Early rheumatoid arthritis (RA) must be differentiated from benign self-limiting polyarthritis because of the risks associated with treatment of RA. Conventional, widely available clinical and laboratory variables, measured at first clinic visit, were studied for their ability to predict persistence in 112 patients with up to 6 months of joint symptoms. Those 65 patients with symmetrical peripheral polyarthritis were followed for 1 year: 36 who underwent spontaneous remission were classified self-limiting synovitis (SLS); the remaining 29 were termed persistent synovitis (PS). Univariate analysis suggested more severe disease in PS at presentation but showing considerable overlap with SLS, making clinical discrimination difficult. Multivariate analysis confirmed this overlap but identified a subset of most helpful variables. The RA latex was the most powerful variable, yet accounted for only 45% of the variability in outcome. Combining a positive RA latex with an ESR > 30 mm/h carried a relative risk for PS of 4.33, with specificity 94% but sensitivity only 69%. Self-limiting synovitis initially could not be distinguished from early RA: hence RA may exist in two forms, the traditional persistent form and a less well recognized abortive form.