The use of admission interview comments to predict clinical clerkship success of medical students was evaluated. Narrative comments made by admissions interviewers regarding an applicant's skills and attitudes were coded, as were narrative evaluations of these students during year III of required clerkships in pediatrics and internal medicine in regard to their clinical skills, attitudes, maturity, and overall clerkship performance. For the 88 students of the University of Arizona College of Medicine, the predictive relationship between these variables was determined using multiple regression. Interview variables accounted for 22 percent of the variance in clinical performance as measured by evaluative comments. By comparison, objective preadmission variables for the same students accounted for 20 percent of the variance of clinical performance as measured by nationally standardized tests. The moderate correlation between these criteria suggests that they may be assessing different aspects of clinical performance, and thus require different kinds of predictors. The findings suggest that admissions interviews are useful in predicting clinical performance and that continuing the current admissions interview process is justified. Under the current selection process, admission interviewers are able to collect information that predicts clinical performance better than grade point averages and Medical College Admissions Tests do. However, the objective admissions measures best predict an objective measure of clinical knowledge. Therefore, choice of admissions variables will depend on which performance measures are of greatest concern. (SW)