We have followed a community-based cohort of recent-onset arthritis for eight years. In addition to Waaler-Rose and latex agglutination tests we used class-specific enzyme immunoassay for the determination of rheumatoid factors (RFs). Eighty-seven patients had erosions by the end of the follow-up period. The seropositive cases in each RF assay were divided into three groups of equal size according to the mean strength of RF reactivity for three or four available follow-up specimens. The radiological index values in the high-titer groups were no higher than the values for the seropositive cases on average. Thus, no severity gradient according to RF titer could be demonstrated.