Despite their widespread use in forensic and correctional practice, surprisingly little research investigates how well actuarial risk assessment instruments (ARAIs) for sexual offenders work within the contexts where they are routinely applied. We examined the predictive validity (M = 4.77 years follow-up) of the two most widely used ARAIs for sexual offenders, the STATIC-99 and Minnesota Sex Offender Sex Offender Screening Tool–Revised (MnSOST-R), as administered in routine practice among 1,928 offenders screened for possible civil commitment as sexually violent predators. Effect sizes for both ARAIs were lower than in most published research and meta-analytic reviews, although the STATIC-99 was a more consistent predictor of recidivism than the MnSOST-R. Recidivism rates for the STATIC-99 were much closer to those expected based on the 2009 norms than the 2003 norms. Offender characteristics (e.g., age at release, prior arrests, release type) were often as or more effective than ARAIs for predicting recidivism. This study, apparently the largest cross-validation study of popular ARAIs for sex offenders, suggests that the predictive validity of these measures in routine practice in the United States may be poorer than often assumed.

This study examined the utility of the Personality Assessment Inventory (PAI) to identify prison inmates in a mandatory sex offender treatment program prone to engage in institutional misconduct. Archival PAI and institutional disciplinary data were coded for 137 inmates in treatment for an average of 1.59 years. The Antisocial Features scale predicted various forms of general and major infractions (e.g., verbal aggression), with no other scales providing any incremental validity beyond this measure. The Treatment Rejection scale was uniquely but modestly correlated (r=.14) with treatment noncompliance, even though such infractions were rare in this sample.

Several studies have concluded that scores from Hare's (2003) Psychopathy Checklist-Revised (PCL-R) predict reoffense among sexual offenders, but most of those studies examined the predictive validity of scores from trained research staff, not clinicians in the field scoring the measure as part of actual forensic assessments. Therefore, we examined the field validity of PCL-R scores that forensic evaluators assigned to 333 male sexual offenders who underwent evaluations during a civil commitment selection process. Overall, no PCL-R score was a significant predictor of sexually violent recidivism. Facet 4 was the only PCL-R score with an area under the curve (AUC) greater than .50 (AUC = .53, p = .85) and the only PCL-R score that approached statistical significance for predicting the combined category of violent or sexually violent offending (AUC = .63, p = .08). However, scores from a subset of evaluators revealed stronger predictive effects, indicating that predictive validity was higher for scores from some evaluators than others. Overall, these results suggest that the stronger predictive validity values in controlled research studies may not apply to all evaluators when the PCL-R is administered in the field.

The Static-99 is the most widely used sex offender risk assessment measure in the world, but predictive effects for the Static-99 have been relatively low in several U.S. samples, particularly in states with high levels of racial or ethnic diversity. The current study uses data from one of the largest predictive validity studies of Static-99 in the U.S. to examine whether Static-99 and Static-99R scores predict recidivism similarly for White (n = 912), Black (n = 411), and Latino (n = 588) offenders. Although predictive effects with both measures were, at times, large enough to reach statistical significance for White and Black offenders, they were never large enough to reach significance for Latino offenders. On the Static-99R, the measure recommended for use in practice, AUC values ranged from .62 to .69 (d = .43 to .71) for Black offenders, .59 to .65 (d = .33 to .56) for White offenders, but only .56 to .58 (d = .17 to .33) for Latino offenders. Findings have implications for fairness in testing and highlight the need for continued research regarding the potentially moderating role of offender race/ethnicity in risk research.

Poisoned Partner EffectIn the early 1980s, a number of studies determined that the presence of a sick "demonstrator" rat could produce a taste aversion in a healthy "observer" (e.g