Overall, the meta-analysis provided solid support for the risk principle in classifying offenders by risk for the purposes of treatment. When studies that reported treatment outcome data by the risk level of the sample were analyzed separately, the findings were even stronger; however, the therapeutic effects of the risk principle applied only when the treatment programs were clinically and psychologically matched to the criminogenic needs of the offenders classified as high-risk. Empirical support for the risk principle of case classification did not vary with the treatment setting (community or institutional, restorative or nonrestorative); with the quality of the study (randomized design, attrition problems, rated comparability of groups, or independent evaluator); or with the offender's ethnicity (majority or minority; or history of violence. Risk did interact with age and gender, however, with the risk principle receiving most support in samples of women and in samples of younger offenders. Suggestions are offered for further research in this area. The two samples of studies created by Andrews et al. (1999) were pooled for this report. The first sample (154 tests of treatment) was derived from Whitehead and Lab (1989), and the additional studies included in Andrews, Zinger, et al. (1990). The second sample (220 tests) consisted of the expanded set reported by Andrews (1996) and Andrews and Bonta (1998), as well as additional studies by one of the authors of this article. Characteristics of the studies are outlined, along with the coding procedure, the calculation of effect sizes, and major variables. 6 tables and 38 references