Although dealing broadly with issues involving program evaluation, this paper is especially concerned with large-scale evaluations of federally funded education programs that are national in scope and, to some extent at least, compensatory in nature. The paper concentrates on a few basic problem areas that seem inherent in the concept of evaluation. They include, for example, problems in defining treatment and outcome variables that will yield useful evaluation results, and difficulties in establishing comparative benchmarks against which program outcomes should be gauged. The paper concludes with a discussion of the role of evaluation vis-a-vis education research, and suggests a strategy by which future evaluations might make a greater contribution to an understanding of important underlying relationships in the teaching-learning process. To illustrate both problems and approaches in evaluation, examples are given from two large studies currently being conducted by System Development Corporation for the U.S. Office of Education: an evaluation of the Emergency School Aid Act (ESAA) and a multiyear evaluation of the sustaining effects of compensatory education. (RC)