ONLY A THEORY-BASED EVALUATION CAN CONTRIBUTE TO THE ACCUMULATION OF PRACTICAL KNOWLEDGE AND HAVE LONG-RUN IMPACT. THEORY-BASED RESEARCH DESIGN CAN APPLY TO ANY PROGRAM DESIGNED TO PREVENT CRIME OR JUVENILE DELINQUENCY INCLUDING DIVERSIONARY, TREATMENT, PREPAROLE OR PAROLE, AND OFFENSE PREVENTION PROGRAMS. THE IDEAL EVALUATION INCORPORATES THEORIES ON ACTIONS TO DEAL WITH PROBLEMS, AND THEORIES THAT DEAL WITH THE ORGANIZATION TAKING SUCH ACTIONS. PROGRAM EVALUATION SHOULD BE DESIGNED TO TEST ABSTRACT PRINCIPLES VALIDATED IN OTHER TYPES OF RESEARCH RELEVANT TO TYPES OF PROGRAMS AND CATEGORIES OF CLIENTELE. FOR EXAMPLE, RESEARCH TO TEST LEARNING THEORY PRINCIPLES WILL REQUIRE THREE TYPES OF STATISTICAL DATA: (1) INDICATORS OF THE SUBJECTS' PRIOR GRATIFICATIONS FROM CRIME; (2) MEASURES ON THE IMPACT OF PUNISHMENT IN REDUCING CRIME-RELATED GRATIFICATIONS; AND (3) EVIDENCE ON THE PROSPECTIVE GRATIFICATIONS OF ALTERNATIVES TO CRIMINAL BEHAVIORS. THEORY-GUIDED EVALUATION BEGINS WITH ABSTRACT POTENTIALLY POLICY-GUIDING IDEAS FROM WHICH ONE CAN DEDUCE THE CONSEQUENCES OF A SPECIFIC TYPE OF PROGRAM FOR A SPECIFIC TYPE OF SUBJECT. THE RESEARCH IS THEN DESIGNED TO DETERMINE IF THE DEDUCED CONSEQUENCES DO IN FACT OCCUR. THE IMPORTANCE OF CONTROLLING SUBJECT VARIABLES IN TERMS OF THE RELEVANT THEORY IS EMPHASIZED. THUS FOR INSTANCE, AN EVALUATION OF HALFWAY HOUSES SHOULD NEVER BE ATTEMPTED FOR ALL THOSE WHO RESIDE WITHIN THEM; RATHER, ONE SHOULD EVALUATE PROGRAM EFFECTIVENESS IN REDUCING RECIDIVISM ONLY FOR SPECIFIC CATEGORIES OF OFFENDERS FOR WHICH SOME THEORY PREDICTS EFFECTIVENESS. MANY TYPES OF ORGANIZATIONAL THEORY MAY BE APPROPRIATE TO PROGRAM EVALUATION. FOR INSTANCE ORGANIZATIONAL SELECTION BIASES MAY FAVOR SELECTION OF ONLY LOW RISK CATEGORIES OF SUBEJCTS OR STAFF MEMBERS MAY FOCUS TREATMENT ONLY ON THE MORE PROMISING AND COMPLIANT SUBJECTS. A POLITICALLY-ORIENTED PROGRAM ANALYSIS OF FUNDING PROCESSES AND NEGOTIATIONS AMONG COMPETING AND COALITION-FORMING INTEREST GROUPS MAY BE DESIRABLE. THE OBJECTIVE OF SUCH ORGANIZATIONAL ANALYSES IS TO DETERMINE CIRCUMSTANCES AND PROCESSES AFFECTING AGREEMENT BETWEEN THE REALITIES OF THE PROGRAM AND THE ORGANIZATION'S PLANS OR CLAIMS. SUCH AN ANALYSIS AIDS IN ASSESSING REASONS FOR FAILURE IN PREDICTIONS AND MAY ALSO LEAD TO ORGANIZATIONAL MODIFICATION NECESSARY TO ACHIEVE OFFICIAL OBJECTIVES. (JAP)