Multiple regression analyses were employed to determine the relationships between achievement variables as predictors of the performance on Part I of the National Board of Medical Examiners examinations (NBME-I) of 366 medical students. The students' examination score averages in first- and second-year basic science courses were sequentially added to a composite Medical College Admission Test (MCAT) score to investigate the increases in prediction accuracy. Four prediction equations were cross validated on two subsequent medical school classes. While the multiple correlation between the predicted NBME-I score and the actual score significantly increased as first- and second-year course examination scores were added to the equations, the overall accuracy in predicting passing or failing did not substantially increase. The most useful equation employed a combination of the MCAT score and the first semester anatomy course examination score. The NBME-I predictions were given by letter to the 1984 and 1985 students approximately a year before they were to take the NBME-I examination so that students at risk of failure could undertake remedial study.