The relationship between rheumatoid arthritis and silicosis was studied by means of a case-control study in South African gold miners. One hundred and fifty seven miners with rheumatoid arthritis classified as "definite" (91) or "probable" (66) were individually matched by year of birth with miners who had no evidence of rheumatoid arthritis. Unmatched analysis of the case-control status for "probable" and "definite" cases yielded an odds ratio of 2.84 (p = 0.0001). Separate analyses yielded an odds ratio of 3.79 (p = 0.0006) for "definite" cases, a non-significant odds ratio for "probable" cases, and an odds ratio of 5.00 (p = 0.0003) for the presence of rheumatoid factor. These results could not be explained on the basis of cumulative dust exposure or intensity of exposure. The rate of progression of silicosis in both the "definite" and the "probable" groups was greater than for the control patients with silicosis, as was the probability of silicosis presenting at the start with larger nodules (type r).