As the proportion of patients with rheumatoid arthritis (RA) who have anti-keratin antibodies (AKA) varies in different ethnic groups, we studied its occurrence in a hospital populations with RA and its association with different disease variables. Sera from 84 consecutive patients with RA, 100 healthy controls and 85 disease controls (polyarticular juvenile rheumatoid arthritis, systemic lupus erythematosus, ankylosing spondylitis) were tested for AKA by an indirect immunofluorescence assay that used rat esophagus as substrate. The proportion of patients with RA who had AKA (47/84) was higher than in healthy controls (2/100; P < 0.001) and in disease controls (2/85; P < 0.001). The frequency of AKA positivity was higher among patients who had severe disease (P < 0.05) and rheumatoid factor. Anti-keratin antibody is present in 56 per cent of our patients with RA and is associated with severe disease.