Some of the problems unique to evaluation research, in particular to the evaluation of longitudinal studies of social programs, are presented. A working model found to be useful in structuring compensatory education program evaluation activities is also offered. Since the goals of social programs such as Head Start are dynamic rather than static, the evaluation design must be flexible in order to accomodate unexpected events and alterations in the program process. Attrition and the transiency of student program participants are highlighted as two of the chief problem areas in evaluating longitudinal studies. (MV)