Polyclonal activation has been proposed as the reason that autoantibodies are produced during autoimmune disease. This model denies a role for specific antigen selection of B cells and predicts instead a multiclonal population of unmutated or randomly mutated autoantibodies. We have found that the genetic features and clonal composition of spontaneously derived immunoglobulin G (IgG) antiself-IgG (rheumatoid factor (RF] autoantibodies derived from the autoimmune MRL/lpr mouse strain are inconsistent with both the predictions of this model and the actual outcome of experimental polyclonal activation. Instead we have found that MRL/lpr RFs are oligoclonal or even monoclonal in origin. They harbour numerous somatic mutations which are distributed in a way that suggests immunoglobulin-receptor-dependent selection of these mutations. In this sense, the MRL/lpr RFs resemble antibodies elicited by exogenous antigens after secondary immunization. The parallels suggest that, like secondary immune responses, antigen stimulation is important in the generation of MRL/lpr RFs.