WEAK EXPERIMENTAL DESIGNS AND LAX METHODOLOGIES ARE PRIMARY OBSTACLES TO THE EFFECTIVE EVALUATION OF CORRECTIONAL REHABILITATION PROGRAMS. WITH NO IMMINENT PANACEA TO MAJOR PROBLEMS IN SIGHT, CONTROLLED SYSTEMATIC STUDY HAS MUCH TO CONTRIBUTE. THE ADVANTAGES OF RANDOMIZED EXPERIMENTAL TESTS SHOULD BE PARTICULARLY CONSIDERED. IF AN EXPERIMENT IS NOT RANDOMIZED, CONCLUSIONS ABOUT TREATMENT EFFECTS ARE FREQUENTLY JEOPARDIZED BY PLAUSIBLE ALTERNATIVE EXPLANATIONS, AND ARGUMENTS FOR THE CAUSAL RELATIONSHIP BETWEEN TREATMENT AND EFFECT MUST TOO OFTEN RELY ON UNSUBSTANTIATED ASSUMPTIONS. IF AN EXPERIMENT IS RANDOMIZED, THE PROBABILITY OF NONTREATMENT VARIABLES ACCOUNTING FOR THE OUTCOME IS MINIMIZED, AND SO THE VALIDITY AND CREDIBILITY OF THE RESULTS CAN BE GREATER THAN THOSE OBTAINED FROM QUASI-EXPERIMENTAL AND NONEXPERIMENTAL STUDIES. RANDOMIZED STUDIES ARE SAID TO BE INAPPROPRIATE OR IMPRACTICAL IN FIELD SETTINGS. WHILE THIS IS SOMETIMES TRUE, IT IS BY NO MEANS ALWAYS THE CASE. INFORMATION IS NEEDED ON WHAT CONDITIONS PERMIT OR PROHIBIT THE USE OF AN EXPERIMENTAL EVALUATION DESIGN. RANDOMIZED STUDIES ARE ALSO SAID TO BE COSTLY AND SLOW. TRUE EXPERIMENTS NEED NOT BE EITHER COSTLY OR SLOW IF THE TREATMENT PROGRAM IS SHORT AND HAS IMMEDIATE EFFECTS. INFORMATION ON THE COMPARATIVE COST OF DIFFERENT EVALUATION METHODOLOGIES IS SCARCE. WHEN CONDITIONS PREVENT RANDOMIZATION, QUASI-EXPERIMENTAL TECHNIQUES ARE USEFUL ALTERNATIVES. THE STATISTICAL CONTROL USED IN SUCH CIRCUMSTANCES CANNOT COMPENSATE FOR EXPERIMENTAL CONTROL, HOWEVER. WHEN AN EXPERIMENTAL DESIGN IS USED, IT MUST BE FOLLOWED BY CAREFUL IMPLEMENTATION, MEASUREMENT, ANALYSIS, AND INTERPRETATION OF THE TREATMENT PROGRAM. FIGURES AND REFERENCES ARE PROVIDED. (AUTHOR ABSTRACT MODIFIED---RCB)