A strategy for evaluation studies is outlined based upon a longitudinal multivariate interactional approach as opposed to the more typical "engineering" approach, which focuses upon input-output differences. It is argued that evaluative research should encompass the process and context of education in addition to the outcomes of education in order to accrue information relevant to program improvement as well as accountability. It is further suggested that with this conception of evaluative research, the evaluation of educational programs provides a vehicle for advancing social science as well as the social welfare. (Author)