Even though recidivism reduction is the ultimate goal of treatment, the low base rate of detected recidivism makes this an impractical outcome criterion in many settings. This paper discusses three general approaches that can be used to advance knowledge: large, well-controlled, single-site research studies; pooling of small studies through meta-analysis; and detailed examination of within-treatment change on dynamic (changeable) risk factors. Each approach has its own strengths and weaknesses. It is only through the accumulation of consistent results from diverse studies that researchers can learn what reduces recidivism for sexual offenders. The confidence in the results of meta-analysis depends on the quality of the individual studies considered. Although Hall's (1995) recent meta-analysis supported the efficacy of treatment, more studies with adequate designs are needed before skeptics will be convinced. Since the base rate of recidivism is low, information alone cannot be used to guide the development of individual treatment programs. Instead, programs can monitor their own efficacy through the analysis of within- treatment changes on dynamic risk factors. Studies of within- treatment changes need to be supported, however, with a parallel stream of research aimed at identifying empirically validated, dynamic risk factors for sexual offenders. 58 references