One hundred ninety-two convicted extrafamilial child molesters were followed for an average of 7.8 years after their conviction. The percentage of men who had committed a sexual, a violent, or any criminal offense by the 12th year was 15.1, 20.3, and 41.6, respectively. The sexual recidivists, compared with the nonrecidivists; demonstrated more problems with alcohol and showed greater sexual arousal to assaultive stimuli involving children than to mutually consenting stimuli with children. The violent recidivists, compared with the nonrecidivists, were more likely to have a history of violence in the families in which they were raised and were rated significantly more psychopathic on the Psychopathy Checklist--Revised (PCL-R). They also showed more sexual arousal to stimuli depicting mutually consenting sexual interactions with children than to adult stimuli. In terms of any criminal recidivism, recidivists were younger, had completed fewer years of school, and were raised in psychologically more harmful family environments compared with nonrecidivists. They also reported that, before 16 years of age, they were more likely to have been physically abused and were more likely to have been removed from their homes compared to those that did not recidivate. In addition, recidivists demonstrated more general hostility on the Buss-Durkee Hostility Inventory and were rated significantly more psychopathic on the PCL-R. The phallometric assessments revealed, that the criminal recidivists, compared to the nonrecidivists, showed more sexual arousal to stimuli depicting coercive sexual activity with children than consenting sexual activities with children. In addition, they showed more sexual arousal to scenes depicting adult rape then adult mutually consenting sex. Finally, the recidivists also had more charges or convictions for violence and any criminal acts. The small number of significant differences between recidivists and nonrecidivists in the sexual and violent categories precluded an attempt to determine which combination of factors meaningfully predicted reoffending. However, for criminal recidivism, a stepwise discriminant function analysis to assess the combination of factors that most successfully distinguished between groups in terms of criminal recidivism was significant, with subjects' age, total number of criminal convictions, and pedophile assault index being retained for optimal prediction. The procedure correctly classified 70.6% of the original group, 82.8% of the nonrecidivists, and 52.6% of the recidivists. PCL-R Total Score alone was equally successful in a similar discriminant function.