Several models for enrollment projections have been developed based on past performance. One of these, a computer-assisted model developed at the California State University at Northridge, was tested for possible use at Los Angeles Pierce College (LAPC). From three to five previous comparable college terms are used in the model to predict enrollments for up to three academic levels; for each course, program, and department; and for the total college. Linear, curvilinear, logarithmic, or exponential models are possible. In testing this model at LAPC, the Earth Science and Business departments were selected. Data on weekly student contact hours and census enrollment were obtained for all courses and programs over the previous five years. Though this model had remarkable success in predicting enrollment within the state and university system, at LAPC the error in prediction for most cases was very large--due primarily to the artificial limitations on or increase in the number of students enrolled in specific programs caused by changing the course and section offerings. Given the predictive failure of this model, how are enrollment predictions best made? Knowledge of local economic and demographic factors permits general enrollment predictions, but decisions about program and course modifications should be based on the participation of department heads and faculty, analyses of five-year course enrollment trends, and consideration of college mission. (AYC)