Pre-enrollment characteristics that distinguish nonpersisting students from students who persist into the second semester of their freshman year were studied at the University of Iowa in 1978. Three sources of information were used to obtain predictor variables: American College Testing (ACT) program test scores, the ACT Student Profile, and high school transcript information from the registrar. Study objectives were to reduce the large number of pre-enrollment variables to a small number of useful preduction variables and to use linear and nonlinear classification techniques to exhaust the predictive power of the predictive pwoer of the predictor variables. Of the 2,850 entering freshmen in the fall 1978 cohort, there was enough information to allow factor scoring for 1,711 of the students. It was found that the difference between using 43 factors and 23 factors was negligible. There were no indications from this study or previous studies that a larger number of hte pre-enrollment variables could account for more the variance of persistence; however, the 23 factors did not predict retention any better than an earlier study that used only three variables. It is concluded that pre-enrollment variables per se did not directly predict persistence. It is recommended that the smallest set of factors that maintains the current level of prediction of persistence should be developed and used in retention studies to save on costs of data storage and analysis. (SW)