Six widely used college admissions variables and four course-specific measures (tests of entering mathematical skills and physics knowledge, and indexes of previous study of mathematics and physics) were correlated with final marks for 495 students in an introductory college physics course. The college admissions variables included scores on American College Test (ACT) mathematics, ACT natural science, ACT composite, School and College Ability Test (SCAT) quantitative, SCAT verbal, and high school percentile rank. Separate stepwise multiple regressions revealed that the complete set of predictor variables accounted for 43.5% of the variance in course mark, whereas the six admissions variables accounted for only 28.6%. Similarly, the specially constructed tests of mathematical skills and physics knowledge were substantially better predictors than the corresponding admissions variables. These results suggest that prediction of success in particular instructional settings can be substantially improved if skills and knowledge specific to each setting can be identified and measured. (Author/RL)