The Tegel records were consulted on all prisoners released between 1979 and 1974 who had spent at least 1 year in prison and whose maximum age at the time of release was 50 years. Thereby, 1,503 prisoners were selected. On the average, they had been in the community for 4.5 years, and were 33 years old. A total of 72.1 prisoners were convicted for property crimes. All the cases were divided into 10 groups. By the time of release, three experimental groups had completed one of the three therapeutic program models. The social-therapy model concentrated on individual therapy using psychological approaches including psychoanalytical and behavior therapy. The school model emphasized compensatory education including courses at primary and secondary levels. The social training model concentrated on group therapy using the approaches of the first model supplemented by courses dealing with problems of postprison adjustment. Each of three other experimental groups had been assigned to one of the program models but had been dropped from the program before release. Among the four control groups, the largest was composed of the 889 inmates released from prison after serving orthodox confinement. Another 114 inmates had requested but were refused assignment to the social-therapeutic unit and were transferred to an open prison. The results showed that exposure to the unit reduced recidivism by 18 percent. Among those prisoners not accepted by the therapeutic unit, the reconviction rate was also lower when they were granted access to other nonpunitive strategies: the open institution, conditional release, and parole supervision. Apparently, the arbitrary diagnosis of prisoners as 'nontreatable' because of their serious crimes was inadequate. However, no significant differences in outcomes were detected among the three therapeutic models when measured by reconviction rates. To compare the models, a much more precise identification of therapeutic processes is necessary. Background of the study, corroboration by other German studies, and future research plans are discussed. Statistical data, footnotes, and over 30 references are included.