The authors identified variables that enhance the predictability of academic performance and retention in an urban, commuter college. They used a longitudinal design with 2 waves of data collection: prior to 1st-semester attendance and again 6 semesters later. The results support the following conclusions: (a) After controlling for precollege indices of academic ability, specific variables within demographic, personal behavior, expectations, and attitudinal domains improved the predictability of grade point average by 15.5% for Whites and 20.3% for Blacks; overall, these analyses accounted for 61% of the variance in grade point average for Whites and 47% for Blacks; (b) for retention, the comparable increment in predictability was 2.5% for Whites and less than 1% for Blacks; overall, the percentage of retention variance accounted for was 42% for Whites and 29% for Blacks; and (c) a paradoxical finding was the combination of a positive correlation between high school average and students' expected grades on tests and a negative correlation between expected and actual grades. The paradox is resolved by citing data that demonstrate (a) a disparity between high school grades and independently measured academic accomplishment; (b) the consequent false attributions students draw about their academic abilities; and (c) the impact of these false attributions on the priority students place on their academic responsibilities.