Does Consideration of Psychopathy and Sexual Deviance Add to the Predictive Validity of the Static-99R?

Abstract

The Static-99 is the most commonly used actuarial risk assessment tool for the prediction of sexual recidivism. In addition, the use of psychopathy and sexual deviance has been common in assessing sexual offenders, based on research suggesting that these factors have predictive validity. It has also become common practice to modify risk assessments based on the Static-99/99R because of the presence of psychopathy and indicators of deviant sexual interests, although to date there has been no research validating this procedure. The current research was conducted to fill this gap in the literature. Using a sample of 272 sexual offenders, the extent to which psychopathy, sexual deviance, and their interaction added to the predictive validity of the Static-99R was examined. Analyses were conducted using the whole sample as well as subgroups of rapists and child molesters. It was found that although the Static-99R predicted sexual recidivism, adding psychopathy and sexual deviance in a Cox regression analysis did not improve the prediction. This held true for child molesters when examined on their own. For rapists, although psychopathy and sexual deviance did not contribute to the prediction of sexual recidivism, for serious (i.e., violent including sexual) recidivism, the inclusion of psychopathy added to the prediction. Results are discussed in terms of implications for practice.

Keywords

sexual offenders risk sexual deviance psychopathy

Those who conduct evaluations of sexual offenders typically rely on actuarial instruments such as the Static-99R (Hanson, Phenix, & Helmus, 2009), which is a collection of 10 static risk factors for sexual recidivism. However, evaluators also often seek to use items outside the actuarials (Doren, 2002) arguing that this provides a more complete assessment of risk. Two of the more commonly used variables in such assessments are psychopathy and phallometrically assessed sexual deviance.

Psychopathy

The assessment of psychopathy as measured by the Psychopathy Checklist–Revised (PCL-R; Hare, 1991, 2003) has become a common practice in North America and appears to play a significant role in sexually violent predator (SVP) assessments in the United States (Jackson & Richards, 2007) and Dangerous Offender proceedings in Canada (Lloyd, Clark, & Forth, 2010). In part, the use of the PCL-R is common because of the relationship between psychopathy and poorer treatment outcome (Hare, 2003). However, there has also been a relatively consistent finding that psychopathy is related to recidivism among sexual offenders. In their meta-analysis of predictors of sexual recidivism, for example, Hanson and Morton-Bourgon (2005) reported an effect size for PCL-R-assessed psychopathy of d = 0.29 (±0.09). Note that a d value in this range is considered to represent a small relationship with the outcome variable (Hanson & Morton-Bourgon, 2004), although it remained one of the more robust predictors in that meta-analysis. Furthermore, Mann, Hanson, and Thornton (2010) noted that a d value of 0.15 corresponds to a difference in recidivism rates of about 5%, whereas a d value of 0.20 corresponds to a difference of about 10%. In noncriminal populations, characteristics associated with psychopathy have also been found to predict sexual aggression. For example, Malamuth (2003) described a program of research on the development of the confluence model of sexual aggression in which features such as impulsiveness, hostility, narcissism, impersonal sexuality, and lack of empathy were significant predictors of sexual aggression.

It is important to note, however, that some researchers have not found that psychopathy predicts sexual recidivism. For example, Barbaree, Seto, Langton, and Peacock (2001) and Langström and Grann (2000) reported that the PCL-R predicted violent and general but not sexual recidivism in their samples. More recently, Murrie, Boccaccini, Caperton, and Rufino (2011) found that psychopathy was unrelated to sexual recidivism in a sample of 333 sexual offenders who underwent an assessment for civil commitment as SVPs.

Sexual Deviance

Among other predictors of sexual recidivism, one of the best individual predictors is sexual deviance. Hanson and Morton-Bourgon (2005) reported that any sexual deviance had a d value of 0.24 (± .12) &x40;±0.12) for the prediction of sexual recidivism, whereas in the Hanson and Morton-Bourgon (2004) meta-analysis, the d value for specifically phallometrically assessed sexual deviance was 0.32.

Mann et al. (2010), in a follow-up to the Hanson and Morton-Bourgon (2004, 2005) meta-analyses, conducted further analyses involving sexual deviance. They noted that although a sexual preference for children was a significant predictor of sexual recidivism in the Hanson and Morton-Bourgon (2004) meta-analysis, sexual violence was not. In the updated meta-analysis (Mann et al., 2010), with one additional study added (Knight & Thornton, 2007), sexualized violence was also a predictor of sexual recidivism (d = 0.18, CI = 0.04, 0.32).

Psychopathy and Sexual Deviance

These relatively robust findings have led some researchers to examine the combination of psychopathy and sexual deviance in predicting sexual recidivism (e.g., Olver & Wong, 2006; Rice & Harris, 1997). It is commonly assumed that the combination of high psychopathy and sexual deviance is a “deadly combination” (Hare, 1999), in that those sexual offenders with both psychopathy and sexual deviance are assumed to reoffend at a very high rate. However, the research results do not lend consistent support for this conclusion. First, a literature review conducted by the first author (J.L.) indicates that there are relatively few studies that examine the combination of psychopathy and phallometrically measured sexual deviance, and second, the findings of the research have been inconsistent. In the following section, we summarize the existing studies in this regard.

Rice and Harris (1997) were the first researchers to examine the effect of psychopathy and sexual deviance in the prediction of recidivism among a sample of 288 sexual offenders. They found that psychopathic sexual offenders with phallometrically assessed sexual deviance (i.e., an absolute preference for deviant stimuli such as children, rape cues, or nonsexual violence cues) had higher sexual recidivism rates, and they recidivated more quickly than psychopathic/nondeviant, nonpsychopathic/deviant, and nonpsychopathic/nondeviant sexual offenders, all of whom reoffended at approximately the same rate. However, they did not find this to be the case for violent recidivism. Harris et al. (2003) using a sample of 155 sexual offenders reported that in a Cox regression analysis the interaction between psychopathy and sexual deviance predicted both violent and sexual recidivism, although the relationship with sexual recidivism was weaker than that for violent recidivism. They also reported that there was a significant main effect for the PCL-R score in survival analyses but not for sexual deviance. In addition, the deviance index was significantly related to violent but not sexual recidivism in simple correlational analysis; see Table 1 of Harris et al. (2003).

Serin, Mailloux, and Malcolm (2001) found, using a sample of 68 sexual offenders and general recidivism as the outcome, that those with high PCL-R scores and evidence of phallometrically assessed sexual deviance reoffended sooner and more often than offenders with high sexual deviance/low PCL-R scores. The high psychopathy/deviant group also reoffended at a higher rate than those with low sexual deviance and either high or low PCL-R scores, the latter two groups of whom reoffended at approximately the same rate. Unfortunately, these researchers did not differentiate between sexual and nonsexual recidivism because of low recidivism rates. Similarly, Gretton, McBride, Hare, O’Shaughnessy, and Kumka (2001), in a sample of 220 adolescent males, found that the high psychopathy/sexually deviant group was more likely to reoffend with any charge during the follow-up period than other groups. They were also more likely to reoffend in a violent fashion than the low psychopathy/low deviance group; however, no other groups differed on this outcome. There were no significant effects for sexual recidivism.

Thus, there are four published studies addressing the relationship among psychopathy, phallometrically assessed sexual deviance, and recidivism. These studies have produced mixed results with (a) those high on both psychopathy and sexual deviance having a higher sexual reoffence rate than other offenders (Rice & Harris, 1997), (b) a strong relationship between the interaction and violent recidivism with a weaker effect for sexual recidivism (Harris et al., 2003), (c) a finding that those who are high on both psychopathy and sexual deviance violently reoffend at a higher rate than others but have no effect for sexual recidivism (Gretton et al., 2001), and (d) a finding that those who are high on both psychopathy and sexual deviance reoffend with nonsexual and nonviolent offences at a higher rate than others (Gretton et al., 2001; Serin et al., 2001).

Proxy Measures of Sexual Deviance

Other researchers, using nonphallometric measures of sexual deviance, have also explored the relationship among sexual deviance, psychopathy, and recidivism. For example, Olver and Wong (2006) used the Sexual Deviance scale of the Violence Risk Scale–Sexual Offender version (VRS-SO) (Wong, Olver, Nicholaichuk, & Gordon, 2004) to assess the presence of sexual deviance. The VRS-SO is a structured clinical rating scale for the assessment of risk of sexual recidivism. The Sexual Deviance scale provides a structured rating of sexual deviance using five items: Sexually Deviant Lifestyle, Sexual Compulsivity, Offense Planning, Sexual Offending Cycle, and Deviant Sexual Preference. Each item is rated on a 4-point scale (0, 1, 2, and 3); the higher the rating, the more the rated individual resembles the item description (Olver & Wong, 2006). These researchers followed 156 sexual offenders who had been rated on the PCL-R and the VRS-SO Sexual Deviance scale, and found that although the high psychopathy–high deviance group had the highest recidivism rate, their rate of reoffence was not significantly different from the other groups. In a Cox regression analysis, the psychopathy–deviance interaction was significant in predicting sexual recidivism. In a survival analysis, however, the high psychopathy–high deviance group differed significantly from only the low psychopathy–low deviance group. Other groups did not differ significantly from each other.

Other studies that use proxy measures for sexual deviance have also been conducted (e.g., Hildebrand, de Ruiter, & de Vogel, 2004; Seto, Harris, Rice, & Barbaree, 2004) and have found similar results. It is important to note, however, that these proxy measures are not the same as phallometric assessment. They tend to be broader in their sampling of exemplars of sexual deviance, thus assess features in addition to penile tumescence. The VRS-SO, for example, as described above, includes results of phallometric testing as part of its measure of sexual deviance but is much broader in its scope. The sexual violence risk (SVR-20) (Boer, Hart, Kropp, & Webster, 1997), which was used in the Hildebrand et al. (2004) study cited above, contains a sexual deviance item, which not only may include phallometrically assessed deviance but also considers diagnosis of paraphilia. Thus, one cannot extrapolate from the results of these studies directly to phallometric studies.

Studies employing other methods of assessing features related to psychopathy and sexual deviance have also found that the combination predicts sexual misbehavior. For example, Knight and Sims-Knight (2003) reported path-analytic research, which indicates that antisociality, sex drive/preoccupation, and callous/unemotional traits predict sexual coercion in community and criminal samples. Also, Malamuth (2003), already discussed above, reported that in his community samples, factors related to antisociality and impersonal sexual behavior were associated with sexual aggression.

Thus, although it has been found that psychopathy, sexual deviance, and their interaction are related to sexual and violent recidivism, the magnitude of the relationship has varied from sample to sample. In addition, much of the research (more than half of the published studies) has used proxy measures for sexual deviance, rather than sexual responding directly assessed by phallometric testing. Note that in only one study (Rice & Harris, 1997) was the “deadly combination” regarding sexual reoffence found.

Actuarial Assessment of Risk

As noted above, evaluators typically use actuarial measures such as the Static-99 in their assessments of risk for sexual reoffence (Hanson & Thornton, 2000). The Static-99 is the most widely used actuarial risk assessment tool for the assessment of risk in sexual offenders (Hanson & Morton-Bourgon, 2009). It has demonstrated reliability (e.g., Barbaree, Langton, & Peacock, 2006) and a moderate relationship with both sexual and violent recidivism. Recently, the developers of this instrument (Helmus, Thornton, Hanson, & Babchishin, 2011) have presented a revision that better accounts for the relationship of age to sexual recidivism. This revised instrument is called the Static-99R and, at least in the developmental research, has a slightly improved relationship to sexual recidivism over the original instrument (Helmus et al., 2011).

It is a relatively common practice in risk evaluations to consider factors outside the actuarials (Doren, 2002), despite the finding that unstructured adjustments of actuarial assessments lead to decreased predictive validity (Hanson & Morton-Bourgon, 2009). Based on Doren (2002), two of the factors most commonly used in such adjustments are psychopathy and sexual deviance. Doren (2002) justifies this recommendation based to a large extent on the research findings discussed above. To date, however, no one has examined the extent to which consideration of psychopathy and sexual deviance adds to the predictive validity of risk assessment. The purpose of the current article is to fill this gap in the literature.

Method

The current research used a sample of sexual offenders assessed and/or treated at the Regional Treatment Centre Sexual Offender Treatment Program (RTCSOTP). The RTCSOTP is a sexual offender treatment program that provides treatment to high-risk/treatment needs sexual offenders serving a sentence of 2 years or more within the Correctional Service of Canada. The RTCSOTP is described in detail elsewhere (Abracen, Looman, & Langton, 2008; Looman, Abracen, & Nicholaichuk, 2000). Briefly, the RTCSOTP is a cognitive-behavioral, group-based treatment program provided in a residential setting. Men are admitted to the Regional Treatment Centre (RTC), which is an accredited psychiatric hospital serving the prison population in Ontario, to complete the program; however, the majority of the offenders treated reside in general population institutions prior to and following participating in the treatment program.

Sample

The current sample consists of 272 sexual offenders assessed between November 1992 and March 2006, drawn from a larger sample of 455 consecutive admissions to the RTCSOTP. Offenders included in the current analyses were those who have Static-99R and PCL-R scores, and phallometric testing results available, and had been released from prison. Various portions of this sample have been used in previous research by the current researchers (e.g., Looman, 2006; Looman & Abracen, 2010; Looman & Marshall, 2005).

Demographic data concerning the current sample are provided in Table 1. Of the current sample, 142 (52.2%) offenders had victims aged 16 or older (i.e., adult rapists), 34 (12.5%) had victims aged 13 to 15 inclusive, 74 (27.2%) had victims aged 12 or younger (i.e., child molesters), 15 (5.5%) offended within the family (i.e., incest offenders), and 7 offenders (2.6%) had victims in more than one age group.

Table 1.

Descriptive Statistics for the Entire Sample.

Variable	M (SD)
Age at testing	35.9 (9.4)
Age at release	38.3 (9.8)
Race (% Caucasian)	79.4
Number of previous sexual offences	3.6 (4.6)
Number of previous violent offences	1.7 (3.5)
Age first arrest	20.1 (7.4)
Sentence length (months)	65.4 (48.6)
Static-99R score	5.4 (2.2)
PCL-R score	22.3 (7.7)
Differential deviance index	0.71 (2.22)
Proportional deviance index	1.70 (2.48)

Note: PCL-R = Psychopathy Checklist–Revised.

Measures

Static-99R

Static-99R (Helmus et al., 2011) scores for the current research were taken from Static-99 scores obtained for previous research with the same sample (Looman & Abracen, 2010). In the Looman and Abracen study, interrater reliability (Intra-Class Correlation [ICC]) for the Static-99 was 0.84, which is consistent with previous studies (e.g., Knight & Thornton, 2007). The revised age item for the Static-99R was calculated based on the offender’s birth date and date of release.

Phallometric assessment

On or shortly after admission to the RTC, each man participated in a phallometric assessment consisting of the assessment of sexual arousal using three separate stimulus sets. The first assessed arousal to adult sexual violence stimuli (Quinsey, Chaplin, & Varney, 1981). The second stimulus set comprised an age/gender assessment that consisted of 21 colored slides depicting adult, pubescent, and prepubescent individuals of each sex, as well as three neutral (scenery) slides. The third stimulus set was a child sexual violence assessment (Quinsey & Chaplin, 1988) that consisted of 22 audiotaped stimuli depicting sexual activity and nonsexual violence between an adult male and a child of either sex. For complete descriptions of the stimuli and procedure, the reader is referred to Looman and Marshall (2001, 2005). For purposes of analysis, for men with offences against adults (aged 16 and older), the adult sexual violence assessment was used, whereas for men with victims below age 16, the more deviant of the child sexual violence assessment or the age gender assessment was used.

All raw peak millimeter-stretch readings were converted to z scores for data analysis (Harris, Rice, Quinsey, Chaplin, & Earls, 1992). Deviance indices were calculated based on these z score data. Harris et al. (2003) suggest that maximal differentiation may be achieved by calculating differential deviance indices in which the mean response to the highest deviant category is subtracted from the mean response to the highest nondeviant category. For purposes of the current study, to ease interpretation, we calculated a deviance differential by subtracting the average of the appropriate stimuli from the average of the deviant stimuli. For the differential indices, positive values indicated a preference for deviant stimuli, values of 0 indicated lack of differentiation, and negative values indicated a preference for appropriate stimuli. Deviance indices are included in Table 1.

In addition, proportional deviance indices (Abel, Barlow, Blanchard, & Guild, 1977) were calculated using untransformed millimeter-stretch data. Proportional indices were included in the analyses based on the consideration of Barbaree and Mewhort’s (1994) analyses, which suggest that z-score transformations may distort the data and that these distortions are not present when raw arousal data, or transformations to percentage full erection, are used. The average arousal to deviant stimuli was divided by the averaged arousal to appropriate stimuli. Thus, values greater than 1.0 indicate a preference for deviant stimuli, an index below 1.0 indicates appropriate preferences, and an index of 1.0 indicates a lack of differentiation.

Following the advice of Harris et al. (1992), no participants were excluded from analyses because of low responding. Harris et al. concluded that omitting low responders does not increase differential validity and excluding such participants would have an effect of decreasing power because of lowering the overall n.

The child sexual violence and age/gender assessments used in the present research have previously been shown to discriminate child molesters from rapists (Looman & Marshall, 2001) in our population, and, for the child violence assessment to be able to discriminate child molesters from nonsexual offenders in the developmental research (Quinsey & Chaplin, 1988). The adult sexual violence assessment has previously been shown (Looman, 2006; Looman & Marshall, 2005) to produce deviance indices in a subsample of the current population that were significantly higher than those found in nonsexual offenders in the developmental research for the stimulus set. In addition, with rapists classified according to the Massachusetts Treatment Center Rapist typology (version 3) (MTC-R3) rapist typology (Knight, 1999), sadistic and vindictive rapists were found to have deviance indices within the deviant range, whereas opportunistic rapists were found to have nondeviant profiles, consistent with their taxometric classifications (Looman, Dickie, & Maillett, 2008).

PCL-R

The PCL-R is perhaps the most widely used rating instrument in the assessment of psychopathy (Hare, 2003). It has been shown to have a high level of reliability as well as construct validity in a wide range of research (Hare, 2003). Scoring of the PCL-R (Hare, 1991, 2003) was completed as part of the pretreatment assessment for offenders in the RTCSOTP. Ratings were made based on both clinical interview and a detailed review of official documentation for all offenders. All raters received training in the administration and scoring of the PCL-R. Previous research examining the interrater reliability for the PCL-R was assessed by comparing ratings made at the RTCSOTP and those completed at Ontario Region’s reception center (Looman, Abracen, & Ismail, 2011). The Looman et al. (2011) sample included 153 men from the current sample. The single-measures ICC for the full-scale PCL-R score was 0.90, p = .001, indicating a high level of agreement. The mean PCL-R score for the sample is displayed in Table 1.

Recidivism

For purposes of the current study, recidivism was defined as a new conviction for another criminal offence following release. All recidivism data were coded according to finger print service records. These data represent a national archive of criminal records collected by the Royal Canadian Mounted Police. Any sexually motivated offence was coded as sexual recidivism (e.g., sexual assault, sexual interference, murder with a sexual component), whereas nonsexually violent offences (e.g., assault, assault causing bodily harm, armed robbery) were coded as violent recidivism, and nonsexual, nonviolent offences (i.e., break and enter, theft, fraud) were coded as general recidivism. If official information (e.g., police reports, parole records) about the new conviction indicated a sexual component (e.g., conviction for assault that was clearly a sexual assault), the offence was coded as both a sexual and violent reoffence.¹ For some analyses, sexual and violent offences were combined for a broader serious recidivism variable. The time-at-risk period was defined as the time from release to first conviction for each of the offence types. Each type of reoffence was considered separately, so that if an offender had both a nonviolent reoffence and a sexual reoffence, each type of offence was coded and time-at-risk calculated for each type of reoffence. Follow-up time was calculated based on the date of the reoffence without taking into account the time the offender may have spent incarcerated for an offence in the intervening period, as we did not have access to information that indicated the actual time served in these cases. It should be noted that Helmus (2009) assessed the effect of using calendar time versus “street” time in calculating recidivism rates and determined that there was no significant effect.

Results

Recidivism

A total of 42 offenders (15.4%) were known to have sexually reoffended during the follow-up period. For the serious recidivism outcome, 101 (37.0%) were known to have reoffended. The average follow-up time for the overall sample was 6.7 (SD = 3.4) years with the range in length of follow-up being from 0.4 to 16.4 years. For sexual recidivism, the average time-at-risk period was 6.0 (SD = 3.4) years, whereas for serious recidivism, the average time-at-risk period was 5.3 (SD = 4.8) years.

Relationship Between Predictor Variables and Recidivism

The relationship between the predictor variables and recidivism was assessed first through correlation coefficients (see Table 2). The Static-99R was significantly related to both sexual and serious recidivism, whereas Psychopathy, for the Total and Factor 2 scores, was only related to serious recidivism. The Factor 1 score approached significance for serious but not sexual recidivism. The proportional deviance index was only related to sexual recidivism, whereas the differential deviance index was not significantly related to either outcome. Given this weak result for the deviance indices and sexual recidivism, further analyses were conducted without the adult rapists in the sample. This was based on the findings of Mann et al. (2010) who noted that sexualized violence has a weaker relationship to sexual recidivism than arousal to children. When the adult rapists were removed from the sample, leaving n = 130, the relationship between deviance and sexual recidivism increased; r = .22, p = .013 for the proportional index. However, for the differential index, the relationship remained nonsignificant, r = .04, p = .65. The r value for the proportional index corresponds with a d value of approximately 0.45 (Rice & Harris, 2005), which exceeds that found for phallometric assessment of sexual interest in children in the Mann et al. (2010) meta-analysis (d = 0.32, 95% confidence interval [CI] = [0.16, 0.47]). For adult rapists, neither deviance index was related to sexual recidivism: proportional index r = .05, p =.57; differential index r = −.05, p = .59. This finding is consistent with the results of Hanson and Morton-Bourgon’s (2004) meta-analysis, which failed to find a significant relationship between sexual interest in rape/violence and sexual recidivism across seven studies. It should be noted, however, that the Mann et al. (2010) meta-analysis added an eighth study and found a significant relationship.

Table 2.

Correlations Between Predictor Variables and Outcome.

	Sexual recidivism	Serious recidivism
Static-99R	.17***	.22****
PCL-R score	.03	.20***
Factor 1	−.01	.10*
Factor 2	.07	.25****
Differential deviance Index	−.01	−.10
Proportional deviance index	.12**	.03

Note: PCL-R = Psychopathy checklist–Revised. N = 273.

p < .10. **p < .05. ***p < .01. ****p < .001.

When assessed via receiver operating characteristic (ROC) area under the curve (AUC), using a fixed 3-year follow-up (n = 232), for the entire sample only the Static-99R was a significant predictor for sexual recidivism, AUC = 0.70 (95% CI = [0.59, 0.82]). For the proportional index, AUC = 0.53 (95% CI = [0.41, 0.64); for the differential index, AUC = 0.45 (95% CI = [0.33, 0.58]); and for the PCL-R, AUC = 0.58 (95% CI = [0.48, 0.67]). For serious recidivism, both the Static-99R and PCL-R scores were significant predictors. For the Static-99R, AUC = 0.63 (95% CI = [0.52, 0.72]) and for the PCL-R, AUC = 0.68 (95% CI = [0.61, 0.76]). In addition, for serious recidivism, the differential deviance index was also a significant predictor but in a negative direction. That is, AUC = 0.40 (95% CI = [0.31, 0.49]) indicated that lower deviance indices are related to serious recidivism. Finally, the proportional deviance index was not a significant predictor of serious recidivism, AUC = 0.44 (95% CI = [0.35, 0.53]).

These analyses were repeated with the adult rapists removed from the sample. For sexual recidivism with the adult rapists excluded (n = 114), the proportional index AUC = 0.53 (95% CI = [0.33, 0.73]) and for the differential index, AUC = 0.48 (95% CI = [0.28, 0.67]); whereas for serious recidivism, the proportional index AUC = 0.62 (95% CI = [0.48, 0.76]) and for the differential index, AUC = 0.49 (95% CI = [0.34, 0.64]).

Psychopathy/Deviance Groupings

Four groups were formed based on scores on the PCL-R and sexual deviance. Men who scored below 25 on the PCL-R were considered to be low on psychopathy, whereas those who scored 25 or higher were considered high. Men whose differential deviance index was 0.0 or higher were considered to be sexually deviant, whereas those whose differential deviance index was below 0.0 were considered nondeviant. Thus, groups were formed of men who had (a) low psychopathy/low deviance, n = 49 (18.0% of the sample); (b) low psychopathy/high deviance, n = 110 (40.4% of the sample); (c) high psychopathy/low deviance, n = 54 (19.9%); and (d) high psychopathy/high deviance, n = 59 (21.7%).

The same types of groupings were formed using the proportional deviance index. Men whose proportional deviance index was 1.0 or higher were considered to be sexually deviant, whereas those whose proportional deviance index was below 1.0 were considered nondeviant. Thus, groups were formed of men who had (a) low psychopathy/low deviance, n = 47 (17.2% of the sample); (b) low psychopathy/high deviance, n = 112 (41.0% of the sample); (c) high psychopathy/low deviance, n = 54 (19.8%); and (d) high psychopathy/high deviance, n = 60 (22.0%).

In terms of offender type, using the groupings based on proportional deviance indices, for the low psychopathy/low deviance (n = 47), 39 (27.3%) were adult rapists, 2 (4.3%) were child molesters, and 2 (4.3%) had teen victims. For the low psychopathy/high deviance group (n = 112), 24 (21.4%) were adult rapists, 50 (44.6%) were child molesters, and 25 (22.3%) had teen victims. For the high psychopathy/low deviance group (n = 54), 48 (88.9%) were adult rapists and 2 (3.7%) were child molesters; there were no offenders with teenage victims in this group. Finally, for the high psychopathy/high deviance group (n = 60), 32 (53.3%) were adult rapists, 20 (33.3%) were child molesters, and 7 (11.7%) had teen victims.

For both sexual and serious recidivism using chi-square analyses, there was no difference in terms of recidivism rates for the psychopathy–sexual deviance groups, using either of the deviance indices. However, using Life Tables Survival analyses, some differences emerged for serious recidivism. Using the differential deviance index and psychopathy groups for the overall analysis, Wilcoxon = 9.59, p = .022. Pairwise comparisons (see Table 3) indicated that the serious recidivism rate differed primarily as a function of psychopathy, with the high psychopathy/low deviance group reoffending at a higher rate than the two low psychopathy groups. The high psychopathy/high deviance group reoffended at an intermediate rate and was not significantly different from any of the other groups.

Table 3.

Sexual and Serious Recidivism Rates for Psychopathy/Deviance Groups.

	Sexual recidivism (n, % reoffend)	Serious recidivism (n, % reoffend)
Low psychopathy/low deviance (n = 49)	6 (12.2)	14 (28.6)^a
Low psychopathy/high deviance (n = 111)	29 (18.9)	37 (33.3)^b
High psychopathy/low deviance (n = 56)	8 (14.3)	27 (48.2)^ab
High psychopathy/high deviance (n = 61)	7 (11.5)	24 (39.3)^ab

Note: For serious recidivism, values with the same superscript differ in survival analyses, p < .05.

Using the proportional index once again with the serious recidivism outcome, differences emerged in survival analysis, Wilcoxon = 9.88, p =.02. Once again, pairwise comparisons (see Table 4) indicated that the serious recidivism rate differed primarily as a function of psychopathy. Also in Table 4, the Static-99R scores for each of the psychopathy–deviance groups are displayed. The two high psychopathy groups have higher Static-99R scores than the low psychopathy–high deviance group.

Table 4.

Sexual and Serious Recidivism Rates for Psychopathy/Deviance Groups Using Proportional Deviance Indices.

	Static-99RScore(SD)	Sexual recidivism (n, % reoffend)	Serious recidivism (n, % reoffend)
Low psychopathy/low deviance (n = 49)	5.3 (1.8)	6 (12.8)	13 (27.7)^a
Low psychopathy/high deviance (n = 111)	4.9 (2.3)^a	21 (18.9)	38 (34.2)^b
High psychopathy/low deviance (n = 56)	5.8 (1.9)^b	8 (14.8)	26 (48.1)^ab
High psychopathy/high deviance (n = 61)	6.2 (2.1)^b	7 (11.7)	23 (39.0)

Note: For serious recidivism, values with the same superscript differ in survival analyses, p < .05. For Static-99R, values with the same superscript differ, p < .05.

The psychopathy–sexual deviance interaction was plotted for both the entire sample and for the rapists (see Figures 1 and 2), for both sexual and serious recidivism, using the process described by Hofmann (2008). Note that for Figure 1, there is no interaction apparent between deviance and psychopathy; however, in Figure 2, for serious recidivism, psychopathy and deviance appear to interact such that rapists who are high on deviance but low on psychopathy reoffend at a higher rate than those high on psychopathy.

Figure 1.

Plots of deviance–psychopathy interaction for the entire sample: (a) sexual recidivism and (b) serious recidivism using the proportional index.

Figure 2.

Plots of deviance–psychopathy interaction for rapists: (a) sexual recidivism and (b) serious recidivism using the proportional index.

Cox regression analyses²

As continuous variables, psychopathy scores, proportional deviance indices, and their interaction term were entered into a Cox regression analysis, to control for the influence of time. For sexual recidivism, no predictors were significant, whereas sexual deviance approached significance. For the final model, χ²(3, N = 272) = 9.64, p = .022 (see top of Table 5 for the final model).

Table 5.

Cox Regression Results for Psychopathy and Sexual Deviance.

	B	SE	Wald	df	p	Exp(B)	CI for E(B)
	B	SE	Wald	df	p	Exp(B)	LL	UL
Sexual recidivism
Sexual deviance	0.364	0.218	2.780	1	.095	1.439	0.94	2.21
PCL-R score	0.035	0.028	1.556	1	.212	1.036	0.98	1.09
Interaction	−0.009	0.010	0.832	1	.362	0.991	0.97	1.01
Serious recidivism
Sexual deviance	0.275	0.198	1.934	1	.164	1.316	0.89	1.94
PCL-R score	0.084	0.020	16.819	1	.000	1.087	1.04	1.13
Interaction	−0.010	0.009	1.177	1	.278	0.990	0.97	1.01

Note: CI = confidence interval; PCL-R = Psychopathy Checklist–Revised; LL = lower limit; UL = upper limit. Only proportional deviance index was used.

Given that none of the individual items were significant whereas the entire model was, it is possible that the predictive power of individual items may be deflated because of multicollinearity. To test this possibility, we calculated “tolerance” and “variance inflation factor (VIF)” values for each predictor (Chen, Ender, Mitchell, & Wells, 2003). Results suggested that the values for each of the variables were within acceptable limits, indicating that multicollinearity is not a concern.

Repeating these analyses for the serious recidivism outcome, psychopathy was the only significant predictor of serious recidivism χ²(3, N = 272) = 22.28, p < .001 (see bottom of Table 5 for the final model).

Static-99R, Psychopathy, and Sexual Deviance

Cox regression analysis was used to examine the extent to which psychopathy, deviance, and the psychopathy–deviance interaction contribute to the prediction of recidivism when the Static-99R has already been considered. For these analyses, psychopathy and deviance were treated as continuous variables. For the entire sample on the first block, the Static-99R score predicted sexual recidivism, χ² (1, N = 272) = 10.09, p < .001, Wald = 10.02 (df = 1), p = .002, E(B) = 1.26. Psychopathy, sexual deviance, and the interaction term all failed to enter the model on the second block with change from the previous block, χ² (3, N = 272) = 3.28, p =.235. See the top of Table 7 for the final model. For serious recidivism, the Static-99R predicted significantly on the first block, χ² (1, N = 272) = 16.46, p < .001, Wald = 16.36 (df = 1), p < .001, E(B) = 1.21. On the second block, the PCL-R score was the only variable that entered the model with change from the first block, χ²(3, N = 272) = 14.01, p = .003. For the total model, χ²(4, N = 272) = 28.73, p < .001, Wald = 10.02 (df = 1), p = .002, E(B) = 1.26. See the bottom of Table 6 for the final model.

Table 6.

Cox Regression Results for the Entire Sample.

	B	SE	Wald	df	Significance	Exp(B)	CI for Exp(B)
	B	SE	Wald	df	Significance	Exp(B)	LL	UL
Sexual recidivism
Static-99R	0.219	0.080	7.70	1	.006	1.25	1.07	1.45
Deviance	0.087	0.120	0.53	1	.466	1.09	0.91	2.04
PCL-R score	−0.004	0.025	0.031	1	.861	0.996	0.96	1.07
Deviance × PCL-R interaction	0.000	0.005	0.003	1	.953	1.00	0.97	1.01
Serious recidivism
Static-99R	0.123	0.05	6.03	1	.014	1.13	1.03	1.25
Deviance	0.089	0.12	0.55	1	.458	1.09	0.86	1.85
PCL-R score	.059	0.018	10.96	1	.001	1.06	1.03	1.12
Deviance × PCL-R interaction	−0.003	0.005	0.220	1	.639	0.997	0.98	1.01

Note: CI = confidence interval; PCL-R = Psychopathy Checklist–Revised; LL = lower limit; UL = upper limit. N = 272. Only proportional deviance index was used.

Offender Groups

For men with victims 12 years of age and younger (i.e., child molesters), there were 73 offenders available for analyses. Of these, 11 (15.0%) reoffended sexually. On the first block, the Static-99R was a significant predictor, χ² (1, N = 72) = 4.47, p = .034, Wald = 4.37 (df = 1), p = .037, E(B) = 1.32. Sexual deviance, psychopathy, and the sexual deviance–psychopathy interaction did not contribute significantly on the second block: change from previous block, χ² (3, 72) = 4.68, p = .196. It is important to note that with the addition of these variables, the relationship between sexual recidivism and the Static-99R was only approaching significance, Wald = 2.88 (df = 1), p = .089.

For serious recidivism, again 73 offenders were available for analyses, with 18 (24.6%) reoffending. Once again, the Static-99R significantly predicted recidivism, χ² (1, 72) = 4.81, p = .028, Wald = 4.68 (df = 1), p = .030, E(B) = 1.28, and the other variables failed to add to the prediction on the second block, with chi-square change for the block being χ²(3, 72) = 1.97, p = .593 (see Table 7).

Table 7.

Cox Regression Results for Child Molesters.

	B	SE	Wald	df	Significance	Exp(B)	CI for Exp(B)
	B	SE	Wald	df	Significance	Exp(B)	LL	UL
Sexual recidivism
Static-99R	0.245	0.144	2.88	1	.089	1.28	0.96	1.67
Deviance	0.224	0.282	0.63	1	.426	1.25	0.72	2.17
PCL-R score	−0.054	0.067	0.67	1	.414	0.95	0.83	1.08
Deviance × PCL-R interaction	−0.003	0.013	0.04	1	.833	0.98	0.97	1.02
Serious recidivism
Static-99R	0.219	0.121	3.25	1	.029	1.24	0.98	1.58
Deviance	0.089	0.291	0.94	1	.759	1.09	0.62	1.93
PCL-R score	0.026	0.057	0.20	1	.653	1.03	0.92	1.15
Deviance × PCL-R interaction	0.000	0.013	0.00	1	.994	1.00	0.97	1.03

Note: CI = confidence interval; PCL-R = Psychopathy Checklist–Revised; LL = lower limit; UL = upper limit. N = 73. Only proportional deviance index was used.

For men with victims 16 years of age and older, the same analyses were conducted. For sexual recidivism, there were 142 offenders available for analyses, with 23 (16.1%) recidivating. Once again, the Static-99R significantly predicted sexual recidivism in the first block, χ² (1, 142) = 8.51, p = .004, Wald = 8.18 (df = 1), p = .004, E(B) = 1.41. As with previous analyses, the deviance–psychopathy interaction failed to enter significantly on the second block. The chi-square change for the block was χ² (3, 142) = 1.27, p = .737 (see top of Table 8 for the complete model).

Table 8.

Cox Regression Results for Adult Rapists.

	B	SE	Wald	df	Significance	Exp(B)	CI for Exp(B)
	B	SE	Wald	df	Significance	Exp(B)	LL	UL
Sexual recidivism
Static-99R	0.334	0.13	6.82	1	.009	1.40	1.09	1.80
Deviance	1.392	1.74	0.64	1	.424	4.02	0.13	122.24
PCL-R score	0.042	0.067	0.39	1	.533	1.04	0.92	1.19
Deviance × PCL-R interaction	−0.036	0.060	0.36	1	.547	0.97	0.86	1.09
Serious recidivism
Static-99R	0.145	0.07	4.07	1	.044	1.16	1.04	1.33
Deviance	2.62	1.09	5.84	1	.016	13.76	1.64	115.51
PCL-R score	0.141	0.045	9.86	1	.002	1.15	1.05	1.26
Deviance × PCL-R interaction	−0.093	0.040	5.39	1	.020	0.91	0.84	0.99

Note: CI = confidence interval; PCL-R = Pyschopathy Checklist–Revised; LL = lower limit; UL = upper limit. N = 142. Only proportional deviance index was used.

Finally, for serious recidivism for men with adult victims, there were 142 offenders available for analysis, and of these, 62 (43.4%) committed a new serious offence. As with previous analyses, the Static-99R significantly predicted recidivism, χ² (1, 142) = 7.06, p = .008, Wald = 6.97 (df = 1), p = .008, E(B) = 1.19. On the second block, PCL-R, deviance, and the deviance–psychopathy interaction term all entered the model with change from the previous block, χ² (3, 142) = 12.10, p = .007. See bottom of Table 8 for the complete model.

Given that an interaction effect was found for serious recidivism but not sexual recidivism for the rapists, an additional analysis was completed examining violent, nonsexual recidivism. For the Static-99R on the first step, the relationship approached significance, χ² (1, 142) = 3.70, p = .054, Wald = 3.68 (df = 1), p = .055, E(B) = 1.51. On the second step, sexual deviance, psychopathy, and the interaction term all added significantly to the prediction with change from the previous block, χ² (3, 142) = 6.14, p = .043, with the overall model, χ² (4, 142) = 10.74, p = .030 (see Table 9 for the final model). This result suggests that the finding of a significant effect for psychopathy and deviance for serious recidivism is related to the violent nonsexual component of serious recidivism rather than the sexual component.

Table 9.

Cox Regression for Adult Rapist for Violent Nonsexual Recidivism.

	B	SE	Wald	df	Significance	Exp(B)	95.0% CI for Exp(B)
	B	SE	Wald	df	Significance	Exp(B)	Lower	Upper
Static-99R	0.112	0.078	2.075	1	.150	1.119	0.960	1.304
Deviance	2.418	1.102	4.815	1	.028	11.220	1.295	97.243
PCL-R Score	0.125	0.047	6.923	1	.009	1.133	1.032	1.243
Interaction	−0.084	0.041	4.180	1	.041	0.920	0.849	0.997

Note: CI = confidence interval; PCL-R = Psychopathy Checklist–Revised. Only proportional deviance index was used.

Risk Percentages for the Combination of Static-99 and PCL-R Scores

The use of logistic regression allows the calculation of risk percentages based on β values derived from the analyses. Using a fixed 3-year follow-up, sexual and serious reoffence probabilities were calculated, using formulae provided in Hosmer and Lemeshow (2000), for the entire sample using Static-99R scores of 3, 5, and 7. Probabilities were also calculated for the combination of the Static-99R at each of those scores and a PCL-R score of 15, 25, and 30 for both low (i.e., proportional index of 0.500) and higher (i.e., proportional index of 1.500) levels of sexual deviance (see Tables 11). As can be seen from Table 10, the inclusion of the PCL-R and sexual deviance did not change the probability of reoffence from the estimate derived from the Static-99R alone, regardless of PCL-R score and level of deviance. However, the inclusion of these additional factors did alter the estimates for serious reoffence. Examination of Table 11 reveals that for a PCL-R score of 15, across all levels of Static-99R score, the estimated serious recidivism rate was lower than that obtained based on the Static-99R alone. This was true for both low and high deviance. For PCL-R scores of 30, the predicted recidivism rate was higher than that based on the Static-99R alone. Again this was true for both low and high deviance. The CIs for the estimates related to the PCL-R of 15 did not overlap with the estimates for PCL-R of 25, indicating that these were significantly different. However, for PCL-R scores of 25 and 30, there was overlap in the CIs.

Table 10.

Sexual Reoffence Probabilities Based on Logistic Regression Analyses for Static-99R and Static-99R + PCL-R/Deviance.

	Static-99R				Static-99R + PCL-R/deviance

Proportional deviance index 0.500
	Probability of reoffence	95% CI			Probability of reoffence	95% CI
	Probability of reoffence	LL	UL		Probability of reoffence	LL	UL
Static-99R 3	.040	0.038	0.043	PCL-R 15	.033	0.020	0.053
				PCL-R 25	.040	0.026	0.060
				PCL-R 30	.044	0.026	0.074
Static-99R 5	.075	0.068	0.084	PCL-R 15	.058	0.038	0.086
				PCL-R 25	.070	0.053	0.091
				PCL-R 30	.077	0.051	0.113
Static-99R 7	.137	0.116	0.162	PCL-R 15	.099	0.060	0.159
				PCL-R 25	.120	0.086	0.162
				PCL-R 30	.130	0.090	0.185
Proportional deviance index 1.500
Static-99R 3	.041	0.037	0.043	PCL-R 15	.037	0.016	0.081
				PCL-R 25	.045	0.021	0.092
				PCL-R 30	.049	0.020	0.114
Static-99R 5	.075	0.067	0.085	PCL-R 15	.064	0.027	0.146
				PCL-R 25	.078	0.034	0.168
				PCL-R 30	.086	0.030	0.224
Static-99R 7	.137	0.113	0.165	PCL-R 15	.111	0.040	0.267
				PCL-R 25	.132	0.048	0.32
				PCL-R 30	.145	0.047	0.369

Note: PCL-R = Psychopathy Checklist–Revised; CI = confidence interval; LL = lower limit; UL = upper limit.

Table 11.

Serious Reoffence Probabilities Based on Logistic Regression Analyses for Static-99R and Static-99R + PCL-R/Deviance.

	Static-99R				Static-99R + PCL-R/deviance

Proportional deviance index 0.500
	Probability of reoffence	95% CI			Probability of reoffence	95% CI
	Probability of reoffence	LL	UL		Probability of reoffence	LL	UL
Static-99R 3	.145	0.123	0.171	PCL-R 15	.098	0.078	0.123
				PCL-R 25	.199	0.174	0.228
				PCL-R 30	.273	0.231	0.320
Static-99R 5	.205	0.168	0.248	PCL-R 15	.125	0.098	0.158
				PCL-R 25	.247	0.217	0.278
				PCL-R 30	.331	0.278	0.389
Static-99R 7	.282	0.223	0.349	PCL-R 15	.158	0.116	0.212
				PCL-R 25	.301	0.248	0.360
				PCL-R 30	.395	0.327	0.467
Proportional deviance index 1.500
Static-99R 3	.145	0.125	0.168	PCL-R 15	.0882	0.071	0.109
				PCL-R 25	.181	0.159	0.206
				PCL-R 30	.251	0.212	0.295
Static-99R 5	.205	0.171	0.244	PCL-R 15	.113	0.083	0.152
				PCL-R 25	.226	0.176	0.284
				PCL-R 30	.306	0.208	0.425
Static-99R 7	.282	0.229	0.342	PCL-R 15	.144	0.094	0.212
				PCL-R 25	.277	0.179	0.403
				PCL-R 30	.367	0.233	0.526

Note: PCL-R = Psychopathy Checklist–Revised; CI = confidence interval; LL = lower limit; UL = upper limit.

Discussion

The current analyses explored the issue of whether consideration of factors outside the Static-99R is necessary for the prediction of recidivism in sexual offenders. Specifically, we explored the contribution of psychopathy and sexual deviance. Previous research has suggested that these factors may be potent predictors both on their own and in combination; however, no one has yet explored whether they add to the prediction of recidivism when actuarially determined risk has already been considered.

Prior to consideration of the Static-99R score, when considering individual predictors for the current sample, sexual deviance, as represented by the proportional deviance index, was significantly related to sexual recidivism, but not other serious recidivism, whereas the opposite was true for psychopathy. When only the child molesters were used in the analyses, the d value for sexual deviance in the prediction of sexual recidivism was the same as that reported in previous meta-analyses (e.g., Mann et al., 2010); however, sexual deviance was not a significant predictor for the adult rapists. When groups were formed based on psychopathy and sexual deviance, they did not differ significantly in terms of their rate of sexual recidivism, whereas differences in rates of serious recidivism were primarily accounted for by the effect of psychopathy.

Thus, results of previous research regarding the prediction of sexual or serious recidivism with the combination of sexual deviance and psychopathy were partially replicated in the current research. The high psychopathy groups differed from the low psychopathy groups in serious recidivism rates, a finding that replicated that of Harris et al. (2003); however, the high psychopathy/high sexual deviance group did not reoffend at a higher rate than the high psychopathy/low sexual deviance group. Thus, the current results add to the already mixed results of the previous research. It should be noted that for the outcome of serious recidivism, sexual deviance actually had an ameliorative effect. That is, an interaction was found that indicated that men who scored high on psychopathy and were not sexually deviant had a serious reoffence rate that was higher than for sexually deviant men. Results suggested that this effect was accounted for by higher rates of nonsexual violent recidivism in the high psychopathy group.

One possible explanation for the limited predictive validity of the phallometric results in the current sample is the fact that more than 50% of the sample consisted of adult rapists and only 27.2% were child molesters. As noted above, the predictive validity for sexual deviance improved greatly when only child molesters were included in the analyses. Previous studies that found a relationship between deviance and recidivism had a higher proportion of child molesters in the sample. For example, both Rice and Harris (1997) and Harris et al. (2003) had approximately 49% child molesters. Serin et al. (2001) had a sample made up of 33 rapists and 35 child molesters, whereas 65.6% of the Gretton et al. (2001) sample had victims below the age of 12.

In addition, sexual deviance was one of the admission criteria for the RTCSOTP; thus, the current sample is overselected for sexual deviance. This fact may lead to questions regarding the representativeness of the current sample. However, this aspect of the sample also makes it more likely to be representative of high–risk/needs samples for which the potential interaction of psychopathy and sexual deviance is most relevant, such as SVP candidates in the United States (Wilson, Looman, Abracen, & Pake, 2010) or Dangerous Offender candidates in Canada. Thus, although further replication of the current results is required, it is reasonable to conclude based on the current findings that in high-risk/needs samples, the utility of sexual deviance as a predictor of recidivism is likely to be more limited than it is in more broadly representative samples.

A finding of particular interest is that the high psychopathy and high deviance group had a mean Static-99R score of more than 6, but their sexual reoffence rate was below 12% over a 6-year period (Table 3). Doren (2010) has demonstrated that 20-year reoffence rates can be accurately estimated by doubling the 5-year rate. Thus, it is reasonable to assume that for this presumably high-risk group, the long-term sexual reoffence rate will be around 25%. Such a rate of reoffence clearly indicates that elevating the risk estimate of an offender with these characteristics beyond that suggested by the Static-99R alone would be inappropriate. For the highest risk group, the serious reoffence rate may be assumed to exceed 50% over the long term (see bottom of Table 11); however, the current analyses indicate that this is primarily because of the influence of nonsexual recidivism in the outcome.

The finding regarding the lack of predictive validity for the differential deviance index in the current study is of interest in that Harris et al. (1992) advocate the use of z-score transformations of phallometric data to maximize discriminant validity. However, prior to this study, there has not been a comparison of transformed and nontransformed data in terms of predicting recidivism. Barbaree and Mewhort (1994) have argued that z-score transformations may distort the information available in raw phallometric data. It may be that this distortion has an effect on the ability of z-score-derived deviance indices to predict recidivism. To reiterate the findings of the current research, in univariate correlational analyses, only the proportional index was significantly related to sexual recidivism, for both the entire sample and child molesters. The differential deviance index was related to violent recidivism but in a negative direction so that deviance was related to a lower probability of nonsexual violence. In multivariate analyses, neither of the indices contributed to prediction of recidivism. It should be noted that in terms of previous research, despite Harris et al’s. (1992) recommendations, the differential index does not appear to be the index most frequently used in the literature. In both Serin et al. (2001) and Gretton et al. (2001), proportional indices were used. Although Harris et al. (2003) calculated a differential index in the same manner as was done in the current study, Rice and Harris (1997) did not specify how they calculated the index, simply stating that “If any phallometric test had indicated an absolute preference for deviant stimuli (children, rape cues, or nonsexual violence cues), the subject was declared deviant” (p. 236). Clearly, further research is needed to clarify the effect of method of summarizing phallometric assessments on prediction of recidivism.

The Static-99R predicted both sexual and serious recidivism in the current sample. When psychopathy and sexual deviance were added in the Cox regression analysis, these additional variables did not add predictive value for either outcome when the complete sample was used, nor for the child molesters examined separately. To some extent, these results should not be particularly surprising, given that the Static-99 (and by extension the Static-99R) is considered to contain indicators of both antisociality and sexual deviance (e.g., Roberts, Doren, & Thornton, 2002).

When the rapists were examined separately, psychopathy and sexual deviance, as well as their interaction term, contributed significantly to the prediction of serious recidivism. When this finding was further explored by analyzing results for violent nonsexual recidivism, the same pattern of results was found. Given that no significant relationship was found for sexual recidivism alone, this significant result for violent recidivism suggests that the relationship found for serious recidivism is related to the violent, rather than the sexual component of the serious recidivism outcome. This result corresponds to the findings of Gretton et al. (2001) and Harris et al. (2003), both of whom found that the psychopathy/deviance combination predicted violent, but not sexual, recidivism.

Calculation of expected sexual recidivism rates based on logistic regression illustrates that sexual deviance and psychopathy do not add to the information provided by the Static-99R score alone (see Table 10). However, for serious recidivism, when the PCL-R score is in the low range (i.e., 15; Hare, 2003), the expected recidivism rate was lower for the combination of Static-99R and PCL-R/deviance than based on the Static-99R alone. For a PCL-R score at a commonly used cut-off for psychopathy (i.e., a score of 25), the expected recidivism rate was the same as that obtained with the Static-99R alone, regardless of the level of sexual deviance. However, when a PCL-R score of 30 was used, the expected recidivism rate was significantly higher than when using the Static-99R score alone (i.e., 39.5% vs. 28.2%, respectively, for the low-deviance estimate). These results suggest that when serious recidivism is the outcome of interest, adding the PCL-R score provides valuable information, at both ends of the score range. However, the consideration of phallometric deviance did not seem to influence the estimate, despite the significant result in the Cox regression analyses.

Harris et al. (2003) and Rice and Harris (1997) report significant main effects for psychopathy with sexual recidivism, as have others (see Hanson & Morton-Bourgon, 2004). It should be noted, however, that other researchers have found that psychopathy is not related to sexual recidivism (e.g., Murrie et al., 2011) and that psychopathy plays a role in explaining behavior in only a minority of sex offences and that most sex offenders, including those who recidivate, do not score high on the PCL-R (Saleh, Malin, Grudzinskas, & Vitacco, 2010). These latter authors conclude that the PCL-R provides only minimal information in the assessment of risk for sexual reoffence, especially for child molesters.

This conclusion is consistent with the current research in that psychopathy predicted serious but not sexual recidivism. This result confirms the relationship between psychopathy and violence that we have found in previous research with the current sample (e.g., Looman, Abracen, Serin, & Marquis, 2005). As well, the current results are consistent with the result of our analysis concerning risk assessment measures (Looman & Abracen, 2010), which failed to find a significant effect for the General Criminality subscale of the Static-2002 (Hanson & Thornton, 2003) in the prediction of sexual recidivism.

Conclusion

Current results indicate that the consideration of sexual deviance, psychopathy, and their interaction term is not necessary when the Static-99R score is known and that these additional variables do not contribute significantly to predicting that outcome once the Static-99R score has been accounted for. Thus, the current results do not support the modification of risk estimates based on the Static-99R because of the presence of sexual deviance and psychopathy when sexual recidivism is the outcome of interest.

Limitations

A possible limitation of the current study is that the majority of the child molesters in the current sample had deviant arousal, possibly limiting the predictive validity of the sexual deviance variable, especially when dichotomized. A possible explanation for this finding is that sexual deviance has been one of the selection criteria for the High Intensity Sex Offender Program. It should be noted, however, that when treated as a continuous variable, the proportional deviance index, when considered alone, did in fact predict sexual recidivism. As well, the high proportion of adult rapists in the current sample may also have limited the predictive validity of sexual deviance for the sample as a whole.

Another possible limitation of the current research is that the majority of the current sample had participated in treatment. This is especially a concern for the sexual deviance variable, given that deviant sexual interests are specifically targeted in sexual offender treatment. However, as noted in the preceding paragraph, the proportional deviance index was found to predict sexual recidivism. Nonetheless, it is unknown to what extent this treatment participation may have affected the relationships among the variables.

Finally, it should be noted that the current sample is a preselected high-risk/need sample; thus, it is possible that the current results only apply to this sample. Replication is needed in other, more typical sexual offender samples. That said, it should be noted that these results may be particularly relevant for offenders being considered for civil commitment in the United States or Dangerous Offender designation in Canada, as these too are high-risk/needs samples where considerations such as psychopathy and sexual deviance are relevant.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

References

Abel

G. G.

Barlow

D. H.

Blanchard

E. B.

Guild

(1977). The components of rapists’ sexual arousal. Archives of General Psychiatry, 34, 895-903.

Abracen

Looman

Langton

C. M.

(2008). Treatment of sexual offenders with psychopathic traits: Recent research developments and clinical implications. Trauma, Violence, & Abuse, 9, 144-166. doi:10.1177/1524838008319633

Barbaree

H. E.

Langton

C. M.

Peacock

E. J.

(2006). The factor structure of static actuarial items: Its relation to prediction. Sexual Abuse: A Journal of Research and Treatment, 18, 207-226. doi:10.1007/s11194-006-9011-6

Barbaree

H. E.

Mewhort

D. J. K.

(1994). The effects of the z-score transformation on measures of relative erectile response strength: A re-appraisal. Behavior Research and Therapy, 32, 547-558.

Barbaree

H. E.

Seto

M. C.

Langton

C. M.

Peacock

E. J.

(2001). Evaluating the predictive accuracy of six risk assessment instruments for adult sex offenders. Criminal Justice and Behavior, 28, 490-521. doi:10.1177/009385480102800406

Boer

D. P.

Hart

S. D.

Kropp.

P. R.

Webster

C. D.

(1997). Manual for the sexual violence risk -20. Vancouver, BC: The British Columbia Institute Against Family Violence, Simon Fraser University.

Chen

Ender

Mitchell

Wells

(2003). Regression with SPSS. Retrieved from http://www.ats.ucla.edu/stat/spss/webbooks/reg/default.htm

Doren

D. M.

(2002). Evaluating sex offenders: A manual for civil commitments and beyond. Thousand Oaks, CA: Sage.

Doren

D. M.

(2010). Empirically based recidivism risk assessment estimate extrapolations across time and outcome measure. In Sclank

(Ed.), The sexual predator: Legal issues, assessment, treatment (Vol. 4, pp. 18-1-18-14). Kingston, NJ: Civic Research Institute.

10.

Gretton

H. M.

McBride

Hare

R. D.

O’Shaughnessy

Kumka

(2001). Psychopathy and recidivism in adolescent sex offenders. Criminal Justice and Behavior, 28, 427-449.

11.

Hanson

R. K.

Morton-Bourgon

K. E.

(2004). Predictors of sexual recidivism: An updated meta-analysis (Research Report. No. 2004–02). Ottawa: Public Safety and Emergency Preparedness Canada.

12.

Hanson

R. K.

Morton-Bourgon

K. E.

(2005). The characteristics of persistent sexual offenders: A meta-analysis of recidivism studies. Journal of Consulting and Clinical Psychology, 73, 1154-1163. doi:10.1037/0022-006X.73.6.1154

13.

Hanson

R. K.

Morton-Bourgon

K. E.

(2009).The accuracy of recidivism risk assessments for sexual offenders: A meta-analysis of 118 prediction studies. Psychological Assessment, 21, 1-21. doi:10.1037/a0014421

14.

Hanson

R. K.

Phenix

Helmus

(2009). Static-99(R) and Static-2002(R): How to interpret and report in light of recent research. Preconference workshop at the 28th Annual Research and Treatment Conference of the Association for the Treatment of Sexual Abusers, Dallas, TX.

15.

Hanson

R. K.

Thornton

(2000). Improving risk assessments for sex offenders: A comparison of three actuarial Scales. Law and Human Behavior, 24, 119-136.

16.

Hanson

R. K.

Thornton

(2003). Notes on the development of Static-2002 (User Report 2003-01). Ottawa, Ontario: Department of the Solicitor General of Canada.

17.

Hare

R. D.

(1991). Manual for the revised psychopathy checklist. Toronto, Canada: Multi-Health Systems.

18.

Hare

R. D.

(1999). Psychopathy as a risk factor for violence. Psychiatric Quarterly, 70, 181-197.

19.

Hare

R. D.

(2003). Hare Psychopathy Checklist–Revised (PCL-R)–2nd ed. Technical manual. Toronto, Canada: Multi-Health Systems.

20.

Harris

G. T.

Rice

M. E.

Quinsey

V. L.

Chaplin

T. C.

Earls

(1992). Maximizing the discriminate validity of phallometric assessment data. Psychological Assessment, 4, 502-511.

21.

Harris

G. T.

Rice

M. E.

Quinsey

V. L.

Lalumière

M. L.

Boer

Lang

(2003). A multisite comparison of actuarial risk instruments for sex offenders. Psychological Assessment, 3, 413-425. doi:10.1037/1040-3590.15.3.413

22.

Helmus

(2009). Re-norming Static-99 recidivism estimates: Exploring base rate variability across sex offender samples. Unpublished master’s thesis. Carleton University, Ottawa, Ontario, Canada.

23.

Helmus

Thornton

Hanson

R. K.

Babchishin

K. M.

(2011). Improving the predictive accuracy of static-99 and static-2002 with older sex offenders: Revised age weights. Sexual Abuse: A Journal of Research and Treatment. Advance online publication. doi:10.1177/1079063211409951

24.

Hildebrand

de Ruiter

de Vogel

(2004). Psychopathy and sexual deviance in treated rapists: Association with (sexual) recidivism. Sexual Abuse: A Journal of Research and Treatment, 16, 1-24. doi:10.1177/107906320401600101

25.

Hofmann

(2008). Workshop: Moderated regression analysis. Retrieved from www.faculty.chicagobooth.edu/./Workshop%20moderated%20regression.ppt

26.

Hosmer

D. W.

Lemeshow

(2000). Applied logistic regression (2nd ed.). New York, NY: John Wiley.

27.

Jackson

R. L.

Richards

H. J.

(2007). Diagnostic and risk profiles among civilly committed sex offenders in Washington state. International Journal of Offender Therapy and Comparative Criminology, 51, 313-323.

28.

Knight

R. A.

(1999). Validation of a typology for rapists. Journal of Interpersonal Violence, 14, 303-330.

29.

Knight

R. A.

Sims-Knight

J. E.

(2003). The developmental antecedents of sexual coercion against women: Testing alternative hypotheses with structural equation modeling. Annals of the New York Academy of Sciences, 989, 72-85.

30.

Knight

R. A.

Thornton

(2007). Evaluating and improving risk assessment schemes for sexual recidivism: A long-term follow-up of convicted sexual offenders. Washington, DC: U.S. Department of Justice. Retrieved from http://www.ncjrs.gov/pdffiles1/nij/grants/217618.pdf

31.

Langström

Grann

(2000). Risk for criminal recidivism among young sex offenders. Journal of Interpersonal Violence, 15, 855-871. doi:10.1177/088626000015008005

32.

Lloyd

C. D.

Clark

H. J.

Forth

A. E.

(2010). Psychopathy, expert testimony, and indeterminate sentences: Exploring the relationship between Psychopathy Checklist–Revised testimony and trial outcome in Canada. Legal and Criminological Psychology, 15, 323-339. doi:10.1348/135532509X468432

33.

Looman

(2006). Correction to Looman and Marshall 2005. Criminal Justice and Behavior, 33, 565-567. doi:10.1177/0093854806288785

34.

Looman

Abracen

(2010). Comparison of measures of risk for recidivism in sexual recidivism. Journal of Interpersonal Violence, 25, 791-807. doi:10.1177/0886260509336961

35.

Looman

Abracen

Ismail

(2011, June). Inter-rater reliability of the PCL-R in an applied setting. Paper presented at the 2nd Annual North American Correctional and Criminal Justice Psychology Conference, Toronto, Ontario, Canada.

36.

Looman

Abracen

Nicholaichuk

T. P.

(2000). Recidivism among treated sexual offenders and matched controls: Data from the Regional Treatment Centre (Ontario). Journal of Interpersonal Violence, 15, 279-290.

37.

Looman

Abracen

Serin

Marquis

(2005). Psychopathy, treatment change and recidivism in high risk high need sexual offenders. Journal of Interpersonal Violence, 20, 549-568. doi:10.1177/0886260504271583

38.

Looman

Dickie

Maillett

(2008). Sexual arousal among rapist subtypes. Journal of Sexual Aggression, 14, 267-279. doi:10.1080/13552600802348864

39.

Looman

Marshall

(2001). Phallometric assessments designed to detect arousal to children: The responses of rapists and child molesters. Sexual Abuse: A Journal of Research and Treatment, 13, 3-13. doi:10.1177/107906320101300102

40.

Looman

Marshall

W. L.

(2005). Sexual arousal in rapists. Criminal Justice and Behavior, 32, 367-389. doi:10.1177/0093854805276402

41.

Malamuth

N. M.

(2003). Criminal and noncriminal sexual aggressors: Integrating psychopathy in a hierarchical–meditational confluence model. Annals of the New York Academy of Sciences, 989, 33-58.

42.

Mann

R. E.

Hanson

R. K.

Thornton

(2010). Assessing risk for sexual recidivism: Some proposals on the nature of psychologically meaningful risk factors. Sexual Abuse: A Journal of Research and Treatment, 22, 191-217. doi:10.1177/1079063210366039

43.

Murrie

D. C.

Boccaccini

M. T.

Caperton

Rufino

(2011). Field validity of the Psychopathy Checklist–Revised in sex offender risk assessment. Psychological Assessment. Advance online publication. doi: 10.1037/a0026015

44.

Olver

M. E.

Wong

S. C. P.

(2006). Psychopathy, sexual deviance, and recidivism among sex offenders. Sexual Abuse: A Journal of Research and Treatment, 18, 65-82. doi:10.1007/s11194-006-9006-3

45.

Quinsey

V. L.

Chaplin

T. C.

(1988). Penile response of child molesters and normals to descriptions of encounters with children involving sex and violence. Journal of Interpersonal Violence, 3, 259-274.

46.

Quinsey

V. L.

Chaplin

T. C.

Varney

(1981). A comparison of rapists and non-sex offenders sexual preferences for mutually consenting sex, rape and physical abuse of women. Behavioral Assessment, 3, 137-135.

47.

Rice

M. E.

Harris

G. T.

(1997). Cross-validation and extension of the violence risk appraisal guide for child molesters and rapists. Law and Human Behavior, 21, 231-241.

48.

Rice

M. E.

Harris

G. T.

(2005). Comparing effect sizes in follow-up studies: ROC area, Cohen’s d, and r. Law and Human Behavior, 29, 615-620. doi:10.1007/s10979-005-6832-7

49.

Roberts

C. F.

Doren

D. M.

Thornton

(2002). Dimensions associated with assessments of sex offender recidivism risk. Criminal Justice and Behavior, 29, 569-589.

50.

Saleh

F. M.

Malin

H. M.

Grudzinskas

A. J.

Vitacco

M. J.

(2010). Paraphilias with co-morbid psychopathy: The clinical and legal significance to sex offender assessments. Behavioral Sciences & the Law, 28, 211-223. doi:10.1002/bsl.933

51.

Serin

R. C.

Mailloux

D. L.

Malcolm

P. B.

(2001). Psychopathy, deviant sexual arousal, and recidivism among sex offenders. Journal of Interpersonal Violence, 16, 234-246.

52.

Seto

M. C.

Harris

G. T.

Rice

M. E.

Barbaree

H. E.

(2004). The Screening Scale for pedophilic interests predicts recidivism among adult sex offenders with child victims. Archives of Sexual Behavior, 33, 455-466.

53.

Wilson

R. J.

Looman

Abracen

J. A.

Pake

D. R.

(2010, October). Comparing sexual offenders at the Regional Treatment Centre and the Florida Civil Commitment Center. Paper presented at the 29th annual meeting of the Association for the Treatment of Sexual Abusers, Phoenix, AZ.

54.

Wong

Olver

M. E.

Nicholaichuk

T. P.

Gordon

(2004). The Violence Risk Scale—Sexual Offender version (VRS-SO). Saskatoon, Saskatchewan, Canada: Regional Psychiatric Centre and University of Saskatchewan.

Does Consideration of Psychopathy and Sexual Deviance Add to the Predictive Validity of the Static-99R?

Abstract

Keywords

Psychopathy

Sexual Deviance

Psychopathy and Sexual Deviance

Proxy Measures of Sexual Deviance

Actuarial Assessment of Risk

Method

Sample

Measures

Static-99R

Phallometric assessment

PCL-R

Recidivism

Results

Recidivism

Relationship Between Predictor Variables and Recidivism

Psychopathy/Deviance Groupings

Cox regression analyses 2

Static-99R, Psychopathy, and Sexual Deviance

Offender Groups

Risk Percentages for the Combination of Static-99 and PCL-R Scores

Discussion

Conclusion

Limitations

Footnotes

Declaration of Conflicting Interests

Funding

Notes

References

Cox regression analyses²