Forensic psychology: Part 4: Chapter 7: Assessing offender populations

Assessment and Prediction of Violent Behaviour in Offender Populations

by Ralph C. Serin, Ph.D., C.Psych.^{Footnote 1}

This review is intended to facilitate the development of "best practices" or standards of care for use by clinicians providing psychological assessment of offenders. Such standards are intended to both guide clinicians and insulate them against criticism regarding their competence in the case of failures. This section should be read as a "draft," with substantive revisions anticipated from critiques by Correctional Service of Canada staff and external reviewers. More than anything else, this review of the literature is intended to stimulate discussion, thereby yielding a final document that is both accepted and acceptable. It is not intended that this section will be used specifically for staff training, however, it may provide a framework for a more detailed review of important issues in risk assessment. While the emphasis of this section is on the assessment and prediction of dangerousness, intervention strategies will also be briefly considered. This section will not specifically address the etiology of violent behaviour, however, the importance of conceptual models (Grisso, 1987) will be discussed.

Several strategies are available for the development of guidelines for clinicians. One strategy would be to carefully review the literature, aiming to identify the most robust predictors and incorporate them into a standardized assessment protocol. Another strategy would be to provide an exhaustive list of correlates of violence in offender populations and then to rely on clinicians' conventional wisdom or judgement to determine their relevance for individual offenders. It should be noted there is debate regarding the utility of clinical judgement (Quinsey & Maguire, 1986). At best, this research implies over prediction and a high false positive rate because of over-reliance on offense related information (Menzies, Webster Butler, 1981). This, despite previous violence being an inconsistent predictor of future violent offending (Holland, Holt & Beckett, 1982). Another approach would be to review issues and approaches presented in the literature, thereby informing clinicians, but permitting professional judgement when dearly articulated. The final goal being the derivation of minimum standards reflecting relevant information and decision points in the form of a model or heuristic. This section adopts the latter approach, but emphasizes key predictors and pitfalls reflected in the review of the literature.

By necessity, imposition of a limit to the length of this review will result in some areas receiving insufficient attention at this point. It is hoped this has not resulted in a lack of clarity regarding the issues. A summary of the literature is presented to highlight the direction of this section in the pursuit of guidelines for clinicians. A more detailed discussion of key issues then follows with a brief summary statement. Finally, a bibliography is provided along with appendices which present exemplars of models/approaches from other jurisdictions.

Some parameters regarding definitions are required before proceeding. Dangerousness, dangerous behaviour and violence are distinct constructs (Mulvey & Lidz, 1984) which may have socio political implications, thereby blurring the question of what is being predicted. For our purposes, dangerousness is being defined as "risk of behaviour harmful to others" (Monahan, 1988), which means that suicide is not included in this section. Improved specificity may be helpful in order to reflect both the nature of the dangerous act and the conditions under which the person will engage in such behaviour (Blackburn, 1993a).

A clearer definition of the outcome will also help prediction because those variables relevant in the prediction of institutional violence will differ from those relating to post-release violence (Blackburn, 1993a; Monahan 1988).

Assessment should include both the gathering of information and the evaluation of the probability of a future act (Simon, 1990). Assessment, then, refers to how risk is measured, whereas prediction applies to those specific factors which predict future violent behaviour. In this way, appropriate assessment is critical to accurate risk management (Quinsey & Walker, 1992).

While the literature suggests that mental health professionals have no special expertise in the assessment and prediction of dangerousness (Ennis & Litwack, 1974), it has been convincingly argued that this is neither entirely accurate, nor socially responsible (Blackburn, 1993a; Magargee, 1976). Work by Monahan (1981,1984,1988), in particular, has served to place the limitations of statistical and clinical assessment and prediction into context.

In summary, at this point there is no agreement regarding the definition of violence or dangerousness. Clinicians should be specific regarding the exact question being asked or behaviour(s) to be predicted in their assessments, the time frames of their predictions, and the contextual (person x environmental variables) issues. Establishing such limits should help to ensure decision makers are not misled by the assessment.

A review of the literature provides consensus for the following conclusions:

1. Notwithstanding some overlap (age, PCL-R), those factors which predict (or postdict) violent recidivism are different from general recidivism (Nuffield, 1982).
2. Mental disorder inconsistently correlates with violent recidivism, but may be an important factor in risk management, i.e., compliance with and response to medication (Monahan & Steadman, 1994). Acute psychotic symptoms appear to have greater predictive validity than lifetime diagnosis of schizophrenia (Link & Stueve, 1994), which may in fact be a protective factor (Harris, Rice & Quinsey, 1993).
3. Literature relating to dangerousness in civil psychiatric patients may have limited application to offender populations.
4. While both actuarial and clinical estimates of dangerousness are important, there is debate regarding their relative strengths (Blackburn, 1993a) or methods to combine approaches. The literature fails to support the predictive validity of clinical judgement.
5. Outcome studies which rely only on official records of recidivism are generally accepted as limiting and constitute an underestimate (Hall, 1987; Monahan, 1988; Williams & Gold, 1972). Official records tend to identify the worst offenders, and self-report the less serious (Blackburn, 1993).
6. Statistical models for one population require replication in the target population, which should also be sensitive to demographic disparities. Models relevant for non-natives may be inappropriate for aboriginals. Weights for males will likely not apply for female offenders.
7. An understanding of base rates for general and violent recidivism is essential (Quinsey, 1980), particularly for specific offender groups, (e.g., sexual offenders, murderers, property offenders).
8. The application of group base rates (nomothetic) to individual cases (idiographic) will facilitate estimates of risk and should identify improved risk management strategies. This addresses the issue of anchoring cases.
9. Notwithstanding recent improvements in actuarial models (Harris, Rice, & Quinsey, 1993; Nuffield, 1982, Serin, in press), it is clear that efforts towards an applied behavioural or functional analysis (Blackburn, 1993a) of intrapersonal (dispositional) and situational variables is important for an informed judgement (Mulvey & Lidz, 1984; Pollock, McBain, & Webster, 1989). It is likely that the perfunctory application of any statistical model, (e.g., General Statistical Information for Recidivism), would fail to meet standards of care.
10. There are limits to the extent to which statistical or actuarial models can predict future violent behaviour in offenders (Blackburn, 1993a). Recent models incorporate theoretical factors, i.e., PCL-R, phallometric data. These models should inform assessment and intervention (i.e., the Level of Supervision Inventory [Andrews, 1982] and the issues of responsivity and need).
11. Psychopathy, as measured by the Psychopathy Checklist — Revised (PCLR, Hare, 1991) is a robust predictor of violence in samples of offenders and psychiatric patients, for sexual and nonsexual recidivism.
12. Both distal (Moffit, 1993) and proximal cues (Pithers, 1990) should be considered as they relate to violence potential or risk management. Although the latter refers to sexual offenders, this approach should also be applied to nonsexual offenders.
13. Psychological test results have limited predictive validity, with the Socialization scale from the CPI appearing more helpful than the standard MMPI clinical scales (Gendreau, Grant, Leipciger, & Collins, 1979; Gough, Wenk, & Rozynko, 1965). Projective techniques have been ignored, but have been recently applied to psychopathy (Gacono & Meloy, 1994). It remains to be seen if these have predictive validity.
14. Short term predictions are more accurate than long term predictions (Monahan, 1984), but more stable outcome results from a longer follow up. It may be that high risk offenders fail sooner (Serin, in press) which speaks to the issue of prediction for the duration of release and the requirement for re-assessment in the community.

i) Group versus individual estimates of risk
Base rates relate to group rates or nomothetic approaches. The base rate is the frequency of violent recidivism for a specified population. The lower the base rate of violent recidivism, the greater the likelihood of committing false positive errors, i.e., the likelihood of an offender succeeding when failure was predicted (Quinsey, 1980). Statistical or actuarial methods provide estimates of risk for groups of offenders, but sampling limits generalizability of findings. Harris et al (in press) suggest actuarial estimates of risk be used to anchor a case in terms of the probability of violent recidivism for that category of offenders with similar characteristics (as reflected in the statistical model). For low risk offenders, the proportion of incorrectly identified false negatives (i.e., the likelihood of an offender failing when success was predicted) will be low, while for high risk offenders, the proportion of false positives will be low. A cutoff score or decision point must, however, balance between these types of error. Offenders with moderate risk will be much more difficult to accurately assess, and represent the majority of offenders. This makes the idiographic approach, based on clinical factors crucial to an overall framework for decision-making. This individualized approach also assists the decision-maker in understanding why the clinician views this particular case to be at a certain risk level. Quinsey & Walker (1992) reinforce this view when they comment on risk management. That is, given a particular level of risk, how can this risk be best managed. The level of risk is therefore appropriately applied to intervention strategies or treatment plans (Shapiro, 1990; Simon, 1990).

In summary, at this point, clinicians who ignore or are uninformed by group membership and base rates would likely fail to meet a standard of care in the assessment of dangerousness. How they incorporate clinical information is a different issue that merits careful scrutiny.

ii) Actuarial versus clinical prediction
Actuarial methods select predictor variables according to their relationship to the outcome variable of interest. Variables are weighted to maximize their ability to correctly identify successes and failures. Since Meehl's (1957) work, there has been agreement in the literature that statistical models are superior, however, opposition exists for philosophical and practical reasons. The latter relates to the absence of viable actuarial models in many jurisdictions. This is not the case for CSC, where the General Statistical Information on Recidivism (GSIR, Nuffield, 1982) is part of Case Management, although staff appear uninformed about its derivation and application. Regardless, the GSIR has not proven to be a valid predictor of violent recidivism. Nuffield (1982) describes a modestly successful actuarial model for violent recidivism. The Michigan Department of Corrections have also developed a decision scale to classify risk of future violence (Murph.y, 1982).

The Harris et al. (1993) findings are very encouraging, representing the current standard in actuarial models and yielding predictions greatly higher than chance using weighted historical and clinical (PCLR, phallometric scores for sexual offenders) information. Harris et al (1993) strongly advocate the adoption of an actuarial model, however, limitations regarding the restrictions of this sole approach have been noted (Serin & Barbaree, 1993; Simon, 1990). Further, the Harris et al (1993) model requires replication with a uniquely offender population. The prevalence of mental disorder in their sample, their use of file only PCL-R ratings, and postdictive nature of their work could result in substantial revisions to the specific weights when applied to Canadian corrections. Also, it is unclear if their variables are independent and whether this is preferred in an actuarial model.

Notwithstanding these cautionary notes, their methodology should be applied prospectively to an incarcerated sample, optimally incorporating theoretically chosen variables as well.

Retrospective or postdictive studies are often limited to case history information and demographic variables, thereby excluding the potential contribution of theoretically derived correlates. At a minimum, ignoring such case history correlates as current age, age of onset of criminality, prior violent offenses, and history of substance abuse, is indefensible.

Commenting on how information was integrated into the final assessment might be very helpful in providing a better understanding of the process. Instead of binary decisions (i.e., the presence or absence of a variable equates to risk level), it may be more helpful to attempt to develop a directional model.

The dichotomy of empathy/ callousness may illustrate this point. High empathy might be considered mitigating (+), indifference has no effect (0), and callousness is a risk factor (-). Clearly, this approach requires some thought, but may prove to be of assistance.

Regardless of how risk is assessed, there appears to be no consensus regarding how it might be categorized. For instance, what is the base rate estimate for a low risk offender? How many categories of risk are defensible or reliable? Do the use of percentages imply greater precision in prediction that the data support? How clinical information such as response to institutional treatment programs effects risk is unclear. Lastly, guidelines need to be developed such that revisions to actuarial estimates of risk (e.g. anchors) are not subject to the same criticisms as clinical judgement.

In summary, reliance on clinical factors alone is insufficient and fails to reflect the accepted standard that actuarial estimates are a valid anchor. Actuarial estimates of general recidivism are not valid to address the question of future violent behaviour.

iii) Assessment versus prediction
In order to meet standards of care (Shapiro, 1990; Simon, 1990), clinicians must demonstrate that they are knowledgeable regarding correlates of violence for their particular setting. Several models or "aide memoirs" provide lists of factors. However, how the information is to be incorporated is typically absent. Clinicians must be able to demonstrate they have considered relevant correlates and integrated them into their assessment, taking into account the individual offender's case. See the appendices for examples by Hall (1987); Marra, Konzelman & Giles (1987); Meloy (1987); Simon (1990); Webster (1987; 1992). Webster (St, Eaves (1993) have been the first to formally incorporate psychopathy into a scheme for assessing dangerousness. Scoring criteria are available, however, their scheme is conceptual to date, without data to support differential weights for the various historical, clinical and risk management factors.

A fundamental requirement is the taking of a complete history and review of prior mental health assessments (Beck, 1990). Interestingly, failure to meet this requirement has been grounds for finding clinicians' liable in civil suits in the United States (Beck, 1990), albeit not with offender populations. Similar findings apply to the area of suicide assessment Gobes & Berman, 1993). Corroboration of self-report also appears vital where possible. Additionally, identification of potential victims(s), plausible motive, and access to weapons, must also be assessed (Beck, 1990).

Given the postdictive nature of many outcome studies, development of a standardized clinicial history could inform prospective studies of violent recidivism and typology research in the identification of homogenous groups of offenders with respect to outcome or treatment needs and responsivity. While relevant content areas might be identified from this review, clinicians should be autonomous to gather additional information as they see fit. Blackburn (1993a) has correctly noted that clinical information could be collated and incorporated as a predictor in a statistical model. This area requires further investigation.

With respect to clinical prediction, Quinsey and Abtman (1979) present evidence that psychiatrists do not contribute anything distinctive to the process, as compared to lay persons. Agreement among raters on the likelihood of future violence for particular cases is typically low (Quinsey & Abtman, 1979; Sepejak, Webster & Menzies, 1984). Offense data and previous violence is heavily weighted (Menzies, Webster & Butler, 1981; Werner, Rose and Yesavage, 1983), but leads to an overprediction. This finding should be anticipated in CSC, given the disparity between prevalence rates of violent convictions (estimates peak at 60% depending on the definition of violence) and rates of violent recidivism (approximately 10-19%, Hann & Harman, 1992).

In summary, clinicians need to demonstrate how they arrive at their estimate of risk, including a thorough history and highlighting key factors which either exacerbate or mitigate risk of violent re-offending. Where possible, an "if…then" statement is preferred.

iv) Psychometric assessment
Only modest correlations exist between traditional psychological test results and violent recidivism (Blackburn, 1993a). Combining various clinical scales with other information yields some improvement (Heilbrun & Heilbrun, 1985), but fails to be impressive. An investigation of construct validity for measures of anger, hostility and aggression yielded discouraging findings. Serin (St. Kuriychuk (in press) noted that scores on various measures purported to relate to violence were highly intercorrelated, but uncorrelated with definitions of violent offenders, using criminal history data. Self-report tests with nonsexual offenders may be similar to phallometric measures in sexual offenders. That is, high scores are indicative of a treatment target, but low and moderate scores are uninformative. Utilizing such scores in prediction would yield an unacceptably high false negative error rate.

Notwithstanding its popularity, Minnesota Multiphasic Personality Inventory (MMPI) clinical scales have not proved to be particularly good predictors of violent recidivism (Gendreau, Grant, Leipciger & Collins, 1979). Combining scales with other information appears more promising (Gough, Wenk & Rozynko, 1965; Heilbrun & Heilbrun, 1985). Hall (1988) reports some scales contributed to the prediction of sex offender recidivism. These findings do not suggest the MMPI is uncorrelated with violent recidivism, simply that the correlations are weak and better predictors exist. No evidence of the predictive validity of MMPI typologies (Megargee & Bohn, 1979) was found. It remains to be demonstrated that the MMPI-2 offers any improvement.

Behavioural rating scales (Quay, 1984) other than the PCL-R show some promise. Coping style has been reported to be important in the prediction of recidivism (Porporino, amble & Higginbottom, 1993), but violent recidivism was not considered. From a theoretical perspective, interpersonal style appears to be increasingly important in understanding individual differences and conflict situations (Blackburn, 1993a; Novaco & Welsh, 1989; Serin & Kuriychuk, 1994). Interpersonal style reflects internal scripts or cognitive schema that yield cues about the manner in which an individual interprets others' behaviour. This area warrants substantial investigation, but existing research appears promising: the personality factor (Factor 1) of the PCL-R (Harpur, Hare, & Hakstian, 1989); aggressive beliefs (Slaby & Guerra, 1988); belligerence (Blackburn, 1987); interpersonal dominance in aggressive juveniles (Agee, 1986, 1992); social-information processing deficits in aggressive boys (Dodge, 1986); process measures of impulsivity (Newman & Wallace, 1993); and judiciously selected projective measures (Gacono & Meloy, 1994). These approaches might also contribute to an improved functional analysis of an offender's violence. That is, the extent to which his violence is related to undercontrol, overcontrol, instrumental aggression, anger, hostility, impaired empathy, undersocialized aggression, ego strength (Meloy, 1987), and so forth. The role of recurring themes or fantasies might also be relevant but would fail to fit an actuarial model because of their infrequency in a group of offenders.

In summary, it would appear the contribution of psychometric assessment is to assist clinicians to identify dispositional variables that might be informative in an analysis of the offenders use of violence. Their predictive validity, however, remains modest.

v) Criminal psychopathy
The Psychopathy Checklist, Revised (PCL-R) has been demonstrated to both correlate with and predict violent recidivism in Canadian offender samples (Serin & Amos, 1995; Serin & Barbaree, 1993; Serin, in press). Psychopaths recidivate sooner than nonpsychopaths, both nonviolently and violently. Victim affiliation appears different for nonpsychopaths, whose victims are more typically family members, as compared to psychopaths whose victims are more likely to be strangers (Williamson, Hare & Wong, 1987). Comparable findings have emerged from the Penetanguishene studies (Harris, Rice & Quinsey, 1993). The PCLR has been demonstrated to be a better predictor of violent recidivism than history of prior violence (Serin, in press) and represents the highest correlate of violent recidivism in the Harris et al (1993) actuarial model. Further, violence and psychopathy appear highly correlated (Hare (St McPherson, 1984; Serin, 1991).

The Level of Supervision Inventory (LSI), which correlates very significantly with the PCL-R, may also be helpful in that it may be more responsive to changes in risk relating to successful intervention. It also assesses criminal associates, an important lifestyle variable. The PCL-R should likely be viewed as a static measure, but complementary measures with a greater scoring range could augment the assessment process in an effort to be sensitive to treatment gains.

It is also important for clinicians to be aware that Blackburn presents an alternate conceptualization of psychopathy. In the meantime, an understanding of the construct of criminal psychopathy should enhance the clinical assessment and prediction of dangerousness.

The following practical and professional issues must be addressed before the PCLR might be incorporated in any standard manner into risk assessment by CSC. The PCL-R is copyrighted, so it must be purchased, and clinicians require specific training to meet ethical standards, although to date there is no reported formal use in a clinical setting. Also, consensus regarding cutoffs and how to use a PCL-R score has not been reached (Serin & Barbaree, 1993). Consensus has also yet to be reached regarding the dimensionality of psychopathy, as measured by the PCL-R. Harris, Rice & Quinsey (1993) report psychopathy to be a taxon and therefore categorical. This research has yet to be replicated with an offender sample and Blackburn (1993a) expresses an opposing view. Beyond the theoretical debate, the perspective has important implications as to how psychopathy might be incorporated into clinical practice. Notably, the issues of developing cutoffs and the utility of intervention directly relate to this concern.

Lastly, the impact of labelling offenders as psychopathic versus high risk has not been addressed. Notwithstanding likely concerns by some clinicians regarding the construct of psychopathy and the manner of its assessment (i.e., the debate between Blackburn and Hare), there is some suggestion the construct might inform intervention (Blackburn, 1992; Serin & Kuriychuk, 1994) and responsivity (Ogloff, Wong & Greenwood, 1989; Rice, Harris & Cormier, 1992).

In summary, the PCL-R is increasingly representing a standard in the assessment and prediction of dangerousness in offender populations. Alone, it may be an insufficient strategy and should be considered as part of a multi-method assessment of violence potential.

vi) Conceptual models
The assessment and prediction process requires the psychologist to exercise professional judgement in the cost-benefit analysis related to risk of future violent behaviour (Pollock, McBain, & Webster, 1989). Some clinicians (Simon, 1990) noted the question is not "Should the offender be released," but "If released, what conditions would enhance response to treatment and supervision." Ethical considerations require that clinicians share with offenders the limits of confidentiality before the assessment process begins, and reviewing their report with the offender is also considered to be good practice. This a priori consent can be accomplished by a disclosure form which the offender signs at the initial contact with the psychologist and at the beginning of any aspect of the assessment process. It is also desirable for clinicians to share with offenders their concerns regarding violence potential, particularly if intervention is intended. Whether the clinician who assesses the offender should also provide treatment is as much a matter of personal style (Shapiro, 1990) as resources. The failure to consult with colleagues regarding cases has been viewed by some courts in the United States as inconsistent with standards of care. Blackburn (1993a) endorses collaborative decision making, noting that pooling judges' opinions increases reliability. Training in a particular approach is also encouraged (Kleinmuntz, 1990), although this is infrequent (Heilbrun & Annis, 1988).

Some of the conceptual models presented in the appendices attempt to incorporate proven correlates of violence with clinical factors (i.e., Marra et al, 1987; Meloy 1987; Simon, 1990). Webster's (1987,1992) approach is to ensure both background/historical factors and clinical factors are reviewed, but it is less clear how they are integrated. All suffer the same impediment as Monahan's (1981) list (i.e., they lack empirically derived weights and tend to be somewhat sample-specific). The Harris et al (1993) model has weights, but by its derivation does not incorporate clinical variables and is still sample-specific. Some would argue the use of the PCL-R in their scoring reflects clinical information, however, it is employed as a weighted score in terms of its relation to violent recidivism rather than the 20 item total score. In the absence of any preferred strategy, it is recommended that clinicians review the factors provided in the appendices and provide a summary for the decision maker. From the current literature, it is unclear whether presence of certain factors or a specific number of factors yield a threshold, similar to a cutoff score for an actuarial score.

Risk is not an entirely static construct. Reassessment will be required at various key stages: post-treatment; as the release process proceeds from higher to lower security; prior to release; and subsequent to release to the community. Each stage represents slightly different questions, but having some minimum standards would assist the clinicians to measure progress and permit the offender to have legitimate treatment gains reflected. For instance, in some cases, improvements in employment and accommodation in the community could be viewed as mitigating factors of risk.

Higher risk offenders warrant more comprehensive assessments to ensure an adequate understanding of the preferred strategy for the management of the case. This speaks to this issue of hierarchical assessment, ensuring that psychological resources are utilized for the most serious cases. It is less dear that assigning comparable resources to the assessment of low risk offenders and the policy of offense related referral criteria are sagacious. In a similar vein, it is dear that for some cases, mental health assessments are unlikely to provide substantive increases in predictability. These issues warrant further consideration.

In summary, in order to demonstrate competence, clinicians must provide a rationale for the assessment of risk in the body of their report. Inclusion of relevant correlates of violence potential and their relation to a treatment plan or management strategy appears mandatory. It is dear this task may be more difficult for some cases and will be related, in part, to the time available to complete the assessment.

vii) Mental disorder
The assessment of mental disorder, organicity and substance abuse will all be considered in separate sections. The literature suggests that mental disorder and organicity are inconsistently related to violence potential depending on the setting (Blackburn, 1993a; Monahan, 1988). Both, however, may impact on risk management strategies. Delusions appear over-represented in violence by psychotics, yet may be contingent on such issues as stability of community situation and compliance with medication regimes. Empirically, schizophrenia has been reported to be a protective factor against violent behaviour (Harris et al, 1993), but their sample was not uniquely correctional. Acute symptoms do appear related to increased levels of violence (Link & Stueve, 1994).

Substance abuse is also relevant to violence potential and management/supervision strategies and covered in a separate section. The prevalence of substance abuse is notably higher in offender populations than community samples (Lightfoot & Hodgins, 1988). There may be specific subtypes of abusers who represent greater risk, but this degree of specificity is currently unavailable. Impulsivity is a somewhat elusive construct (Newman & Wallace, 1993; Pulkinnen, 1986), but appears to be related to personality disorders in the clinical lore and warrants further investigation, both as a predictor and moderator variable. Conceptual and psychometric improvements, however, may be required before gains can occur. Most of these latter areas have a genetic link which should be explored in prospective studies. Currently, it would appear all of these areas should be considered by clinicians and articulated in their assessment of dangerousness.

viii) Antecedents to violence
In addition to the notion of triggers from the anger control literature (Novaco & Welsh, 1989), the utility of considering proximal and distal cues is being seen in a relapse prevention view of re-offending.

This work has flourished in the areas of substance abuse (Annis, 1986) and sexual offenders (Pithers, 1990). Loeber (1991) describes distal cues and this work has been extended by Moffit (1993). Their work highlights the importance of developmental patterns in high risk offenders. The manner in which antecedents to violence are measured will likely yield divergent results. Differentiating between situational and dispositional, (i.e., personality) variables would seem important. Violent offenders, however, are a heterogenous group and the relevant distal and proximal cues will surely vary among offenders.

In as much as correlates of general and violent recidivism differ, it is reasonable to anticipate that antecedents will also differ. Self-reported "priming events" such as mood state and substance abuse preceded criminality in parolees by interfering with effective problem solving (Zamble and Quinsey, 1991). Strategies to derive more homogenous groups should serve to inform assessment, intervention, and risk management by providing greater specificity. For instance, differentiating between persistently violent offenders who select nonfamilial victims and situational offenders against family members may prove informative. Clearly, such information cannot be expected to be gathered from cursory contact with an offender. Some preliminary work in this area has begun with sexual offenders (Knight & Prentky, 1990) and should be applied to nonsexual aggression.

ix) Criminal history
A review of an official criminal history would appear to be a minimum standard. Consideration is required of length and variety of criminal history; proportion of violent to total convictions; number of violent convictions; offense severity; age of first arrest; and, juvenile record. Routine carrying of a weapon, possession of a weapon at the time of the offense, and use of a weapon during the offense distinguish differential issues regarding weapon use. How these relate to violence potential is unclear, however, an investigation of weapon use appears prudent. History of escapes and breaches of trust is also important. Further, degree of planning, and therefore premeditation, would seem important, but again of ambiguous relation to violence potential. Likely, criminal history variables are not independent of endorsement of criminal values, but existing measurement of procriminal sentiments (Andrews & Wormith, 1990) has yet to be demonstrated to have predictive validity. Intuitively this has validity and is an item on the LSI, which is predictive of general recidivism.

In summary, additional variables are presented in the literature, which are rarely reflected in existing models. The purpose of their inclusion (i.e., treatment issues versus prediction) should be made clear to the decision maker.

x) Treatability, treatment and dangerousness
As long as the expectation of intervention is both the alleviation of personal distress and reductions in future violence, decisions about treatment goals, treatment efficacy, and outcome criteria will be blurred. In addition to considering a constellation of personal and situation variables (Blackburn, 1993a; Rogers & Webster, 1989), amenability to treatment also relates to responsivity (Andrews, 1989). This includes matching treatment needs to intervention, type of intervention, motivation and response to prior treatment opportunities, and contra-indications (Heilbrun, Bennett, Evans, Offult, Reiff, & White, 1988). Risk and treatability are not necessarily highly correlated, but concretely addressing treatability issues in high risk offenders makes good correctional sense. An example of this is the work on denial and minimization in sex offender treatment (Barbaree, 1993).

With respect to psychopharmacological intervention, Blackburn (1993a) summarizes: "The case for using drugs to control violence is pragmatic, they sometimes enable disturbed patients to function more effectively, but theoretical and empirical justifications remain weak. Diagnosis is rarely sufficient to explain violence, and while psychophysiological and biochemical models continue to be proposed, the effects of drugs on aggressive behaviour appear largely nonspecific. Moreover, voluntary acceptability of psychoactive agents is limited by their unpleasant side effects." Psychopharmacological treatment, however, remains an important consideration in risk management strategies, despite concerns about the use of a "chemical straightjacket."

Anger control programs abound in incarcerated settings, mainly utilizing a cognitive behavioural framework. There is increasing concern about the utility of this approach with males who abuse partners (Edleson & Tolman, 1992). At a minimum, the issues of abuse of power and control need to augment the more traditional approach and to be delivered from a feminist perspective. Many offenders use aggression instrumentally, so this inclusion should probably apply to regular anger control programs. Treatment efficacy data for community samples of aggressive juveniles (Lochman & Lenhart, 1993) cannot be generalized to this population. Hughes (1993) presents some outcome data, albeit with some methodological difficulties. At this point, any CSC outcome research on nonsexually violent offenders should be encouraged. Development of more rigorous investigations would appear to be forthcoming some time in the future. In the meantime, evaluation of consumer satisfaction, pre-post assessments by staff of interpersonal style and institutional misbehaviour, and pre-post assessment of anger knowledge could begin immediately. I believe that decision makers view successful program completion as important for a positive decision. However, there is no empirical support for this halo effect.

Outcome data for substance abuse programs with offenders are available. Treatment tends to be hierarchical, with more serious abusers receiving treatment of longer duration or supplemented by individual sessions with a clinician. Eno, Weekes, Millson, Porporino, Robinson, Fabiano, and Rosine (1993) provide encouraging findings regarding the utility of treatment gain in predicting outcome. More serious abusers had poorer outcomes, however, data for an untreated group matched in terms of severity of abuse are unavailable, so that treatment efficacy is unclear.

Although Blackburn (1989, 1993b) makes a cogent argument for long term therapeutic intervention of personality disordered offenders, such programs do not exist in Canadian corrections. The psychiatric centres in Pacific and Prairies have some form of specialized programs for violent offenders, however, outcome data are unavailable. Ogloff et al (1989) noted that psychopathy was related to poor performance in one such program. Rice, Harris, and Cormier (1989) present data that psychopaths deteriorated as a function of intervention, at least with respect to violent recidivism.

Current understanding regarding viable treatment targets and delivery methods, however, would be critical of some of these earlier efforts.

In summary, involvement in institutional treatment programs may be unrelated to outcome. Clear demonstration of reduction of criminogenic need cannot be discounted, but clinicians should be wary of generalizing to community performance.

xi) Decision guidelines
Given that the decision to grant or deny release is not a uniquely psychological issue, it is recommended that clinicians refrain from making explicit recommendations. Rather, their report should highlight key issues which inform as to an understanding of an offender's likelihood of future violence. This may be summarized in an "if... then statement."

Although disclaimers are increasingly popular, it would appear that if crucial information is absent, the clinician has a responsibility to advise the decision maker of these limitations to the task requested. In extreme situations, insufficient information may warrant the clinician to abstain from completing an assessment.

The development of a process to review assessments with colleagues for consensus appears important. This may have to be restricted to particularly worrisome, high risk cases because of resource issues. Multidisciplinary reviews are also encouraged.

The goal of this section was to familiarize CSC psychologists with the literature and clinical issues relating to the assessment and prediction of violence. Incorporating theses guidelines would reflect "best practices." Meeting these standards of care should then insulate clinicians from retrospective examinations of their assessments and suggestions of incompetence.

Although failures will occur as long as offenders are granted release, meeting a standard of care should better ensure these failures will be viewed as decision errors, rather than mistakes.

The following areas reflect a synthesis of the above review, highlighting areas considered to be central in the assessment of violence. They are intended to be guidelines, but lack sufficient detail to be applied by rote. At this point, it is recommended that the incorporation of this information into an assessment of violence potential reflect "best practices" for CSC psychologists.

• Clinical history:
• review of prior mental health contacts;
• developmental issues (age of onset and type of problem behaviour);
• substance abuse;
• functional analysis of violence (attending to antecedents and triggers);
• identified victim or pattern of victims;
• medical history (major injuries, including neurological concerns);
• employment, financial difficulties, life skills;
• heterosexual relationships & marital situation;
• interpersonal style;
• suicide & self-mutilation history; and
• endorsement of criminal values.

2) Criminality:

• prior escapes;
• density, versatility, length of criminal history;
• age of onset;
• most serious offense;
• victim affiliation, victim injury;
• degree of planning;
• response to previous conditional release(s); and
• prior sexual assaults (paraphiliac/male, prepubescent victim).

3) Psychological testing:

• tests administered and their purpose;
• validity check for malingering;
• relevant offender norms;
• specific relation of test results to criminogenic need;
• relation of test results to treatment requirements;
• hypothesized relation to outcome; and
• neuropsychological assessment, if warranted.

4) Institutional performance:

• involvement in treatment programs;
• meet correctional treatment plan objectives (reduce criminogenic need); and
• adjustment difficulties.

5) Base rate:

• determining and reporting base rate for general and violent recidivism; and
• if a discrepancy exists between base rate and clinical estimate, a rationale for this difference should be provided.

6) Psychopathy:

• consider estimates of failure based on PCL-R score; and
• reflect on relation to group base rate.

7) Release plan:

• matching treatment follow-up needs to level of community intervention available;
• delineation of support versus risk factors;
• matching intensity of supervision to risk level;
• recommendations regarding mental health follow-up;
• clarifying offender's expectations about release; and
• prescriptive risk management strategy by identifying temporal, contextual factors and specific behaviours where possible.

Chapter 7 Appendix A, B, C, D, E, F, G

Agee, V. (1986). "Institutional treatment programs for the violent juvenile," in S. Apter and A. Goldstein (Eds.), Youth Violence: Programs and Prospects. New York: Pergamon.

Agee, V. (1992). Rating Scale for Aggressive Juveniles. Unpublished manuscript.

Andrews, D. A. (1982). The Level Of Supervision Inventory (LSI). Toronto: Ministry of Correctional Services (Ontario).

Andrews, D. A. (1989). "Recidivism is predictable and can be influenced: Using risk assessments to reduce recidivism. Forum on Corrections Research, 1, 11-17.

Andrews, D.A. and Wormith, J. S. (1990). A Summary of Normative, Reliability and Validity Statistics on the Criminal Sentiments Scale. Unpublished manuscript.

Annis, H. (1986). "A relapse prevention model for treatment of alcoholics," inW. E. Miller and N. Heather (Eds.), Treating Addictive Behaviors (pp. 407-435). New York: Plenum.

Barbaree, H. E. (1993). "Denial and minimization among sex offenders: Assessment and treatment outcome," Forum on Corrections Research, 3, 30-33.

Beck, J.C. (1990). "The basic issues," in J.C. Beck (Ed.), Confidentiality Versus the Duty to Protect: Foreseeable Harm in the Practice of Psychiatry (pp. 9-22). Washington: American Psychiatric Press, Inc.

Blackburn, R. (1993a). The Psychology of Criminal Conduct. Chichester, England: Wiley & Sons Ltd.

Blackburn, R. (1993b). "Clinical programs with psychopaths," in K. Howells and C. Hollin (Eds.)., Clinical Approaches to the Mentally Disordered Offender (pp. 179-208). Chichester: Wiley.

Blackburn, R (1989). "Psychopathy and personality disorder in relation to violence," in K. Howells & C. Hollin (Eds.), Clinical Approaches to Violence (pp. 61-88). New York: Wiley & Sons.

Dodge, K. A. (1986). "A social-information processing model of social competence in children," in M. Perlmutter (Ed.), Minnesota Symposium on Child Psychology, Hillsdale, NJ: Erlbaum.

Edleson, J. L. and Tolman, R. M. (1992). Intervention for Men Who Batter. Newbury Park: Sage.

Ennis, B.J. & Litwack, T.R. (1974). "Psychiatry and the presumption of expertise: Flipping coins in the courtroom," California Law Review, 62, 693-752.

Eno, J. Weekes, J. R., Millson, W. A., Porporino,F. J., Robinson, D. Fabiano, E. and Rosine, L. (1993). Offender Substance Abuse Treatment: The Prelease Program. Paper presented at the Annual Convention of the Canadian Psychological Association, Montreal, Quebec.

Gacono, C. and Meloy J. R. (1994). Research Assessment of Aggressive and Psychopathic Personalities. Hillsdale, NJ: Lawrence Erlbaum Associates.

Gendreau, P., Grant, B. A., Leipciger, M., (St. Collins, S. (1979). "Norms and recidivism risks for MMPI and selected scales on a Canadian delinquent sample," Canadian Journal of Behavioural Science, 11, 21-31.

Gough, H. G., Wenk, E. A., & Rozynko, V. V. (1965). "Parole outcome as predicted from CPI, MMPI and a base expectancy table," Journal of Abnormal Psychology, 70, 432-441.

Grisso, T. (1987). "The economic and scientific future of forensic psychological assessment," American Psychologist, 42, 831-839.

Hall, H. V. (1987). Violence Prediction: Guidelines for the Forensic Practitioner. Springfield, Ill: Charles C. Thomas.

Hall, G.C.N. (1988). "Criminal behavior as a function of clinical and actuarial variables in a sex offender population," Journal of Consulting and Clinical Psychology, 56, 773-775.

Hann, R. G. and Harman, W. G. Predicting Violent Risk for Penitentiary Inmates. User Report 1992-08. Ottawa, Ontario: Solicitor General Ministry Secretariat.

Hare, RD. (1991). The Hare Psychopathy Checklist - Revised (PCL-R). Toronto, Ontario: Multi-Health Systems.

Hare, RD. & McPherson, L.M. (1984). "Violent and aggressive behaviour by criminal psychopaths," International Journal of Law and Psychiatry, 7, 35-50.

Harpur, T. J., Hare, R. D., & Halcstian, A. R (1989). Two-factor conceptualization of psychopathy: Construct validity and assessment implications," Psychological Assessment: A Journal of Consulting and Clinical Psychology, 1, 6-17.

Harris, G.T., Rice, M.E., & Cormier, C.A. (1992). "Psychopathy and violent recidivism," Law and Human Behavior, 15, 625-637.

Harris, G.T., Rice, M.E., & Quinsey, V.L. (1993). "Violent recidivism of mentally disordered offenders: The development of a statistical prediction instrument," Criminal Justice and Behavior, 20, 315-335.

Harris, G.T., Rice, M.E., & Quinsey, V.L. (1993b). "Psychopathy as a taxon: Evidence that psychopaths are a discrete class," Journal of Consulting and Clinical Psychology, 62, 387-397.

Hart, S.D., Hare, R.D., & Harpur (in press). "The Psychopathy Checldist - Revised (PCL-R): An overview for researchers and clinicians," in J. Rosen & P. McReynolds (Eds.), Advances in Psychological Assessment, Vol. 8. New York: Plenum.

Heilbrun, K. S. and Heilbrun, M. R. (1985). "Psychopathy and dangerousness: Comparison, integration, and extension of two psychopathic typologies," British Journal of Clinical Psychology, 24, 181-195.

Heilbrun, K. S. and Annis, L. V. (1988). "Research and training in forensic psychology: National survey of forensic facilities," Professional Psychology: Research and Practice, 19, 211-215.

Heilbrun, K. S., Bennett, W. S., Evans, J. H., Offutt, R. A., Reiff, H. J., and White, A. J. (1988). "Assessing treatability in mentally disordered offenders: A conceptual and methodological note," Behavioral Sciences and the Law, 6, 479-486.

Holland, T. R., Holt, N., & Beckett, G. E. (1982). "Prediction of violent versus nonviolent recidivism from prior violent and nonviolent criminality," Journal of Abnormal Psychology, 91, 178-182.

Hughes, G. V. (1993). "Anger management program outcomes," Forum on Corrections Research, 5, 1, 5-9.

Jobes, D.A. & Berman, A.L (1993). "Suicide and malpractice: Assessing and revising policies, procedures, and practice in outpatient settings. Professional Psychology: Research and Practice, 24, 91-99.

Kleinmuntz, B. (1990). "Why do we still use our heads instead of formulas: Towards an integrative approach," Psychological Bulletin, 107, 296-310.

Knight, R A. and Prentky, R.A. (1990). "Classifying sexual offenders: The development and corroboration of taxonomie in W. L Marshall, D. R. Laws, and H. E. Barbaree ( models," Eds.), Handbook of Sexual Assault: Issues, Theories and Treatment of the Offender (pp.23-52). New York: Plenum.

Lightfoot, L.O., & Hodgins, D. (1988). "A survey of alcohol and drug problems in incarcerated offenders," The International Journal of the Addictions, 23, 687-706.

Loeber, R (1991). "Risk factors and the development of disruptive and antisocial behaviour in children," Forum on Correctional Research, 3, 3, 22-29.

Lochman, J. E. and Lenhart, L. A. (1993). "Anger coping intervention for aggressive children: Conceptual models and outcome effects," Clinical Psychology Review, 13, 785-805.

Marra, H.A., Konzelman, G.E., & Giles, P.G. (1987). "A clinical strategy to the assessment of dangerousness," International Journal of Offender Therapy and Comparative Criminology, 31, 291-299.

Meehl, P. E. (1957). Clinical and Statistical Prediction, Minneapolis: University of Minnesota Press.

Megargee, E. I. (1976). "The prediction of dangerous behavior," Criminal Justice and Behavior, 3, 3-21.

Megargee, E. I. and Bohn, M. J. (1979). Classifying Criminal Offenders. Beverly Hills, Ca: Sage.

Meloy, J. R. (1987). "The prediction of violence in outpatient psychotherapy," American Journal of Psychotherapy, 41, 38-45.

Menzies, R. J., Webster, C. D., & Butler, B. T. (1981). "Perceptions of dangerousness among forensic psychiatrists," Comprehensive Psychiatry, 22, 387-396.

Moffit, T. E. (1993). "Adolescence-limited and life-course-persistent antisocial behavior: A developmental taxonomy," Psychological Review, 100, 674-701.

Monahan, J. (1981). Predicting Violent Behavior. Beverly Hills, CA: Sage Publications.

Monahan, J. (1984). "The prediction of violent behavior: Toward a second generation of theory and policy," American Journal of Psychiatry, 141, 1-15.

Monahan, J. (1988). "Risk assessment of violence among the mentally disordered: Geneiating useful knowledge," International Journal of Law and Psychiatry, 11,249-257.

Mulvey, E.P, & Lidz, C.W. (1984). "Clinical considerations in the prediction of dangerousness in mental patients," Clinical Psychology Review, 4, 379-401.

Murphy, T.H. (1980). Michigan risk prediction: A replication study. Unpublished manuscript. Michigan Department of Corrections.

Newman, J.P., & Wallace, J.F. (1993a). "Diverse pathways of deficient self-regulation: Implications for disinhibitory psychopathology in children," Clinical Psychology Review, 13, 721-740.

Newman, J.P., & Wallace, J.F. (1993b). "Psychology and cognition," in K.S. Dobson and P.C. Kendall (Eds.), Psychology and Cognition (pp. 293-349). Orlando, FL Academic Press, Inc.

Novaco, RW., & Welsh, W.N. (1989). "Anger disturbances: Cognitive mediation and clinical prescriptions," in K. Howells and C.R. Hollin (Eds.), Clinical Approaches to Violence (pp 39-60). London: John Wiley & Sons Ltd.

Nuffield, J. (1982). Parole Decision-making in Canada: Research Towards Decision Guidelines. Ottawa, Ontario: Ministry of the Solicitor General.

Ogloff, J.P.R., Wong, S., & Greenwood, A. (1990). "Treating criminal psychopaths in a therapeutic community program," Behavioral Sciences and the Law, 8, 81-90.

Pithers, W. D. (1990). "Relapse prevention with sexual aggressors: A method for maintaining therapeutic gain and enhancing exte rnal supervision," in W. L. Marshall, D. R. Laws, and H. E. Barbaree (Eds.), Handbook of Sexual Assault: Issues, Theories and Treatment of the Offender (pp. 343-361). New York: Plenum.

Pollock, N. McBain, I., & Webster, C. D. (1989). "Clinical decision making and the assessment of dangerousness," in K. Howells and C.R. Hollin (Eds.), Clinical Approaches to Violence (pp. 89-118). London: John Wiley & Sons Ltd.

Porporino, F. J., Zamble, E., & Higginbottom, S. F. (1993). Assessing models for predicting risk of criminal recidivism. Unpublished manuscript.

Pulkinnen, L (1986). "The role of impulse control in the development of antisocial and prosocial behavior," in D. Olweus, J. Block and M. Radke.Yarrow (Eds.), Development of Antisocial and Prosocial Behavior: Research and Theories and Issues. New York: Academic Press.

Quay, H.C. (1984). Managing Adult Inmates. American Correctional Association.

Quinsey, V. L (1980). "The baserate problem and the prediction of dangerousness: A reappraisal," The Journal of Psychiatry and Law, 8, 329-340.

Quinsey, V. L and Abtman, R. (1979). "Variables affecting psychiatrists' and teachers' assessments of the dangerousness of mentally ill offenders," Journal of Consulting and Clinical Psychology, 47, 353-362.

Quinsey, V. L. & Maguire, A. (1986). "Maximum security psychiatric patients: Actuarial and clinical prediction of dangerousness," Journal of Interpersonal Violence, 1, 143-171.

Quinsey, V.L. & Walker, W.D. (1992). "Dealing with dangerousness: Community risk management strategies with violent offenders," in R. DeV. Peters, (Ed.), Aggression and Violence Throughout the Life Span (pp 244-262). Newbury Park, CA: Sage.

Rice, M.E., Harris, G.T., & Cormier, C. (1992). "Evaluation of a maximum security therapeutic community for psychopaths and other mentally disordered offenders," Law and Human Behavior, 16, 399- 412.

Rogers, R. and Webster, C. D. (1989). "Assessing treatability in mentally disordered offenders," Law and Human Behavior, 13, 19-29.

Serin, R.C. (in press). "Violent recidivism in criminal psychopaths," Law and Human Behavior.

Serin, R.C. (1991). "Psychopathy and violence in criminals," Journal of Interpersonal Violence, 6, 423-431.

Serin, R. C. and Amos, N. (in press). "The role of psychopathy in the clinical prediction of dangerousness," International Journal of Law and Psychiatry.

Serin, R.C., and Barbaree, H.E. (1993). "Decision issues in risk assessment," Forum on Corrections Research, 5, 2, 22-25.

Serin, R.C. and Kuriychuk, M. (in press). "Social and cognitive processing deficits in violent offenders: Implications for treatment," International Journal of Law and Psychiatry.

Sepejak, D. S., Webster, C. D., and Menzies, R. J. (1984). "The clinical prediction of dangerousness: Getting beyond the basic questions," in D. J. Muller, D. E. Blackman and A. J. Chapman (Eds.), Psychology and Law. Chichester: Wiley.

Shapiro, D. L. (1990). "Standard of care in the prediction of violent behaviour," Psychotherapy in Private Practice, 8, 43-53.

Simon, R.I. (1990). "The duty to protect in private practice," in J.C. Beck (Ed.), Confidentiality Versus the Duty to Protect: Foreseeable Harm in the Practice of Psychiatry (pp. 23-42). Washington: American Psychiatric Press, Inc.

Slaby, R. G. and Guerra, N. G. (1988). "Cognitive mediators of aggression in adolescent offenders: Assessment," Developmental Psychology, 24, 580-588.

Webster, C. D. (1987). Dangerous Behaviour Clinical Guide (DBCG). Toronto: Clarke Institute of Psychiatry.

Webster, C.D. (1992). Actuarial and Clinical Factors of Dangerousness. Unpublished manuscript.

Webster, C. D. (St.. Eaves, D. (1993) The HCR-20: A Scheme for the Assessment of Dangerousness and Risk. Mental Health, Law, and Policy Institute, Department of Psychology, Simon Fraser University.

Werner, P. D., Rose, T. L., and Yesavage, J. A. (1983). "Reliability, accuracy and decision-making strategy in clinical predictions of imminent dangerousness," Journal of Consulting and Clinical Psychology, 51, 815-825.

Williamson, S., Hare, RD., & Wong, S. (1987). "Violence: Criminal psychopaths and their victims," Canadian Journal of Behavioral Science, 19, 454-462.

Zamble, E. and Quinsey, V. L. (1991). Dynamic and Behavioral Antecedents to Recidivism: A Retrospective Analysis. Research Report R- 17. Ottawa: Correctional Service of Canada.