The initial sample totaled more than 2,300 persons from 7 U.S. cities. Approximately half were randomly assigned as experimentals; the remainder were assigned to be controls. The sample design called for the followup of all sample members for a minimum of 18 months after their enrollment; 68 percent were assigned 27-month followup interviews, while 22 percent were assigned 36-month followup interviews. Participants in the program could earn entry level wages as well as bonuses for good performance and attendance; at the end of a fixed period they were required to leave the program whether or not they had obtained a regular job. Supported work is primarily distinguished from other employment and training programs by its emphasis on the following techniques: group support, graduated stress, and close supervision. Key findings showed an initially strong, positive effect of program participation on ex-offenders' employment, but after 12 months of followup, the experimental-control differential in hours worked per month was virtually nonexistent. There was some evidence that supported work may have led to reductions in drug use, but the favorable findings were concentrated in the early months after enrollment and among the earliest enrollees who were followed for 3 years. There was no evidence of a short-term or long-term effect on criminal activity and no overall indication that the probabilities of arrest and drug use were associated with employment outside of supported work. Some positive effects were observed for older participants, those who received welfare or food stamps in the month prior to their enrollment, and those who were early enrollees in the supported work program. The study concluded that those who are older and have families may be more prepared to take advantage of the opportunities that supported work provides. Footnotes and 40 tables are given. Five appendixes provide supplementary tables, study methodology, and assessment of the effects of the length of time spent in supported work on program impacts, employment trends among controls, and approximately 50 references.