To examine racial differences in the incidence of systemic lupus erythematosus (SLE). A population-based registry of SLE patients in Allegheny County, Pennsylvania, was used to identify incident cases of SLE diagnosed between January 1, 1985 and December 31, 1990, from 3 sources, by medical record review (University of Pittsburgh Lupus Databank, rheumatologists, and hospitals). Capture-recapture methods using log-linear models were used to estimate the level of case-finding and to calculate 95% confidence intervals (CI). Incidence rates were calculated per 100,000 population. A total of 191 definite and 78 probable incident cases of SLE were identified, and the overall annual incidence rates were 2.4 (95% CI 2.1-2.8) and 1.0 (95% CI 0.8-1.3), respectively. The crude incidence rates of definite SLE were 0.4 for white males, 3.5 for white females, 0.7 for African-American males, and 9.2 for African-American females. The annual incidence rates of definite SLE remained fairly constant over the study interval. African-American females with definite SLE had a younger mean age at diagnosis compared with white females (P < 0.05). Since the overall ascertainment rate was high (85%; 95% CI 78-92%), the ascertainment-corrected incidence rate for definite SLE, 2.8 (95% CI 2.6-3.2), was similar to the crude rate. Our rates clearly confirm previous reports of an excess incidence of SLE among females compared with males and among African-Americans compared with whites. We have used capture-recapture methods to improve the accuracy of SLE incidence rates, and we advocate their use to facilitate comparisons across studies.