The performance of the University of Arizona College of Medicine's graduating class of 1976 on the National Board of Medical Examiners (NBME) Parts I and II was analyzed in a structural equation model. Although the primary emphasis of this paper is toward model development, the effects of several background and demographic variables such as age, sex, socioeconomic and marital status were estimated, together with those of several major events in these students' medical school careers. About half of the variance in NMBE I total score could be explained with knowledge of overall undergraduate Grade Point Average (GPA), Medical College Admissions Tests (MCAT) scores, and the students' sex. Over 85% of the variance in NMBE II total score could be explained using the variables shown in the structural equation model. NMBE I performance did not greatly affect NMBE II performance when other variables were included. The methodological techniques used to isolate the effects of these variables can be applied to many other analyses in health professions education. ( Author/CTM)