 17, 1–17 C© 2012 The British Psychological Society

The British Psychological Society

An appraisal of the risk–need–responsivity (RNR) model of offender rehabilitation and its application in correctional treatment

Save your time - order a paper!

Get your paper written from scratch within the tight deadline. Our service is a reliable solution to all your troubles. Place an order on any task and we will take care of it. You won’t have to worry about the quality and deadlines

Order Paper Now

Devon L. L. Polaschek∗ School of Psychology, Victoria University of Wellington, New Zealand

The science of effective offender rehabilitation remains a very young field: dominated theoretically and empirically by the work of a small group of Canadian psychologists. Their achievements include the ‘what works’ research literature, and the RNR model of offender rehabilitation. First disseminated in 1990, over the following 20 years, the Risk, Need and Responsivity Principles became the core of the theoretical framework used in those correctional systems around the world that use science as a basis for offender rehabilitation. This paper evaluates the strengths and weaknesses of the RNR model as a Level I rehabilitation framework. It proposes that unrealistic expectations and mistranslations of the model into practice are contributing to concerns about its validity and utility, and stifling needed innovation in the development both of mid-level treatment resources, and of RNR-adherent interventions. It concludes that although the RNR model’s empirical validity and practical utility justify its place as the dominant model, it is not the ‘last word’ on offender rehabilitation; there is much work still to be done.

The scientific study of criminal justice interventions has a short history, from which the first publications on ‘the RNR model of offender rehabilitation’ emerged little more than 20 years ago. Founded on three core principles of offender classification—risk, need, and responsivity—today the RNR model remains the only empirically validated guide for criminal justice interventions that aim to help offenders to depart from that system.

Despite the progress made, the RNR model and its growing knowledge base have had limited impact internationally on correctional responses to offenders. The highly emotive and politicized nature of law-and-order issues in our communities can leave little room for the influence of science. Instead, ‘truthiness’—judging the validity of ideas by their subjective appeal, without reference to facts, logic, or data (Colbert, 2005) —may prevail. Even in nations that both commit criminal justice resources to rehabilitation and that have adopted the model as a matter of policy, the scope and impact of potentially effective interventions is small compared to the systemic resources dedicated to ineffective but ‘true’ approaches, such as increasingly severe and diverse forms of punishment and

∗Correspondence should be addressed to Devon Polaschek, School of Psychology, Victoria University of Wellington, P. O. Box 600, Wellington 6140, New Zealand (e-mail:




2 Devon L. L. Polaschek

surveillance. This wider punitive context often threatens to swamp the application of this knowledge, along with the efforts of those offenders who strive to desist from crime, and those of the rehabilitative programme workers who support them. Implementing knowledge gains—applying what we have learned—despite the forces of ‘truthiness’ is a significant challenge.

Nevertheless, we know much more than we did about what can work, due in large part to the concerted efforts of Canadian psychologists during the 1980s and 1990s. Although their work started much earlier (Wormith, Gendreau, & Bonta, in press), at that time, they adopted the new technique of meta-analysis to create a turning point in the scientific understanding of how to reduce re-offending risk (Andrews, Zinger et al., 1990). Meta-analysis transcended the limitations of narrative reviews, by empir- ically agglomerating superficially disparate programmes and research methodologies, putting them on a common metric: the effect size. For the first time it was possible to use large amounts of information systematically to identify the ‘essential ingredients’ in programmes that worked, and to suggest what was wrong with the many that did not. The ’what works’ research literature had an important role in the development of the RNR model, which was first published in full form in 1994 (Andrews & Bonta, 1994).

However, effect sizes can only take us so far theoretically. Andrews and colleagues theoretical ideas and original research were at least as important in shaping the RNR model. Andrews began in the 1970s to translate promising crime theories into effective correctional service practices (e.g., Andrews, 1980, 1982b).

Although Andrews and colleagues have published numerous papers on the RNR model, its most detailed descriptions are contained in the five editions of their graduate text, The Psychology of Criminal Conduct (Andrews & Bonta, 1994, 1998, 2003, 2006, 2010a). Its strengths as a theoretical framework for offender rehabilitation are substantial; it distils a very large volume of aetiological and intervention-related information into a series of transparently simple principles for application. Notwith- standing on-going development since 1994, areas of weakness remain, but perhaps more troubling are some of the ways in which their work has been translated into practice.

Just as with aetiological theories, the conceptual resources that inform effective programme design vary in their abstractness and complexity, and different levels of models are associated with different purposes and capacities. This paper describes the RNR model, then locates it within a tiered conceptual framework of intervention model development, to guide an appraisal the strengths and weaknesses of the RNR model: what it does well, and what it does not, and—in some cases—cannot do. The paper aims to promote constructive discussion about future developments in the research and practice of offender rehabilitation.

What is the RNR model? The RNR model of rehabilitation is a theoretical framework that outlines both the central causes of persistent criminal behaviour, and some broad principles for reducing engagement in crime. The dissemination of the model began in 1990 with just the three ‘core’ principles: risk, need, and responsivity (along with a fourth, professional discretion; Andrews, Bonta, & Hoge, 1990). The full model includes not just the list of principles for effective rehabilitation, later expanded to 18 (Andrews, 2001; Andrews,



The RNR model of offender rehabilitation: An appraisal 3

Bonta, & Wormith, 2011), but also the body of empirical, theoretical, and practical work on which they rest, known as General Personality and Cognitive Social Learning (GPCSL1) perspective, and its more specific theory of the central mechanisms of criminal conduct, the Personal, Interpersonal, and Community-Reinforcement (PIC-R) perspective.

The model assumes that (a) intervening to help offenders reduce their involvement in crime benefits them and the community around them, and (b) that the only way to intervene effectively is through compassionate, collaborative, and dignified human service intervention that targets change on factors that predict criminal behaviour, (i.e., it is a risk reduction model). It also assumes that correctional rehabilitation is usually resourced by, and accountable to government; although offenders have the same rights as others do to assistance with all aspects of their functioning (e.g., psychological and physical health), correctional programmes do not have a mandate to address those needs that do not lead to reduced involvement in crime (Blanchette & Brown, 2006).

The risk principle has two parts. First, at any point in time, people differ from each other in the likelihood of engaging in criminal behaviour, and this likelihood can be predicted from a wide range of factors, including current attributes and previous criminal behaviour. Level of risk is important because, all other things being equal, more crime can be prevented by targeting higher rather than lower risk offenders for service. Therefore, offenders’ current risk level should be identified prior to making intervention decisions (Andrews, Bonta et al., 1990). Second, significant reductions for higher risk cases require intensive intervention; brief or narrowly focused programmes have little impact (Andrews & Bonta, 1994).

The need principle refers to the targets for change. Criminogenic needs are dynamic attributes of offenders and their circumstances that, when changed, are followed by changes in recidivism (Andrews, Bonta et al., 1990). Andrews and Bonta use the term ‘need’ in their characteristically pragmatic manner, to refer to ‘problem- atic circumstances’ (2010, p. 21), or correctional treatment needs (Ogloff & Davis, 2004).

Andrews and Bonta (1994, 1998) listed six and later eight broad ‘risk/need’ fac- tors: divided into the ‘big four’ (anti-social attitudes, anti-social associates, anti-social temperament/personality, and a static factor: a history of diverse anti-social behaviour) and the ‘moderate four’: family/marital circumstances, social/work, leisure/recreation, and substance abuse. Their presence and ordering is based on meta-analytic results. To assist their translation into treatment targets, Andrews and Bonta listed ‘promising targets for change’ (1994, p. 233); later these were aligned with specific criminogenic needs encompassing both offenders themselves (e.g., reduce drug use) and their social contexts (e.g., change the quality of family supervision).

The third core principle is responsivity, also described as the ‘how’ of intervention (Andrews & Bonta, 2010a): designing and delivering services in ways that engage offenders, help them to learn and change. General responsivity refers to general techniques and processes: behavioural and cognitive-behavioural techniques such as teaching skills and reinforcing prosocial behaviour.

Specific responsivity refers to variations among offenders in the styles and modes of service to which they respond. For example, women offenders may not engage as well with a programme based on men’s preferences and needs. Young, physically active men

1In earlier accounts of the RNR model, the GPCSL is referred to instead as the ‘general personality and social psychology approach’ (GPSPA; e.g., Andrews & Bonta, 1994).



4 Devon L. L. Polaschek

may learn better from shorter, more interactive sessions. Poorly motivated offenders may engage when they understand better how an intervention serves their interests (Andrews & Bonta, 1994, 1998, 2003, 2006, 2010a).

The three core principles are accompanied by ‘overarching principles’, ‘additional clinical principles’, and ‘organizational principles’ (Andrews & Bonta, 2010a; Andrews et al., 2011). Overarching principles include (a) respect for the person and the normative context, (b) basing the programme on empirically validated psychological theory, and (c) the importance and legitimacy of services that prevent crime, even when those services are located outside the criminal justice system. Additional clinical principles state that programmes should target multiple criminogenic needs (breadth), should assess strengths, both for risk prediction and responsivity, use structured assessments of risk, and use professional discretion occasionally on well-reasoned and well-documented grounds. Organizational principles recognize intervention contexts and needed resources. They state that community-based interventions are preferable, that staff practice both the relationship and structuring principles with offenders, and that management must provide, develop, and support the staff and other resources needed (Andrews & Bonta, 2010a). None of these additional principles is new; they draw out and state explicitly material that was present but less accessible in their earlier writing.

The principles derive both from empirical research and from their personal, inter- personal, and community reinforcement (PIC-R) perspective. The PIC-R is described as a specific theory derived from a ‘general personality and cognitive social learning’ (GPCSL) perspective to understanding criminal conduct (Andrews & Bonta, 1994, 1998, 2003, 2006, 2010a). The PIC-R began around 1982 (Andrews, 1982a) as an ambitious attempt at theory knitting (Ward, Polaschek, & Beech, 2006), integrating promising aspects of diverse sociological and psychological aetiological theories and research on crime, and distilling their contributions down to the level of the individual. The primary source theory at the individual level is social learning theory; crucial roles are accorded to cognition and modelling and observational learning (Andrews & Bonta, 1994, 1998, 2003, 2006, 2010a).

Although the fifth edition of the Psychology of Criminal Conduct (Andrews & Bonta, 2010a) is more than twice the length of the first (Andrews & Bonta, 1994), the foundations of the approach remain unchanged. Over successive editions, Andrews and Bonta continued to elaborate, and contextualize their model, and to evaluate its empirical accuracy, based on steady growth in relevant research. The supporting material on each RNR principle, and the PIC-R is extended, incorporating disparate theoretical and empirical advances in criminological psychology and other relevant domains (e.g., clinical, neuro-social, and personality psychology).

The other notable change over time is an increase in the clarity and accessibility of presentation; language is less technical, there is more elaboration of basic concepts, technical aspects of research findings are set beside the main text or in appendices, rather than embedded, and the principles themselves are described in more detail. The potential impact of such apparently superficial changes is substantial for a framework that will be applied to real-world issues; promoting a more complete understanding of the model. Some of these changes represent responses to critiques that have encouraged the authors to pull aspects of the model more clearly into the foreground. For example, the term ‘strengths’ is now featured in the principles themselves, more emphasis is given to resistance and motivation, and there are more explicit statements on attention to non-criminogenic needs.



The RNR model of offender rehabilitation: An appraisal 5

Contextualizing the RNR model as a rehabilitation framework Theories and models are developed to serve particular purposes. They vary on di- mensions that are unrelated to their content domain, such as detail, complexity, and abstractness. No theory—no matter how well crafted—can be all things to all people. Theory evaluation should be conducted with reference to the intended type or level of theory proposed, because different levels of theories inevitably tend to have particular types of strengths and weaknesses.

Integrated and systematic theory building in the criminal justice system—especially intervention theory—remains neglected. Some years ago, Ward and Hudson (1998) recognized a similar problem with aetiological theories informing sex offender treatment, and organized them into three tiers, differing in level of abstraction and compre- hensiveness. This proposed structure—known as the ‘meta-theoretical framework’— when applied to correctional rehabilitation theories similarly enhances understanding of current theories and gaps, identifying high-priority areas for future development (Polaschek, 2011).

This adapted ‘rehabilitation meta-theoretical framework’ is also valuable here. Locat- ing the RNR model within the tiered framework provides clearer expectations about how to judge its quality relative to the intended purpose and inherent strengths and weaknesses of theories of that type.

In Ward and Hudson’s (1998) meta-theoretical framework, theories in the top tier— Level I—are global and multi- factorial. They leave unspecified important details about the inner workings of the phenomenon they seek to explain. Level II theories instead deal with a single aetiological factor: specifying mechanisms, and describing how the factor interacts with other factors. Finally, Level III theories are local theories of the offence process itself (Ward & Hudson, 1998).

In the proposed adaptation for intervention theories, Level I multifactorial reha- bilitation theories are also global, necessarily broad in focus and lacking sufficient detail to directly shape the design of specific interventions. They are hybrids (Ward, Melser, & Yates, 2007): informed partly by aetiological theories, but also incorporating the underlying values and assumptions of intervention, therapy strategies, change processes, programme context and setting, and implementation issues, all in an abstract, ‘high level’, way. Their main purpose is to provide general parameters, in which rehabilitative endeavours will operate, and to support developments on the other two levels (Polaschek, 2011).

Turning next to the bottom, most specific, level of conceptual development: At Level III reside intervention theories for specific programmes. They describe the programme, its processes and content, therapist characteristics, intended client group, expected changes, and so on. They explain their characteristics with reference to Level II and Level I theory.

In the original meta-theoretical framework (Ward & Hudson, 1998), a descriptive theory of the offence process is an impoverished one if it draws only on multivariate aetiological theories of sexual offending. Similarly, the breadth and abstractness of Level I theories makes them unsuited to informing Level III—specific interventions— either for therapists working individually with offenders, or for expert teams of programme designers.

Level II contains the conceptual guides representing broad intervention approaches under the umbrella of a Level I rehabilitation theory. These resources may vary by offender characteristics (e.g., level of criminal risk, gender and age, psychopathy,



6 Devon L. L. Polaschek

criminal gang members) or by the presence of particular needs (e.g., alcohol and drugs) or a focus on change in particular domains (e.g., cognitive skills, vocational skills, reintegration planning, and implementation) or particular settings (court-based, correctional, hospital). Clearly, just as Level I rehabilitation theories are hybrids rather than true theories, these are also not theories in the conventional sense, but again comprise elements of theory, research, and practice resources, but at an intermediate level of abstractness and detail.

What do Level II intervention resources look like? Some concrete models are available for the treatment of sexual offending: they outline how to rehabilitate sexual offenders across a range of risk levels and settings. For example, Marshall, Marshall, Serran, & O’Brien (2011) draw together resources from research, decades of evolving programme practice, and new theoretical developments in positive psychology. They outline a Level II theory: a general approach to ‘strength-based’ sexual offender rehabilitation, and include several Level III derivatives.

Finally, Ward and Hudson (1998) distinguished between Level I theories and ‘theoretical frameworks’. The latter are less well developed and their main function is ‘to help organise research by providing a loose set of constructs with which to approach empirical problems’ (Ward & Hudson, 1998, p. 54). At this point in its development, the RNR model is best thought of not as a fully fledged Level I multivariate rehabilitation theory, but as a framework (Ward et al., 2007). It is not fully developed yet.

Model appraisal criteria Typically, explanatory or aetiological theories are evaluated against a series of criteria that includes empirical validity (retrospective and direct prospective tests), scope, discriminability, generalizability, internal and external consistency, ability to unify existing theories, simplicity, falsifiability, research and practical fertility, explanatory depth, and even originality, creativity or excitement (Sternberg, 2006; Ward et al., 2006).

Although values other than empirical standing are important in evaluating theory quality (see Ward et al., 2006), empirical validity and practical application have primacy when evaluating intervention theories (Gendreau, Smith, & French, 2006; Ogloff & Davis, 2004). This stronger—though not exclusive—emphasis on whether theories or models can prove themselves against both existing evidence, and new data—especially data from the model’s application—is justifiable for two reasons.

First, the criteria listed above apply to more conventional psychological theories. Rehabilitation theories are not aetiological or explanatory theories in the usual sense, but hybrids that bring together theoretical and other resources needed for treatment (Ward et al., 2007). Second, their main purpose is application, rather than knowledge generation for its own sake. So arguably, although conceptual criteria are important, they become irrelevant if empirical evidence does not support the model, or it cannot be adequately implemented in the intended practice setting. The practical consequences of using such a theory may be serious.

Strengths Constructed as it was from existing theory and data, the RNR model has strong unifying power and external consistency. The GPSPCA/PIC-R drew on an embarrassment of



The RNR model of offender rehabilitation: An appraisal 7

riches: even finding merit in Freudian theory, a notable inclusion for committed behaviourists (Wormith, 2011).

It has significant explanatory depth in several important respects. For example, it can explain why programmes as diverse as boot camps, pet therapy, and anger management programmes do or do not work in reducing recidivism, and why programmes’ ‘labels’ do not reliably signal their potential. For example, intensive supervision programmes (ISPs)—punitive community supervision regimes—are usually ineffective, but an ISP in- corporating features of the RNR model can reduce recidivism (see Paparozzi & Gendreau, 2005). Other examples include its ability to explain why a positive relationship between staff and offenders is a necessary but not sufficient condition for effective outcomes (i.e., the relationship and structuring principles) and to pinpoint retrospectively potential sources of loss of programme effect when demonstration programmes are implemented more widely (see examples in Andrews & Bonta, 2010a).

Importantly, they sought to demonstrate how factors from multiple levels of influence, including the political, societal, and neighbourhood, can be seen in people’s behaviour, through translation into social learning principles. Thus, they circumvented a big weakness of social-level aetiological theories: explaining why some individuals are more affected by these factors than others in apparently similar environments. With elegant simplicity, these cognitive–behavioural principles also unify the aetiological with the intervention aspects of the model.

Turning now to empirical validity, it is difficult to fault the RNR model on its consistency with pre-existing empirical data. Newer research has fleshed out but not falsified any major aspects of the model; the abstractness of Level I theories often renders them difficult to falsify. Direct prospective tests of the model, have been based on (a) the core RNR principles (Andrews & Bonta’s 2010 meta-analysis, and independent research, e.g., Hanson, Bourgon, Helmus, & Hodgson, 2009; Lowenkamp, Latessa, & Smith, 2006), and (b) the model itself, using the Correctional Programme Assessment Inventory (CPAI; see Andrews, 2006; Goggin & Gendreau, 2006). This research continues to find that programmes’ ability to reduce criminal offending is related to the extent to which design and implementation conform to the RNR model.

The fertility and practical utility of a rehabilitation theory is demonstrated in new research and new applications. The fertility of the RNR model is also significant. Others have used it to understand the effects of diverse types of programmes: including prisoner re-entry (Listwan, Cullen, & Latessa, 2006), supervision (Pearson, McDougall, Kanaan, Bowles, & Torgerson, 2011; Taxman, 2006), family interventions (Dowden, & Andrews, 2003), and ISPs (Lowenkamp, Flores, Holsinger, Makarios, & Latessa, 2010). It has inspired new assessment tools and approaches (Coebergh, Bakker, Anstiss, Maynard, & Percy, 1999; Gordon & Wong, 2010; Latessa, Lemke, Makarios, Smith, & Lowenkamp, 2010; Ware & Coebergh, 2004), and new tools for assessing programme quality (e.g., the CPAI, see Goggin & Gendreau, 2006). It has found its way into the development of programme accreditation systems (e.g., Bowen, 2011; Hanson, 2005; Lipton, Thornton, McGuire, Porporino, & Hollin, 2000; Lösel, 2010).

It is credited with inspiring a flurry of intervention design; in the last decade or so, several criminal justice systems (e.g., New Zealand, the UK) moved to widespread adoption of assessments and development of programmes based on Andrews and Bonta’s work. Encouragingly, there has been a steady range of research and practice publications on the least well-understood principle of responsivity, on staff behaviour and treatment process (e.g., Beyko & Wong, 2005; Dowden & Andrews, 2004; Hubbard & Pealer, 2009; McMurran, 2009). All in all, the RNR model has made an original substantive



8 Devon L. L. Polaschek

contribution to criminal justice assessment, intervention, research, programme accred- itation, and programme integrity.

Weaknesses The PCC is the only detailed source of information about the RNR model; article summaries often only touch on the three core principles; at best, they discuss very briefly the underlying theory. However, the volume and complexity of material in the PCC makes familiarizing oneself with the full model and its underpinnings a committing task. Although the principles are succinct (Wormith et al., in press), in its totality, the overall framework lacks simplicity or parsimony.

In the early editions of the PCC, this complexity led to confusion, especially with respect to the relationships between different aspects of the model (e.g., the GPSPA, the PIC-R, and the RNR principles;(Ward & Maruna, 2007). The Ward et al. (2007) reformulation demonstrated that this problem could be partially remediated. Later editions of the PCC have made some improvements, expanding the listing of key principles (see also Andrews et al., 2011), and presenting the core messages in diverse and interesting ways that serve to clarify the authors’ views.

The model’s explanatory depth is limited in several key areas. For example, despite its centrality to the model, the responsivity principle is the least developed of the three core principles (Andrews & Bonta, 2010; see Dowden & Andrews, 2004, for an exception). It is theoretically unsophisticated: a catch-all category. Yet, it contains much of what makes the application of the model both humane and effective, so its underdevelopment may have important consequences (Ogloff & Davis, 2004).

For example, it is still unclear why demographic variables such as gender and ethnicity are important, and there remains limited detail about offender motivation and engagement in treatment, despite decade-long Level II developments in motivational interventions (McMurran, 2009) and more recently in treatment readiness (Day, Casey, Ward, Howells, & Vess, 2010). One of the most useful aspects of critiques from strength- emphasizing perspectives is in reminding us of the importance of giving offenders reasons to want to engage in desistance and change (e.g., Porporino, 2010; Ward & Maruna, 2007), not just the capacities to do so. Developing responsivity theoretically could go some way to reduce the impression one can get from the RNR model, that ‘simple operant conditioning’ provides an inherently unsatisfying explanation for how treatment works, and that effective RNR treatment is only concerned with the external manipulation of contingencies for behaviour (Ward, Yates, & Willis, in press).

Another important limitation on explanatory depth, and so a priority for future development is the conceptual gap that exists between an identified list of high- priority dynamic risk factors or targets for change—the central eight—and the theoretical resources needed to translate these factors into intervention design, individual clinical formulations, treatment plans, and change monitoring (Porporino, 2010).

The central eight are empirically well established as correlates both of criminal propensity, and of programmes that reduce re-offending. But each is a proxy, or placeholder for a series of phenomena that Mann, Hanson, and Thornton (2010) recently labelled ‘psychologically meaningful risk factors’; factors that potentially have an actual causal role in on-going involvement in criminal behaviour. A list of broad categories of treatment targets cannot and is not meant to substitute for an adequate understanding of (a) the central mechanisms driving current criminal propensity, and therefore (b) how



The RNR model of offender rehabilitation: An appraisal 9

different risk factors are related to each other, and further (c) how change processes work on these mechanisms for different offenders (Ward & Maruna, 2007).

But rather than expecting a Level I framework—abstract and lacking in detail as it is—to bridge on its own these explanatory gaps, we should instead promote the development at Level II both of single-factor aetiological theories and intervention theories. Our scientific understanding both of treatment targets and treatment change processes remains unsophisticated, and that lack of sophistication must inevitably be imported into Level I theories.

Translation issues The extent to which theorists themselves should be held responsible for others misunderstanding or misapplying their work is no doubt a matter for debate. Regardless of how clearly or carefully a theory is presented, there is always potential for important aspects of it to be ‘lost in translation’. This final section of the critique considers some areas of concern that rather than being inherent limitations of the model itself, may instead be difficulties with knowledge transfer.

Andrews et al. (2011) acknowledge that a lack of clarity and comprehensibility may have contributed to some of the criticisms levelled at their model, and further, that the style of presentation may even have lacked charisma or attractiveness (Andrews & Bonta, 2010a). Experience in teaching with the PCC text since its first edition suggests that they are correct. Both the detached and formal academic writing style and some of the terminology (e.g., ‘clinically appropriate human service’; Andrews & Bonta, 1998, p. 270; and ‘the magnitude of the effect of any one signaled reward for any class of behaviors depends upon the signaled density of other rewards for that class of behaviors’ (Andrews & Bonta, 2006, p. 142) telegraph to students that the roots of this model precede the birth of their parents. Though undoubtedly technically correct, both style and terminology may discourage today’s students of correctional psychology from reading the PCC thoroughly. Instead they may focus more on summaries of the model (e.g., Andrews & Bonta, 2010b; Bonta & Andrews, 2007), including those of other commentators, which are, by definition, social constructions that may be prone to disadvantageous reductionism (Cullen, in press).

If the next generation of therapists and policy makers do not fully understand, or cannot even bring themselves to fully read the relevant material, then their appraisals of it will be superficial or inaccurate, leading to problems with translation into practice. Although really good books should not be judged by their covers, research shows that they are. Simpler, clearer, and more accessible language serves to reduce misunderstandings, and has other important practical implications: readers are likely to like the model more and view it as easier to use (Song & Schwarz, 2008, 2009).

Ward and colleagues have provided the most systematic and thought-provoking critiques of the RNR model. Their critiques include a number of points they suggest may be better thought of as problems with implementation ( Ward et al., 2007; Ward & Maruna, 2007). Several of their observations point to failures to implement core tenets of the RNR model.

For example, Ward and Maruna (2007) suggest that in their experience, programmes derived from the RNR model primarily are conducted for the benefit not of the offender, but of the community at large. Relatedly, these programmes do not motivate or engage offenders since they do not build or promote valuable positive skills and capacities. Further, they cannot support offenders’ interests in desisting from crime and hamper



10 Devon L. L. Polaschek

the development of a good therapeutic alliance. Lastly, in these programmes, offenders’ treatment goals are primarily avoidant, because the overall goal of intervention is to reduce re-offending (Ward & Maruna, 2007).

No doubt such programmes exist. It is hard to imagine, however, that they could be effective in reducing recidivism; they differ in important ways from both the letter and the spirit of the RNR model. Considering these points in turn, first, RNR programmes are conducted for the benefit of the offender and the community; throughout the PCC, Andrews and Bonta refer to the benefits of human service (i.e., rehabilitative) interventions for the offender (e.g., 2010, p. 346). The underlying assumption of the RNR model—and the rehabilitative ethos from which it emerged—is that it benefits both offenders and society to get offenders away from the criminal justice system and into pro-social community life, and RNR rehabilitation is about assisting them in ways that will achieve that goal (see also Bonta & Andrews, 2003).

Typical RNR programmes, even those for the highest risk offenders devote the majority of treatment time to helping prisoners learn skills that we would all find helpful: mood regulation, behavioural regulation, distress tolerance, problem solving, communication skills, and so on (Cortoni, Nunes, & Latendresse, 2006; Polaschek, 2011; Polaschek & Dixon, 2001; Wong, Gordon, & Gu, 2007). Programme environments often also include vocational skills and education components, assistance in reconnecting with families, as well as practical assistance with preparing for community re-entry. Do offenders have other important yet non-criminogenic needs, and would they benefit from other forms of assistance? Without doubt, but the presence of other needs does not make risk-related needs unimportant (Ward et al., in press): and their amelioration of no benefit to the offender.

RNR programmes are fundamentally capacity building: the most effective way to address criminogenic needs is by shaping the development of and differentially rewarding the adoption of positive capacities. This idea goes back at least to early behaviourism (Miltenberger, 2004), and in the RNR model, is incorporated into the core process of ‘building up rewards for non-criminal alternatives’ (Andrews et al., 2011, p. 741).2

For example, Andrews and Bonta (1994, p. 233) noted that promising targets for change include ‘promoting family affection and communication’; ‘replacing the skills of lying, stealing, and aggression with more prosocial alternatives’; ‘promoting identifi- cation/association with anticriminal role models’; and making non-criminal activities in familial, academic, vocational, recreational, and other behavioural settings more rewarding than criminal activities, so that offenders will prefer them. These examples (see also Andrews & Bonta, 2010, pp. 58–60) show too that it is not correct that RNR- based interventions primarily set avoidant treatment goals with offenders. Yes, the overall policy goal—the ultimate goal—of correctional interventions is avoidant: reducing re- offending (Blanchette & Brown, 2006), but this ultimate goal of risk reduction is best achieved through some combination of avoidant and approach goals for offenders.

Ward and colleagues (Ward et al., 2007; Ward & Maruna, 2007) drew attention to these areas of confusion in the context of proposing the Good Lives Model (GLM). There are important differences between the RNR model and the GLM, and encouragingly,

2Theoretically, however, RNR-adhering programmes only develop capacities that either increase responsivity, or reduce criminogenic needs, or both. This is one important point of difference with the Good Lives Model, which instead seeks to develop a wider range of capacities, linked to primary goods (Ward et al., in press).



The RNR model of offender rehabilitation: An appraisal 11

these differences are becoming clearer as the authors of each model respond to each others’ comparisons of the two (e.g., Ward et al., in press; Wormith et al., in press). However, observations such as those above suggest that some differences between the two models that seem substantive on paper, evaporate, or at most, become differences in emphasis when the models are applied (Blanchette & Brown, 2006; Cullen, in press). The basis for these misunderstandings seems to be omissions from the RNR model’s description that occurred because the model was first developed contemporaneously with an evolving correctional rehabilitation ethos that made these points self-evident to those involved at the time.

A good working alliance between therapist or correctional worker and offender is also a founding assumption of the RNR model. It is implicit in the Responsivity principle, but explicitly embodied in the Relationship and Structuring principles: in human service interventions, workers collaborate with offenders (Andrews et al., 2011) and they are warm, empathic, and flexible (Andrews & Bonta, 1994). In fact, Andrews was one of the first correctional researchers to investigate the importance of high-quality relationships between human service workers and offenders in effective interventions; demonstrating that when probation officers are warm and empathic, actively challenge anti-social thinking, and actively reward pro-social thinking, very substantial reductions in recidivism can be achieved (Andrews, 1980; see also Andrews, 1982b; Andrews & Kiessling, 1980).

A second source of difficulties with knowledge transfer can be traced to inferences made about the model based not in the published literature, but on observations of interventions intending to operationalize it. The 18 principles of the RNR model derive in part from the ‘what works’ meta-analyses, in turn based on diverse programme types. By contrast, over the last decade or so, adoption of evidence-based rehabilitation policies has led to a series of large ‘rollouts’ of RNR-based programme suites in countries such as the UK, Canada, and New Zealand, that primarily translate RNR into one style of intervention: structured, cognitive-behavioural closed-group based treatment programmes (Porporino, 2010). These programmes may vary in their target population, and in intensity and setting, but fundamentally they represent at best, only one version of the diversity that is possible within the RNR ambit.

More concerning, programmes subject to large-scale implementation encounter many challenges, especially with regard to treatment integrity (Goggin & Gendreau, 2006), leading these programmes, and the policies that surround them to be highly structured and manualized. But, effective treatment requires a focus not just on good content and technique. Effective relationship skills and treatment process are necessary to achieve treatment goals. The former can be specified in a written manual but the latter is much more difficult to manualize (Marshall, 2009) and may be compromised by efforts to achieve the former. Highly specified content has been used to (a) enable inexperienced, or even undertrained programme staff to deliver treatment simply by ‘reading aloud’ (Mann, 2009), (b) facilitate accreditation processes that tend to evaluate programmes primarily from written documentation (Maguire, Grubin, Lösel, & Raynor, 2010; Marshall, 2009), and (c) enable a simplistic approach to monitoring treatment integrity (i.e., by checking off presented content against manual content; Polaschek, 2011). The use of overly scripted content-focused manuals cannot make up for a lack of well-trained and constructively supervised therapy staff, but does constrict application of the responsivity principle. Therapists cannot adequately tailor the treatment to the client, cannot respond to obvious crises as naturally occurring ‘teachable moments’, appear unempathic and inflexible, rightly leading to ‘cookie cutter’ criticisms (Ward et al., 2007). Adherence to



12 Devon L. L. Polaschek

the responsivity principle requires intervention ‘guides’ rather than manuals (Marshall et al., 2011), well-trained and experienced staff, and practice supervisors who are sufficiently well informed and skilled to monitor integrity—and enhance practice— by evaluating whether goals were achieved rather than whether treatment content that demonstrates only one method of achieving those goals was presented (Polaschek, 2011). Group-format programmes are not necessarily one-size-fits-all (Kirsch & Becker, 2006). They can be responsive to individual offenders’ needs, but only if therapists can use RNR-adherent therapy process and content together to do so.

The lack of effectiveness of some of these widespread roll-outs, by comparison with the demonstration programmes, on which they were based, may be a result of problems with implementation integrity, but they also reflect a failure to adhere to the core RNR principle of specific responsivity. This problem—of omitting or compromising the Responsivity principle—has been explicit in some importations of the RNR model (e.g., Kemshall, 2010). Goggin and Gendreau (2006) found that very few ‘real world’ programmes meet criteria for all three core principles. Although the responsivity principle is the least developed of the three core principles, it does not follow that it is the least important, but perhaps instead that it is the hardest to implement.

Conclusions and future directions The RNR model of rehabilitation seems set to remain the ‘premier rehabilitation theory’ (Ward et al., 2007, p. 222) as long as it continues to enjoy strong empirical support, and as long as scientific data are held in higher esteem than ‘truthiness’. The achievements of the RNR model are quite remarkable: developing and promoting from the quagmire of quackery, confusion, and contradictory findings that preceded ‘what works’ (Latessa, Cullen, & Gendreau, 2002) a deceptively simple series of principles to guide offender rehabilitation.

Yet this appraisal has noted (a) difficulties with complexity, accessibility of language, and clarity in the model itself, (b) large-scale operationalization of a narrow range of RNR programmes, and (c) implementation of interventions that emphasize the core principles of risk and need at the expense of other, equally important principles such as responsivity and core staff practices. Together, these factors may foster a sense of disenchantment with, and misunderstanding about, the model and its value. However well supported it is empirically, futur