A 2-year study of 206 college students at an urban research university in a southern state examined personal, environmental, academic, and non-academic predictors of: (1) course grade performance; and (2) academic retention. Of particular interest were "high-risk" or "disadvantaged" students and how they compensated for their marginal educational preparedness. Data sources included two student questionnaires, instructor grade records, and student transcript records for the subsequent 2 years. Automatic Interaction Detection, developed by J. A. Sonquist et. al. (1973), was used to search for interaction effects. The technique revealed that the relationship between course performance and predictor variables differed for advantaged and disadvantaged students. As was consistent with previous research, personal academic factors accounted for a significant increase in grade performance, and environmental factors tended to affect disadvantaged students more than others. Analysis of retention rates revealed reading comprehension ability to be the best predictor. The implications of the findings are discussed with reference to institutional interventions. Two tables and two figures present study data. (Author/SLD)