We have focussed on the differences in origin and physiological properties of two classes of self-reactive T cells. Autoreactive T cells described in many laboratories are activated in the course of normal immune responses to foreign antigen. These T cells can be shown under well-defined conditions to be the direct progeny of antigen-stimulated precursors. This, together with evidence that their activation requirements can be distinguished from those of antigen-specific, MHC-restricted T cells, leads us to suggest that they represent a particular physiological state that recapitulates the conditions of thymic selection and is induced in many antigen-specific, MHC-restricted peripheral T cells as a result of normal antigen-dependent activation. Although it appears that the associated physiological properties can be stable in some in vitro maintained lines, it is possible that this is normally a transient state in vivo. Available evidence concerning the specificity of these T cells indicates only that they can be activated in the absence of any identifiable foreign antigen by class II MHC-syngeneic but not MHC-allogeneic stimulators. We have suggested that such T cells are specific for the same elements, possibly an association of MHC and other self-peptides (Singer et al. 1987), that are the basis for positive selection in the thymus. The properties of these autoreactive T cells need to be distinguished from those of T cells associated with autoimmune pathology. It is presumed that autoimmune T cells are directly activated in a resting state by specific self-peptides. Our interest in distinguishing these self-reactive T-cell populations has focussed on different predictions concerning the diversity of their associated self-reactive repertoires. The relative complexity of the immune repertoire expressed in autoreactive T cells expanded by positive selection and restimulated in the course of normal antigen-specific immune responses should be considerably greater than that of autoimmune T cells constrained by negative selection and a narrow window of escape from self-tolerance. We were greatly hindered in our initial efforts in this analysis by the considerable effort required to characterize any specific immune repertoire. A published technique employing poly(A) tailing (Frohman et al. 1988) did not work efficiently in our hands, although others (Loh et al. 1989) have apparently had some success. We describe above an alternative approach, linker-facilitated PCR, which we have employed for efficient repertoire analysis. Using this method we have been able to identify dominant utilization of the Va4 family in T cells specific for the synthetic peptide YYEELLKYYEELLK.(ABSTRACT TRUNCATED AT 400 WORDS)