A sample of 593 pilot trainees in Officer Training School were administered the Strong Vocational Interest Blank and the Officer Biographical and Attitudinal Survey. Their performance in pilot training was monitored in order to assess the effectiveness of biographical, attitudinal, and interest data in predicting pilot training criteria. Using item analysis, four eliminee keys were developed from the survey data; regression analysis was utilized to select various models of predictors to be used as possible Pilot Selection Composites (PSC). Using the optimal cutoff score for PSC Model 1, 38 percent of the eliminees were correctly identified and only 10 percent of the graduate group incorrectly labeled as potential failures. For Model 2, 18 percent of the eliminee category was identified as high-risk for attrition along with seven percent of the successful graduates. With Model 3, 45 percent of the eliminees were labeled as potential failures along with 20 percent of the graduate group. The possibility of increasing the accuracy of prediction by using non-cognitive data and the operational usefulness of the composites are discussed. (Author/BW)