BASED ON AN ANALYSIS OF RECIDIVISM RATES, THE FISHMAN STUDY CONCLUDED THAT THE 18 PROJECTS FAILED TO REHABILITATE CLIENTS. PROBLEMS IN THE STUDY'S FINDINGS AND GENERALIZATIONS PERTAIN TO THE LIMITED NATURE OF THE SAMPLE, THE USE OF A SINGLE OUTCOME MEASURE (RECIDIVISM), AND THE EVALUATION DESIGN (VALIDITY OF THE GROUP PRETEST-POSTTEST ONLY METHOD, LACK OF COMPARISON ALTERNATIVES, AND LENGTH OF PARTICIPATION REQUIRED BY PROJECTS). CONSIDERING THESE PROBLEMS, FISHMAN'S RESULTS ARE OF LIMITED UTILITY AND GENERALITY AND DO NOT PROVIDE A SUFFICIENT BASIS FOR MAKING DECISIONS ABOUT REHABILITATION AND DIVERSION SERVICES. A MODEL FOR EVALUATION RESEARCH SHOULD INCORPORATE CONSTRUCT VALIDITY, INTERNAL VALIDITY, PROCESS AND OUTCOME EVALUATION MEASURES, AND EXTERNAL VALIDITY. FURTHERMORE, PROGRAM EVALUATIONS SHOULD BE USED TO PROVIDE INFORMATION ABOUT CERTAIN ASPECTS OF A PROGRAM RATHER THAN TO DETERMINE OVERALL PROGRAM SUCCESS. FOOTNOTES ARE INCLUDED. (DEP)