This report presents data concerning the validity of a set of experimental computerized and paper-and-pencil tests for measures of on-job performance on global and job elements. It reports on the usefulness of 30 experimental and operational variables for predicting marks on 42 job elements and on a global criterion for Electrician's Mate, Personnelman, Sonar Technician, and Apprenticeship rating groups. There was little or no evidence of consistency of the job element characteristics across ratings. The job elements which were highly predictable were those which were important and central to the duties of particular ratings. For the technical ratings, the most effective predictors of job element marks were experimental tests, with the best such tests being computer-administered. Use of time required and importance ratings as moderators for prediction of global marks from the marks for job elements did not result in any practical increase in validity coefficients. Generally, low correlations were found between empirically-derived estimates of importance of personal attributes for particular job elements and similar estimates based on the judgments of personnel experts. Synthetic validity was generally not as accurate as multiple regression for predicting job performance. (Author/DEP)