Questions are raised about the future of research and development evaluation. Without a set of overarching policy questions or theoretical questions, evaluation data are simply not cumulative and cannot be organized into any data base. Education has never been a data-based enterprise. The meaning of evaluation should be broadened to include the generation of systematically collected information relevant to policy decisions about programs, and the development of models to improve understanding of problems and interventions. Evaluation should adapt to the life cycle of the program. A series of small, single-objective, studies may be more useful than a single large and complex study. The plans for evaluation should be negotiated with all the interested parties and the results should be addressed to the interests of all these groups and effectively synthesized. The federal government is likely to support large data banks and efforts to synthesize evaluation data, and to reduce support for evaluation of separate projects. (Author/CTM)