The multivariate approach of canonical correlation was used to assess selection procedures of the Air Force Academy. It was felt that improved student selection methods might reduce the number of dropouts while maintaining or improving the quality of graduates. The method of canonical correlation was designed to maximize prediction of academic performance, military performance, and attrition. High school course and grade information, scores on the Admissions Testing Program Verbal and Mathematics Aptitude Tests, a physical aptitude measure, an athletic activity index, and an index of nonathletic extracurricular activities were analyzed for 257 individuals. The criterion variables were cumulative grade-point average at the end of the freshman year, an index of military performance, and a zero-one attrition variable that identified subjects who left the Air Force Academy after their freshman year. Findings are discussed, and it is concluded that the selection process can be improved by using additional data from standard high school transcripts and modifying weights of current selection variables. (SW)