Psychology, Public Policy, and Law Copyright 1998 by ihe American Psychological Association Inc. 1998, Vol. 4, No. 1/2. 116-137 ' ' 1076-8971/98/53.00 WHAT WE KNOW AND DO NOT KNOW ABOUT ASSESSING AND TREATING SEX OFFENDERS Judith V. Becker William D. Murphy University of Arizona University of Tennessee, Memphis Findings of the Supreme Court in Kansas v. Hendricks (1997) raise a number of issues related to public policy, law, and mental health. The purpose of this article is not to reargue these issues, which are covered in a number of articles in this special issue. Instead, the purpose is to review the clinical literature, empirical literature, and standards of assessment and treatment within the sex offender field and to examine how these impact implementation of the laws. This article outlines an approach to operationalizing the concept of mental abnormality and describes accepted assess- ment approaches, validated risk criteria, and currently acceptable treatment ap- proaches. Limitations of researchers current knowledge about implementing sexual predator laws are also discussed. Public policymakers have made increased efforts during the past decade to protect the public from individuals who commit sexual offenses. As a result, the population of incarcerated sex offenders has grown rapidly. Between 1988 and 1990, the population of sex offenders in American prisons grew by 48%. Currently, one third of the prison population in some states consists of sexual offenders, and every state has laws regarding the registration of sex offenders. Most states require that community members be notified when a sex offender moves into their neighborhood (the level of notification, however, varies according to the serious- ness of the offender's past crimes). A number of states have passed sexual predator laws. Society has struggled for many years with the question of how to deal with sex offenders. In 1938, half of the states had enacted sexual psychopath laws. The underlying assumption was that sex offenders were mentally ill and should receive treatment; once "cured," they could be safely released to the community (Matson & Lieb, 1997). During the 1960s, the trend to adopt these laws had peaked, and by the 1970s and 1980s, the majority of states had repealed these laws. Of the 13 states that retained these statutes up to the early 1990s, few had used them. With the increased public focus on dangerous sex offenders in the late 1980s and early 1990s, the notion of postprison commitment gained in popularity. A number of states (Arizona, California, Illinois, Kansas, North Dakota, Washington, and Wisconsin) have statutes that allow for the involuntary civil commitment of highly dangerous sex offenders upon release from prison (Matson & Lieb, 1997). Matson and Lieb conducted a survey of those six states having postprison commitment and reported that as of August 1997, 240 offenders have been civilly committed in six states. Judith V. Becker, Department of Psychology, University of Arizona; William D. Murphy, Department of Psychiatry, University of Tennessee, Memphis. Correspondence concerning this article should be addressed to Judith V. Becker, Department of Psychology, University of Arizona, 1503 East University, Tucson, Arizona 85721. Electronic mail may be sent tojvhecker@u.arizona.edu. 116 This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. ASSESSING AND TREATING SEX OFFENDERS 117 In June 1997, the United States Supreme Court decided the Kansas v. Hendricks case (1997). The Association for the Treatment of Sexual Abusers (ATSA), an international organization with more than 1,100 members, filed an amicus brief in that case. The ATSA took no position on the constitutionality of the Kansas Sexually Violent Predator Act, but addressed the issue of current diagnostic processes as well as the treatment of sexual offenders. We are both past presidents of the ATSA, and William D. Murphy assisted in the development of the brief for the ATSA. It should be noted, however, that the opinions stated in this article are those of the authors and are not official positions of the ATSA. The goals of this article are basically threefold. The first is to review the concept of mental abnormality as described in the Kansas statute in the context of paraphilic diagnoses and civil commitment. Second, we attempt to acquaint those less familiar with the sex offender literature with the current knowledge about sex offenders in areas most relevant to sexual predator laws. Finally, we outline the implications of our current knowledge, both in terms of the limitations of these laws and recommendations of how they can most effectively be implemented if the goal is to reduce victimization. The Concept of Mental Abnormality The Kansas Sexually Violent Predator Act required as a precondition to civil commitment that the defendant have a mental abnormality. The definition of mental abnormality is vague and has been a controversial aspect of these laws. In their brief, the ATSA (1997) argued that mental abnormality is governed by a statutory definition, and its clinical meaning is derived when professionals give it specific content (p. 3). They further pointed out that terms such as mental disorder or mental illness are generally lacking in specificity, and mental health profession- als should be content-specific when testifying in legal situations as to the nature of the mental illness, disorder, or abnormality. The ATSA also argued, and we concur, that the process of determining whether someone suffers from a mental abnormal- ity is no different than determining whether someone is mentally ill under more typical civil commitment procedures. Generally, in civil commitment situations, this is a two-prong process. First, there is a determination of whether one has a recognizable psychiatric disorder, generally a disorder defined by the fourth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV; American Psychiatric Association, 1994). Following the determination, the mental health professional attempts to determine whether the individual with the disorder is a danger to himself on herself or others. The determination of dangerousness is usually based not solely on the disorder but also on known risk factors. For example, frequently individuals are civilly committed because of their potential for harming themselves. Generally, in making this determination, the mental health professional determines whether the individual has a mental disorder such as a major depression. The presence of a major depression, however, in and of itself is not sufficient for civil commitment. The second step is the determination of the individual's risk for suicide, and, in making this determination, the clinician assesses major factors known to increase the risk of suicide such as age, gender, abuse of alcohol and drugs, previous suicidal attempts, and the presence of a personality disorder including an antisocial personality disorder (Kaplan & Sad- This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 118 BECKER AND MURPHY dock, 1991; Stoudemire, 1994). A similar process could be applied to individuals being considered for civil commitment under sexually violent predator laws. First, the mental health professional would determine whether the individual has a definable mental disorder, and in general, within the context of sexual predator laws, this disorder would be a paraphilia. The mental health professional then would be assigned the task of determining that individual's risk of danger to others, and again this level of dangerousness would be based on looking at known risk factors for reoffending. The Role of Paraphilic Diagnosis Inherent in the discussion above is our belief that if sexually violent predator laws are to be applied, they are most frequently going to be applied to those with paraphilic disorders. There may be other clinical conditions, such as the rape-prone individual (Freund, 1990) to which these laws may be applicable. However, it is not appropriate for mental health professionals to begin creating disorders that are not clearly accepted by a majority of mental health professionals. The DSM-IV (1994) establishes the criteria for the diagnosis of paraphilia: The person must have recurrent, intense, sexually arousing fantasies and sexual urges or behaviors generally involving nonhuman objects or involving the suffering or humiliation of one's partner, children, or other nonconsenting persons that occur over a period of 6 months or more. The fantasies, urges, or behaviors cause significant distress or impairment in social, occupational, or other areas of functioning. It is also accepted by many in the sex offender field that there is a subset of rapists who are paraphilic, although they are not included in the DSM-IV definition (Freund, 1990). It is important for public policymakers and for mental health professionals to realize that sexual predator laws cannot be applied to all individuals with diagnoses of paraphilias. First, an unknown number of individuals probably have paraphilic diagnoses and have recurrent urges to engage in some type of illegal or deviant sexual behavior but never act on these urges. Second, by statute, the sexual predator laws generally would encompass only a certain subset of persons with paraphilia, primarily those who are pedophilic, sexually sadistic, or rape-prone. In addition, it needs to be recognized that even within these subtypes, there is a great deal of heterogeneity (Knight & Prentky, 1990), and not all will show associated risk factors that would suggest they are at high risk to reoffend. Finally, there is also a subset of individuals who may commit sex offenses but who are not paraphilic. In our opinion, sexual predator laws should not be applied to these nonparaphilic individuals. The prime example is the individual with an antisocial personality disorder who may commit an occasional sex offense as part of his or her overall pattern of norm-violating behavior. These individuals do not have a recurrent pattern of sexually disordered behavior and would not appear to be appropriate for sexually violent predator programs. There is clear evidence that psychopathy (a subset of antisocial personality disorders) is a risk factor for reoffending among paraphilic individuals. This fact should not lead to the conclusion that the presence of an antisocial personality disorder alone should lead to civil commitment in the absence of a paraphilic diagnosis. The nature of a paraphilic diagnosis and the concept of sexual urges leads to the question of whether these urges are irresistible impulses or, as noted by a This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. ASSESSING AND TREATING SEX OFFENDERS 119 reviewer of this paper, impulses not resisted. Unfortunately, there is little literature on this question, nor is there much on measuring strengths of these impulses. However, there is adequate evidence in the sex offender field that offenders are not cognitively impaired in the sense that someone with a major psychosis may be, which leads to the question of whether paraphilic acts are volitional in nature. The apparent volitional nature of sex offending is inherent to a number of the arguments against sexually violent predator laws. However, although we feel there is no data that really bears on this question, the general nature of volitional control must be considered. First, in any type of civil commitment, what is volitional is not clearly defined, and the idea of volitional control of a behavior is generally a threshold determination where different thresholds may apply in different situa- tions. For example, the degree of volitional control necessary for the plea of not guilty by reason of insanity is different than the definition of volitional control necessary for civil commitment. Many individuals have been civilly committed who do not totally lack volitional control. Tt is not uncommon for individuals with major depressive disorders not to show their symptoms in all situations, and even if they are harboring severe suicidal thoughts, they can function in certain aspects of their lives. Individuals with borderline personality disorders, a very severe type of personality disorder, are often civilly committed because of their self-injurious behavior, although in most instances these individuals are not cognitively impaired. In summary, in the last two sections, we have tried to point out that the process of determining whether an individual meets criteria for commitment under sexual predator laws does not substantially differ from the process of civil commitment for individuals with other types of disorders. We do believe that mental health professionals should use recognized diagnostic criteria and disorders that are accepted by a body of mental health professionals and should not idiosyncratically interpret concepts such as mental abnormality. We have argued that if these laws were applied, the most frequent diagnosis would be paraphilia, and we have argued that the presence of an antisocial personality disorder alone should not be sufficient for civil commitment. The issue of volitional control is very complex, and there is no data that would allow us to equate or compare in some way the sexual urges of persons with paraphilia with the suicidal ideation of an individual with a major depression. However, many individuals civilly committed are not completely cognitively impaired and show volitional control in many areas of their lives. Etiology The sex offender treatment field, like many areas of mental health, is lacking in data related to the etiology of sex offending, which is unfortunate because a better understanding of etiology would have direct implications for more effective prevention programs and most likely more effective treatment programs. Although data is lacking, policymakers and, more importantly, mental health professionals who may be entering the sex offender treatment field for the first time need to understand what we do and do not know about the development of sexually abusive behavior and sexual disorders. As is the case for many psychiatric disorders, early theories of the etiology of sex offending were psychoanalytic in nature (Karpman, 1954). These early writers focused on theories of infantile sexuality, castration anxiety, and the oedipal This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 120 BECKER AND MURPHY complex. Basically, there was little to no support for such theories, and they generally have fallen by the wayside. Similarly, there has long been a focus within the incest subpopulation of sex offenders on family dynamics as a cause of incest (Alexander, 1985). These theories have tended to place the blame for incest on family enmeshment, confused mother-daughter roles, and family isolation. Within the subset of incest cases, these theories tend to be frequently mentioned; however, there is almost no empirical support for family dysfunction as a cause of incestuous behavior (Crittenden, 1996). The theories that currently tend to predominate the sex offender treatment field have been the behavioral theories, which focus on conditioning of deviant sexual arousal and on the development of cognitions that justify and maintain deviant sexual behavior. Behavioral theories have also focused on the role of poor social or interpersonal skills that make it difficult for individuals to maintain appropriate relationships. Although behavioral and cognitive theories also have mixed support, they clearly have had significant heuristic value and have driven the development of what appear to be more successful treatment programs. Biological factors have many times been posited as causative of deviant sexual behavior. There has been specific focus on the role of the temporal lobe and temporal lobe dysfunction in deviant sexual behavior (Blumer, 1970; Langevin, 1990). However, studies of structural brain damage have not been coordinated with random samples, and many times results have not been replicated across studies. More recently, there has been a focus on the role of neurotransmitters (Kafka, 1997) in modulating sexual behavior. Although a new area of focus, this area does have some promise because it is well established that, in general, neurotransmitters do modulate sexual behavior and can be related to obsessive- compulsive behavior, and some subset of sex offenders do seem to share similar symptoms with persons with other obsessive-compulsive disorders. There are also theories of sex offending that are more specific to offending itself. Primarily, these are related to feminist theories of male socialization (Brownmiller, 1975) and to the role of victimization in the etiology of offending. One primary observation that can be made of sex offenders is that the vast majority are male. This fact has led to a focus on the role of male socialization, especially the development of attitudes supportive of sexually aggressive behavior, such as sexual stereotyping and rape myth acceptance, in the role of offending. There has been support for some of these attitudes at least among a subset of sexually aggressive persons (Malamuth, Sockloskie, Koss, & Tanaka, 1991). However, it is not clear that the research on the relationship between attitudes and sexually aggressive behavior, which has primarily been with college students, applies to the more violent rapists who are likely to fall under the sexual predator laws. Finally, there is a generally accepted popular notion that being sexually abused is a direct cause of sex offending. However, there are a number of problems with this hypothesis; first, it appears that at least among adult offenders, only approximately 30% have actually been sexually abused. However, there are subpopulations, such as those who abuse young boys, who show higher rates of having been victimized (Hanson & Slater, 1988). Although this is a much higher rate than expected in the general population, it is clear that having been a victim of sexual abuse in and of itself will not explain all sexually aggressive behavior. However, it does appear that being a victim can place one at risk for offending. Also, it should be made clear This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. ASSESSING AND TREATING SEX OFFENDERS 121 that the vast majority of young children who are victimized do not become offenders. The above discussion of etiology clearly indicates one important factor about the sexual offender field. That is, there is no one theory that will explain the heterogeneity of offending, and it is likely that these disorders are multicausal. More recently, there has been a focus on developing typologies (Knight & Prentky, 1990) that are multidimensional and that will hopefully reduce some of the heterogeneity so that we can more clearly understand some of the factors causative of at least subtypes of offending. The etiology literature tells us, however, that because of the multicausal nature of sex offending, treatment programs must be comprehensive. Treatment programs must be able to address a wide variety of issues in offenders that may range from general personality disturbance to attitudes supportive of sexual violence and violence in general. Programs that focus on only one aspect of offending behavior are unlikely to be successful. In addition, theories regarding etiology assist in directing us to areas where we can begin prevention efforts. There is sufficient data to suggest that there are certain attitudes associated with sexually aggressive behavior, even if this evidence does not extend to the more violent sexual predator. Prevention efforts should be directed at young men in terms of the attitudes they hold toward women, sexuality, and violence. In addition, although it is clear that not all sexually abusive behavior derives from an offender being a victim, there is some subset where this variable does seem relevant. From a prevention standpoint, those young boys who have been sexually abused must be identified and given adequate interventions early on. Assessment Evaluation or assessment of sexual offenders has developed into a specialized area. The field has its own ethical standards and principles for evaluation (Association for the Treatment of Sexual Abusers [ATSA], 1997) that should be reviewed by any mental health professional, regardless of his or her discipline, who becomes involved in the evaluation process with sexual offenders. In addition, the field has its own process of assessment, areas of assessment, and some specific assessment instruments. Furthermore there are various points at which the mental health professional may get involved in the legal process, and a variety of questions may be asked by the legal system. Mental health profession- als' ability to answer questions posed by the legal system varies considerably. In this section, we briefly outline the process and areas of assessment, the typical questions posed, and the current clinical status and empirical knowledge related to some of the questions. We also briefly address the topics of penile plethysmogra- phy and polygraphy given the role they play in offender assessment and their appearance of objectivity. Guidelines for assessing sex offenders strongly advise that the clinician collect collateral material as part of the assessment process before evaluation. Because of the problem of denial among offenders, it is very important that an assessor attempt to gather information from more than just the offender. Generally, sources of data are victim statements, police reports, previous mental health and medical records, juvenile and adult criminal records, and probation or parole reports when available. The key part of any assessment of an offender is the collection of a This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 122 BECKER AND MURPHY detailed psychosocial and psychosexual history. The psychosocial history includes family, medical, criminal, and mental health history in addition to history of substance abuse. The sexual history focuses on both deviant and nondeviant behaviors in addition to the individual's ability to develop relationships. It is generally accepted within the sex offender treatment field that specific areas should be assessed. These can be classified as general assessment areas and offender specific areas. The general assessment areas are those that would be assessed in most individuals presenting to a mental health professional. They include intellectual level, general measures of psychopathology and personality, general social competence and social skills, self-esteem, and, more recently, assessment of attachment and bonding. It is important to recognize that this aspect of the assessment is not to determine a specific profile of a sexual offender. For example, there have been numerous studies using the Minnesota Multiphasic Personality Inventory (MMPI) to evaluate sex offenders, and it is clear that there is a wide variety of profiles (Levin & Stava, 1987). These profiles do not necessarily differ from other criminal or mental health populations, and there is no single MMPI profile of sexual offenders (Murphy & Peters, 1992). One specific instrument, the Hare Psychopathy Checklist, is increasingly being used with sex offender populations and has consistently been found to be a strong predictor of recidivism. In addition to assessment of general areas, there are sex offender specific areas that are generally assessed. As noted above, detailed sexual histories are part of the sex offender assessment and include the duration of the deviant behavior, the frequency of the behavior, the number and variety of deviant behaviors engaged in by the individual, the grooming process used, and the determination of specific factors that place the individual at risk to reoffend. In addition, sex offender specific assessment focuses on what are generally termed distortions and minimizations, that is, the rationalizations and excuses offenders use. These are assessed either through clinical interviewing or through a number of standardized instruments that have been developed specifically to assess this area. Another major area of sex offender specific assessment is the assessment of denial and degree of denial and the specific assessment of sexual arousal patterns and deviant sexual arousal. Phallometric assessment is a frequent sex offender specific measure of deviant arousal. In a phallometric assessment, changes in penile tumescence are measured during the presentation of a variety of sexual stimuli (deviant and nondeviant) to the individual. The relative magnitude of those responses reflects the amount of sexual interest elicited by the types of stimuli shown. Murphy and Barbaree (1988) have provided a comprehensive review on the use of phallometric assessment and its psychometric properties in decision making. This form of assessment should not be administered without the client's consent and should never be used exclusively to aid in forensic decision making. There are a number of threats to the validity to this form of assessment including the faking of erectile responses. On the basis of their review, Murphy and Barbaree (1988) have concluded that there is insufficient evidence to support the use of erection measures in making decisions about releasing an individual from prison or a treatment program. There is moderate support for the use of erection measures in classifying individuals as to the type of paraphilia they have, particularly persons with pedophilia. The This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. ASSESSING AND TREATING SEX OFFENDERS 123 strongest support is in the use of erection measures to target deviant arousal and to determine treatment, specifically, to use this form of assessment as a pre- and posttreatment measure. Because many sex offenders either deny or minimize their offenses, some clinicians have used polygraphy in an attempt to validate or invalidate an individual's self-report. Unfortunately, there are, to our knowledge, no reports of controlled studies using polygraphy with this population in the literature, and the utility of polygraph methodology in this area is difficult to assess. Regarding this method of assessment, clinicians should consider the unknown effects of repeated testing, the high possibility of false positive errors in some cases, and the issue that physiological responses measured by the polygraph are not specific correlates of lying (Lalumiere & Quinsey, 1991). The specific instruments used and the reliability and validity of specific instruments have been thoroughly reviewed (Milner & Murphy, 1995; Prentky & Edmunds, 1997). However, the mental health professional is usually asked to use these methods to answer specific questions. Before a trial, a mental health professional may be asked to evaluate an alleged offender to determine whether he or she fits a "specific profile of a sex offender." Basically the underlying question is, did this offender commit the specific crime he is charged with? There is clear evidence in the sex offender treatment field that there is no specific profile of a sexual offender (Murphy & Peters, 1992; Peters & Murphy, 1992). It also should be very clear that a mental health professional cannot answer the question of whether someone, even if he or she is a known sex offender, has committed a specific offense at a specific time. Mental health professionals should not attempt to address this question for the legal system. Traditionally, mental health professionals also have been asked to address the offender's amenability to treatment, treatment needs, and risk to the community at either a pretrial stage or the postconviction, preincarceration phase of the criminal justice process. In general, the assessment process outlined above can help in determining amenability to treatment. There is some empirical evidence related to risk prediction. Given our current knowledge of risk assessment and treatment of offender populations, a thorough assessment can provide the legal system some guidelines for the types of treatment needed and, as importantly, the types of external controls necessary. These controls include the need for adequate probation or parole monitoring and an evaluation of the type of support system the individual offender would or should have if he or she is placed in community-based treatment. Assessment at the civil commitment level is in some ways similar to the assessment of risk in community-based treatment that many mental health professionals already provide for the courts. Again, the focus is on known risk factors, and the next section will outline at least moderate support for using actuarial methods to make such assessments. The next level of assessment, and probably one that is most difficult, is determination of successful completion of treatment. This determination of risk following treatment in an outpatient program, a prison-based program, or a program related to civil commitment is much more difficult and one for which empirical support is more limited. Again, as is clearly outlined in the next section, we have much more knowledge of what are termed static risk factors, (i.e., those that are fixed and unchangeable), than dynamic risk factors, which are often the This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 124 BECKER AND MURPHY focus of treatment and which include cognitive distortions, deviant sexual arousal, and empathy. In summary, there are a variety of instruments used in the assessment of sexual offenders. Individually, many of these instruments have adequate psychometric properties. However, the actual validity of the assessment process varies consider- ably depending on the question one is attempting to answer. Our current knowledge does not allow us to determine whether someone has committed a specific crime or, in more general sense, fits the profile of a sexual offender. There is more empirical support for using assessment to design treatment programs and to recommend treatment needs. There is moderate empirical support for using actuarial methods for risk assessment. There is much more limited data, however, in assessing change in risk after someone has undergone a treatment program. In the posttreatment situation, a mental health professional must focus not only on assessing the individual offender, but also on outlining the types of external controls (i.e., monitoring, continued community-based treatment, support systems, living conditions, etc.) that can be put in place to lower an individual's opportunity to reoffend. Risk Assessment For mental health professionals, the findings of the court in Kansas v. Hendricks (1997) presented a challenge—to identify those offenders most at risk to reoffend. The purpose of this section is to acquaint general psychology professionals and mental health practitioners, through a very brief review, with the current knowledge base regarding sex offender risk assessment. Although far from perfect, there has been significant advances in our ability to predict violent behavior in general (Borum, 1996; Mossman, 1994; Rice, 1997; Rice & Harris, 1995) and to predict sex offender recidivism (Quinsey, Lalumiere, Rice, & Harris, 1995; Rice & Harris, 1997) through the use of actuarial methods. Before reviewing the actual data, there are a few points that need to be made. First, as noted previously, improvements in prediction have been made using actuarial methods. We do not want to belabor the point, but, as Meehl pointed out in his classic text (Meehl, 1954), actuarial prediction continues to outperform clinical prediction (Mossman, 1994; Rice, 1997). The focus of this section will be on actuarial prediction. Second, data analytic procedures have changed and the primary procedures now being used are receiving operator characteristics (ROC) analyses. Measures of area under the curve, somewhat similar to effect size, in meta-analyses are frequently used to describe the strength of the prediction. Third, it should be recognized that the accuracy of predicting sex offender recidivism is similar to the accuracy of predicting other violent offenses when actuarial methods are used (Mossman, 1994; Rice, 1997). Finally, sex offenders as a group probably do not reoffend at any higher rate than other violent offenders (Rice, 1997). However, society seems especially concerned about the sex offender population, at least partly because offenses are frequently directed toward children. A wide variety of recidivism rates (Furby, Weinrott, & Blackshaw, 1989) can be found in the sex offender literature. Quinsey, Lalumiere, et al. (1995), in a review of the literature, provided weighted averages for various subgroups. For rapists, the mean weighted rate was 22.8%, and for child molesters, the rate was 20.4%. However, as we will discuss later, at least for child molesters, this rate This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. ASSESSING AND TREATING SEX OFFENDERS 125 varies depending on the sample characteristics. Finally, although this article is only briefly dealing with some select studies in the area, a very thorough meta-analysis of risk factors for 61 data sets is available from Hanson and Bussiere (1996). It has long been recognized that certain victim characteristics, such as sex of victim and relationship of offender to victim, are related to recidivism (Frisbie & Dondis, 1965). More recently, Quinsey, Lalumiere, et al. (1995), again looking at weighted averages, found that offenders against unrelated females had a recidi- vism rate of 18.3%, unrelated males 35.2%, and incest cases 8.5%. In general, similar data have been reported by others (Frisbie & Dondis, 1965; Hanson & Bussiere, 1996; Hanson, Steffy & Gauthier, 1993), although there have been exceptions. Prentky, Knight, and Lee (1997) did not find differences on the basis of the sex of the victim. However, this was a fairly high-risk group of offenders with all having over three known sex offenses; therefore, the variable regarding the sex of the victim may not have added additional risk in such a high-risk group. Similarly, Abel, Mittelman, Becker, Rathner, and Rouleau (1988) did not find differences between those with male or female victims, although they found that those who victimized both sexes had higher recidivism rates. However, this study had a rather large dropout rate, making it very difficult to interpret. Barbaree and Marshall (1988) also did not find a difference, although the sample size was relatively small, with only 35 subjects. There are also a number of criminological variables that predict criminal behavior (Mossman, 1994) and sex offender recidivism. Previous history of criminal offenses, both sexual and nonsexual, are generally found to be predictive of sex offender recidivism (Hanson et al., 1993; Hanson, Scott, & Steffy, 1995; Quinsey, Rice, & Harris, 1995). Similarly, age of arrest (young) and marital status (never married) have also been linked to recidivism (Hanson & Bussiere, 1996; Rice & Harris, 1995; Prentky et al., 1997). Psychological test data have not been highly predictive of recidivism (Hanson & Bussiere, 1996; Proulx et al., 1997). However, two factors—deviant sexual arousal, as measured by penile plethysmography, and psychopathy, as measured by the Psychopathy Checklist (Hare, 1970)—have been strong predictors of recidi- vism in sex offenders. For example, in terms of deviant sexual arousal to children, Malcolm, Andrews, and Quinsey (1993) found a correlation of .42 between arousal measures and sexual recidivism. Quinsey, Rice, and Harris (1995) report correlations of approximately .20 between sexual arousal and both sexual and nonsexual recidivism. In their meta-analysis, Hanson and Bussiere (1996) found a mean correlation of .32 between arousal to children and recidivism, with data being less strong for the relationship between sexual preference for rape and recidivism. Quinsey, Lalumiere, et al. (1995) and Rice and Harris (1997) have found significant relationships between Psychopathy Checklist (PCL) scores and sexual and nonsexual recidivism. Rice and Harris (1997) indicate that there is an interaction between deviant arousal and psychopathy, with those individuals scoring high on both deviant arousal and psychopathy recidivating the most quickly. A number of individual variables that are related to sex offenders have been identified. More recently, these variables, along with other variables from the general criminological field, have been combined in a variety of prediction schemes. Rice and Harris (1997) applied the Violence Risk Assessment Guide (VRAG; Rice, 1997; Rice & Harris, 1995) to sex offenders. The VRAG contains This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 126 BECKER AND MURPHY 12 items, primarily drawn from the criminological area, that include factors such as previous violent offenses, never married, and PCL score. When applied to sex offenders, the VRAG did significantly predict recidivism. However, it did a better job of predicting violent recidivism in general with an ROC area under the curve of approximately .77 than the prediction of sex offender specific recidivism with the area under the curve ranging from .60 to .62. This level of prediction was still significantly better than chance, and Rice and Harris felt that sex offender recidivism prediction could be improved by combining the VRAG with more sex offender specific measures such as victim characteristics and deviant sexual arousal. Hanson (1997a) presents data on a simple four-item scale that includes factors such as previous sexual offenses, age less than 25, extrafamilial victims, and boy victims. He found the average area under the curve to be .70, which is considered a moderate effect size. Quinsey, Rice, et al. (1995) combined a number of measures that included PCL scores, previous criminal histories, victim characteristics, and physiological measures of sexual arousal. These variables correctly predicted recidivism 77% of the time, which was a 44% improvement over chance. Very similarly, Prentky et al. (1997), usi