The study used a dataset on 251 sex offenders (136 rapists and 115 child molesters) who were discharged over a 25-year period to examine changes in recidivism as a function of changes in dispositional definition of reoffense (e.g., arrest or conviction), changes in the domain of criminal offenses that are considered, and changes in the length of exposure time. The data show that both rapists and child molesters remain at risk to reoffend long after their discharge, in some cases 15-20 years after discharge. There was a marked underestimation of recidivism when calculating a simple proportion that consisted of those who were known to have reoffended during the follow-up period. There was also a marked underestimation of recidivism when the criterion was based on conviction or imprisonment. There is a need to address the methodological problems identified so as to improve the accuracy of risk-based, high-stakes decision-making about sex offenders by incorporating consideration of taxonomically differentiated subgroups. The base rate problems identified also affect the evaluation of treatment outcome. Policy implications are also discussed. 4 tables and 28 references