Preliminary findings from a longitudinal outcome study of sex offender treatment are presented. Ss included 76 child molesters who were treated in a comprehensive relapse prevention program, and 79 molesters who were randomly assigned to a control (no treatment) condition. Three sets of findings are reported: survival analyses of time to reoffense; in-treatment change data relevant to the program's treatment goals; and the relationships between treatment measures and risk of reoffense. Over an average followup period of 38 months, there was a trend for treatment Ss to be at lower risk for both sex and violent crimes than were controls. Treatment Ss showed significant progress towards treatment goals of increased acceptance of personal responsibility for their crimes and decreased levels of deviant sexual arousal. Measures of personal responsibility, however, were not related to risk of rearrest for new sex crimes. Treatment Ss with high levels of both deviant and nondeviant sexual arousal were more likely to commit new sex offenses, but not other violent offenses. The strongest predictor in the study was a measure of the Ss' skills in applying the relapse prevention model, with highly skilled Ss being less likely to commit new sex crimes.