This investigation of the familial aggregation of rheumatoid arthritis in Rochester, Minnesota, was prompted by the considerable variability in previous reports and the need to interpret findings in light of the recently established human lymphocyte antigen (HLA)-DR4 association. The historical cohort methodology was applied to determine the incidence of adult-onset rheumatoid arthritis in 1631 biologic relatives of 78 probands compared with the Rochester population incidence. The ratio of the age- and sex-adjusted rates in first-degree relatives compared with the general population was 1.7 (95% confidence interval 1.0-2.9). The increase was concentrated in the 16- to 40-year-old age group, suggesting some disease heterogeneity. However, the level of familial risk was not significantly affected by the proband's sex, seropositivity, age, or parental disease status. Integrating these findings with prior research in which case ascertainment was complete led to the conclusion that familial aggregation of rheumatoid arthritis is weak. The apparent discrepancy between weak familial aggregation and the known strong HLA-DR4 association with rheumatoid arthritis was resolved by examining the mathematical relationship between the measures of association in the two different types of studies. Results show that to be consistent with weak familial clustering, any putative susceptibility gene must have very low penetrance, and/or there must be a large residual of sporadic cases.