AN INTIMATE RELATION EXISTS BETWEEN A STUDY'S RESEARCH DESIGN AND DATA ANALYSIS PROCEDURES. THE EXPERIMENTAL DESIGN IS THE PRIMARY DETERMINANT OF STATISTICAL ANALYSIS. INTERNAL VALIDITY OF RESEARCH FINDINGS IS CONCERNED WITH THE EXTENT TO WHICH ONE CAN BE ASSURED OF LOGICALLY CORRECT INFERENCES. THE TRUE EXPERIMENTAL METHOD AVOIDS CERTAIN THREATS TO INTERNAL VALIDITY (E.G., HISTORY, MATURATION, INSTABILITY, TESTING, INSTRUMENTATION, REGRESSION ARTIFACTS, SELECTION, EXPERIMENTAL MORTALITY, AND SELECTION-MATURATION BIAS). EXTERNAL VALIDITY MEANS THE DEGREE TO WHICH RESULTS CAN BE GENERALIZED OR APPLIED TO DIFFERENT SETTINGS. IN ORDER TO REFLECT REALITIES OF A SITUATION AND PRAGMATIC CONCERNS OF AN EVALUATOR, QUASI-EXPERIMENTAL OR PRE-EXPERIMENTAL DESIGNS HAVE BEEN SUGGESTED AS A COMPROMISE APPROACH TO EVALUATION. THESE DESIGNS SHOULD BE USED WITH CAUTION DUE TO INTERNAL VALIDITY THREATS. EVALUATION RESEARCH HAS AS A PRIMARY GOAL THE GENERATION OF INFORMATION TO AID DECISIONMAKING. MISUSES OF EVALUATION, TERMED 'PSEUDO-EVALUATIONS,' CAN OCCUR. DIFFERENCES BETWEEN BASIC RESEARCH (ACQUISITION OF NEW KNOWLEDGE) AND EVALUATION RESEARCH (APPLICATION OF KNOWLEDGE) ARE DIFFERENTIATED. PRIMARY AND SECONDARY BENEFITS OF EVALUATION RESEARCH IN CORRECTIONAL SETTINGS ARE NOTED. REFERENCES ARE CITED. (DEP)