ALTHOUGH THE CLASSICAL METHOD OF EVALUATING THE IMPACT OF A PROGRAM IS EXPERIMENTAL RESEARCH DESIGN, SUCH A DESIGN IS SELDOM POSSIBLE WITH HUMAN SUBJECTS. QUASI-EXPERIMENTAL RESEARCH IS MORE OFTEN USED IN CRIMINAL JUSTICE RESEARCH, THEREFORE, BUT SINCE SUCH RESEARCH CANNOT CONTROL FOR THE EFFECT OF ALL OTHER VARIABLES, IT IS NEVER POSSIBLE TO REACH UNEQUIVOCAL RESULTS. OTHER TYPES OF DESIGN, SUCH AS CROSS-SECTIONAL RESEARCH, WHICH CORRELATES EVENTS AT A PARTICULAR POINT IN TIME, HAVE SIMILAR FLAWS. IN ADDITION, DECIDING ON THE CRITERION OF SUCCESS OR FAILURE GREATLY COMPLICATES THE EVALUATION TASK. THE TECHNICAL ASPECT OF EVALUATION RESEARCH IS DATA ANALYSIS, WITH METHODS RANGING FROM SIMPLE DESCRIPTION IN PERCENTAGES TO COMPLEX FORMS OF MULTIPLE REGRESSION ANALYSIS AND STATISTICAL INFERENCES. THE GOAL OF MOST ANALYSES IS TO DISTINGUISH A CORRELATION BETWEEN AN INDEPENDENT VARIABLE (NUMBER OF POLICE, FOR EXAMPLE) AND A DEPENDENT VARIABLE (CRIME RATE). THE KIND OF STATISTICAL TECHNIQUE APPLICABLE DEPENDS UPON THE LEVEL OF MEASUREMENT THAT HAS BEEN ACHIEVED--NOMINAL, ORDINAL, OR INTERVAL. THE PROBLEMS LIES IN THE FACT THAT STRONG CORRELATIONS ARE NOT FREQUENT IN CRIMINAL JUSTICE RESEARCH. THUS, THE UNCERTAIN NATURE OF MUCH EVALUATION RESEARCH LEADS DECISIONMAKERS TO EITHER IGNORE IT OR USE IT TO SUPPORT PREVIOUSLY HELD POSITIONS. EVEN IF THE FINDINGS ARE CLEAR, HOWEVER, THEY ARE SELDOM IMMEDIATELY IMPLEMENTED. THOSE WHO EVALUATE ARE OFTEN ACADEMICALLY TRAINED AND CONCERNED WITH REACHING 'OBJECTIVE' CONCLUSIONS THAT HAVE THEORETICAL RATHER THAN APPLIED RELEVANCE. THOSE ADMINISTRATORS WHOSE PROGRAMS ARE BEING EVALUATED ARE OFTEN MORE CONCERNED WITH PROTECTING THEIR OWN INTERESTS AND PROJECTING A POSITIVE IMAGE THAN WITH OBJECTIVE TRUTH ABOUT THE PROGRAM. ACTION RESEARCH, WHICH BRINGS THE ADMINISTRATORS INTO THE CHANGE PROCESS, IS A PROMISING STEP TOWARD RESOLVING THIS CONFLICT. FOOTNOTES AND A BIBLIOGRAPHY ARE PROVIDED. (AUTHOR ABSTRACT MODIFIED--AOP)