Several recent case-control studies have suggested that nulliparity is a risk factor for rheumatoid arthritis (RA). We studied parity for its association with the incidence of RA in a cohort of adult Finnish women examined by the Mobile Clinic of the Social Insurance Institution in 1966-72. The 15,441 women were > or = 30 yr of age and had neither arthritis nor a history of it at the time of the baseline examination. They were followed up by examining records kept at the population register of the institution, to identify patients entitled to free anti-rheumatic medication. Up to late 1989, 269 women had developed RA. Of the incident cases, 176 were rheumatoid factor (RF)-positive. Compared with nulliparous women, the age-adjusted relative risks (95% confidence intervals) of parous women developing RF-positive RA, RF-negative RA or either type were 1.26 (0.82-1.92), 0.80 (0.48-1.32) and 1.06 (0.77-1.46), respectively. The number of childbirths presented an n-shaped relationship with the risk of RF-positive RA, but there were distinct differences between subpopulations. In female farmers the association was inverse, but in housewives and in other occupational groups (mainly employees) it was clearly positive. The results challenge the view that there is a causal relation between parity status and the development of RA.