Student and institutional characteristics related to college freshman persistence were studied. Persistence was examined for 5 institutions (5,950 students) at 4 points in time: end of first term, reenrollment in the spring, end of spring term, and reenrollment in the fall of the sophomore year. Data from a variety of sources were used, and predictor variables were selected to meet the components of the model of persistence of V. Tinto. Three surveys were developed for the study. Both linear and logistic regression were used to develop separate prediction models for each institution. Estimated success rates and accuracy rates of the models for identifying high-risk students were calculated from the probabilities generated by logistic regression. Results support Tinto's view that persistence models are specific to individual institutions and to the time period being examined. An example is provided for using the results to identify high-risk students and to develop intervention strategies targeting key factors related to student retention. A discussion of the practical issues involved in collecting retention data is also provided. Fourteen tables present study findings. (Author/SLD)