While the excess risk of clinical cardiovascular disease among persons with diabetes mellitus is well established, most autopsy studies have not been able to elucidate reasons for the excess, to assess potential selection bias, or to adjust for other cardiovascular risk factors. The purpose of this study was to examine the predictive relation between diabetes and autopsy evidence of coronary atherosclerosis and myocardial lesions. Among 8,006 Japanese-American men examined at baseline in 1965-1968 as part of the Honolulu Heart Program, 7,591 were free of cardiovascular disease, and 1,515 of these men died over a 17-year follow-up period. Protocol autopsies were performed for 83 diabetic men and 159 nondiabetic men. Diabetes status was defined using self-reported history and treatment at several examinations, as well as physician diagnoses during hospitalization between 1965 and 1984. An excess of coronary artery atherosclerosis, assessed by mean panel score (3.4 vs. 3.0, p = 0.017) and percentage of intimal surface with raised lesions (56.6% vs. 47.4%, p = 0.024), was present among diabetic men but diminished to nonsignificant levels (3.3 vs. 3.0, p = 0.102, and 53.9% vs. 48.8%, p = 0.183, respectively) after adjustment for other cardiovascular risk factors. Myocardial lesions (acute, healing, or fibrotic) occurred significantly more frequently among diabetics than among nondiabetics (77.7% vs. 63.4%, p = 0.035), even after adjustment for other risk factors. Potential autopsy selection bias assessed in several ways appeared minimal. Among men with mild atherosclerosis, diabetics had more small and large myocardial lesions than did nondiabetics, although differences were not statistically significant (p < 0.10). It appears that the more adverse risk factor profile among diabetics accounts for some of the observed excess of coronary atherosclerosis. However, diabetes was independently associated with myocardial lesions, and these findings suggest a role for nonatherosclerotic mechanisms, such as clotting abnormalities or microvascular disease, in accounting for the excess clinical heart disease found in persons with diabetes.