Rheumatoid factor (RF) is often detected in subjects who do not have rheumatoid arthritis (RA). To determine if the presence of RF is predictive of the development of RA, 2,712 Pima Indians of Arizona, 20 years of age or older, initially without RA, have been observed for up to 19 years with biennial examinations. These included a medical history, physical examination of the joints, radiographs, and RF determination by sheep cell agglutination test. During the study period, 70 new cases of RA developed. When the population at risk was stratified by RF titer, with reclassification at subsequent examinations if the RF titer changed, the age- and sex-adjusted incidence of RA increased with higher titers of RF. The incidence of RA (in cases per 1,000 person-years) according to RF titer was 2.4 (RF titer less than 1:2); 6.7 (titer 1:2-1:16); 11.0 (titer 1:32-1:256); and 48.3 (titer greater than 1:256) (P less than 0.001). The same trend was also found within each age and sex group, and within groups defined by the number of American Rheumatism Association criteria present before the definite diagnosis. We conclude that the presence of RF, in subjects without RA, is a risk factor for the development of RA, and that this risk is related to the RF titer. It is also suggested that RF may represent a marker of the earliest phases of the pathogenetic process of RA, that may be detectable before the appearance of other features that permit a clinical diagnosis.