AN EXPERIMENTAL DESIGN WAS USED IN WHICH, RANDOMLY, 188 PERSONS WERE ASSIGNED TO A WORK RELEASE GROUP AND 93 TO A CONTROL GROUP. FOLLOW-UP INTERVIEWS WERE CONDUCTED IN THE COMMUNITY, AND RECIDIVISM DATA WERE OBTAINED FROM DIVISION OF CORRECTIONS FILES AND FROM FBI RAP SHEETS. EIGHTEEN DIFFERENT INDICES OF RECIDIVISM WERE EXAMINED USING BOTH DISCRETE AND CONTINUOUS MEASURES. THE WORK RELEASE AND CONTROL GROUPS SHOWED NO APPRECIABLE DIFFERENCES ON ANY OF THE RECIDIVISM MEASURES AND LENGTH OF TIME IN WORK RELEASE ALSO FAILED TO BE CONSISTENTLY RELATED TO RECIDIVISM. CONTROLS WERE INTRODUCED FOR 16 DEMOGRAPHIC, PRIOR RECORD, AND JOB-RELATED VARIABLES, WHICH DID NOT ALTER THE ORIGINAL FINDINGS. POLICY IMPLICATIONS OF THE RESULTS ARE DISCUSSED. THE TEXT ALSO CONTAINS STATISTICAL TABLES REPRESENTING FINDINGS OF THE EXPERIMENTATION. (AUTHOR ABSTRACT)....KAP