Under comparison were an experimental, therapeutic community and a traditional regime which emphasized training rather than treatment for rehabilitating delinquents. After a four-year controlled trial, in which delinquent boys were assigned to the units randomly, the results indicated similar post-treatment success rates. A ten-year follow-up provided no evidence of delayed differential treatment effects. Nearly half of the group has been in prison, and there is a high level of persistent offending within the sample. Although these results were inconsistent with the prevailing medico-psychological model, they were consistent with social learning theory explanations, which emphasized the importance of an individual's current environment in providing relevant opportunities and reinforcements. Treatment regimes also control these factors and the offender's behavior during treatment, but when the offender is returned to his original environment, his delinquent responses reemerge. These results had little immediate impact on rehabilitation policy in terms of applying behavioral methods to the offender in his natural environment. Furthermore, the efficacy of controlled trials in penal evaluation was questioned in terms of the problems of implementing the research design and securing interpretable results. Because of the shortcomings of the controlled trial, alternative evaluative strategies, such as the cross-institutional design were discussed. Following this particular trial in Great Britain, a closer relationship developed between theory, research, and policy, leading to a situational crime prevention approach. The author contends that rehabilitation efforts might be improved were they to be more closely geared to a rational choice approach to criminal behavior, if they stressed the rational aspects of offending, had a crime-specific focus, and expressed the need to pay attention to the circumstances of the particular crime and the growth of certain forms of crime. However, because criminal behavior is complex, involving many factors, intervention will only be temporarily effective for particular offenses. 40 references.