Methodological difficulties inherent in the design of studies of college impact are reviewed and an attempt is made to provide an improved research design. The discussion of research problems focuses on those of inferring causation. A characteristic three-component model of student development, comprised of student output, student input, and college environment, is analyzed to illustrate the problems in the research design presently employed. Inferential errors resulting from the use of this design are categorized and discussed. Three statistical methods for analyzing the model components, matching, actuarial tables, and linear multiple regression analysis, possible measurement techniques, and various methods of data collection are individually evaluated. The paper concludes with suggested resolutions of the inferential dilemmas. (PR)