With the exception of recidivism data, the study variables were coded retrospectively from institutional files. The files generally contained psychiatric and psychological evaluations, police records, criminal history, treatment plans and evaluations, the hospital's annual advice to the court about the need for prolonging the TBS order, and family background data. Psychopathy was measured by using the PCL-R, which is a reliable and well-validated 20-item structured clinical assessment instrument. Participants were assigned a rating on each of the PCL-R items, measuring characteristics such as glibness/superficial charm, lack of empathy, need for stimulation/proneness to boredom, poor behavioral controls, impulsivity, and juvenile delinquency. The determination of sexual deviance was based on the person's documented history of offending and/or the patient's acknowledgement of the sexual deviance during treatment. Any diagnosed paraphilia constituted sexual deviance in the study, ranging from exhibitionism to sadism to pedophilia to more atypical paraphilias. Data on recidivism were obtained from the Judicial Documentation Register of the Dutch Ministry of Justice. The follow-up period after release ranged up to 23.5 years, with the medium period being 11.8 years. Base rates for sexual, violent nonsexual, violent (including sexual), and general recidivism were 34 percent, 47 percent, 55 percent, and 73 percent, respectively. For all types of offending measured, offenders who scored high on the PCL-R were significantly more often reconvicted than other offenders. The sexual deviance score was found to be a significant predictor of sexual reconviction. Survival analyses yielded considerable evidence that psychopathic sex offenders with sexually deviant preferences were at substantially greater risk of committing new sexual offenses than psychopathic offenders without deviant preferences or nonpsychopathic offenders with or without sexual deviance. These findings are discussed in terms of their practical and clinical implications. 3 tables, 1 figure, and 68 references