Evaluations occur within a political decision-making mileau, where multiple stakeholders are contending for limited funds. Given the subjective basis of empirical information, different conclusions or recommendations about a program may result from different ideological, theoretical, and disciplinary perspectives. The logic behind the interpretation of results, and the assumptions necessary for such interpretations, must be specified and explained to facilitate the most appropriate use of an evaluation. Because of the complexity of many statistical techniques presently used, much work is needed to identify what assumptions must be met for meaningful and useful interpretations of results in a specific decision-making situation. The rationales for both the inclusion and the exclusion of the variables to be considered in an evaluation should be made explicit. The problem of obtaining a matched control group is often nearly impossible. The relatioship between the statistical analysis and the evaluation question is often based on tenuous assumptions. The evaluation of Project Follow Through is used to exemplify these problems. (Author/CTM)