There is evidence indicating that autoreactive B cells constitute a substantial part of the B cell repertoire. This autoreactive repertoire secretes the so-called natural autoantibodies (NAA), characterized by their broad reactivity mainly directed against very well conserved public epitopes. They fulfill the definition of an autoantibody because they are self-reactive, but they are not self-specific. As yet, NAA directed against determinants of polymorphism have not been reported. Their germinal origin is suggested by their early appearance during ontogeny, by their expression of cross-reactive idiotopes, and by structural studies of their sequence. As for the physiological role of this repertoire, we can assume that it may play a major role as a first barrier of defense. It is presently unknown whether these polyreactive B cells could constitute a preimmune template that, through an antigendriven process, may be involved in the production of immune high-affinity antibodies. The presence of this repertoire in normal conditions challenges the clonal deletion theory as a unique explanation for self-tolerance. However, if we take into account that this autoreactive B cell repertoire is not self-specific, this contradiction may not be a real opposition. This repertoire may be the price to pay for ensuring selftolerance. It may escape clonal deletion and be induced into a compartment submitted to stringent regulation by its connectivity. In contrast, true anti-self-B cell clones may be submitted to very strict regulation through clonal deletion, receptor editing, or very stringent anergic mechanisms. However, it is important to keep in mind that pathogenic autoantibodies share binding specificities with NAA, in the sense that they are not self-specific and recognize public epitopes shared by all individuals of a given species and even shared by individuals belonging to phylogenetically distant species. The relationship between NAA and pathological autoantibodies is also presently unclear. It has not been definitively established whether pathological autoantibodies are the consequence of polyclonal stimulation, resulting in increased levels of germline-encoded NAA, or whether they are derived via an antigen-driven somatic mutation process operating in these NAA precursors. Although there are convincing results, indicating that the generation of pathogenic autoantibodies may be the consequence of an antigen-driven process, this has not been firmly established in the case of human pathogenic autoantibodies. It cannot be excluded that both polyclonal