The existence in the mature T cell repertoire of a high precursor frequency of cells which recognise allogeneic MHC molecules appears to contradict the well-established dogma of positive selection for self MHC restriction. In order to explore the possibility that alloreactive cells are derived from a fraction of the repertoire that is not self-MHC-restricted, the contribution of in vivo-primed T cells to "primary" alloresponses was investigated. Peripheral blood T cells were separated into virgin and memory populations by sorting for low and high levels of LFA-3 expression, and their proliferative responses to MHC incompatible stimulator cells was quantitated. The results demonstrated that approximately half of a "primary" alloresponse is contributed by previously primed T cells that, by definition, must be self-MHC restricted. Furthermore it was possible to define the original MHC-restricted antigen specificity of two T cell clones raised against the allospecific HLA-DR1 from a DR4Dw4/DRw13DW19 responder. The emerging consensus view that anti-MHC alloreactive T cells, like antigen-specific T cells, are specific for MHC/peptide complexes, and have a parental self-MHC restriction, begs a structural explanation. Comparison of multiple DR beta 1 domain sequences reveals that DR molecules fall into groups that have extensive homology in the residues on the beta 1 domain alpha-helix that are predicted to point up towards the T cell receptor (histotopic), and thus to determine MHC restriction. Given that the DR alpha chain is invariant this creates the possibility that anti-DR allorecognition can mimic self-restricted recognition. Within these groups of histotopically similar DR products there are multiple differences in the peptide-binding residues that lie on the inner aspects of the alpha-helix or on the floor of the antigen-binding groove. As a consequence, it is predicted that a different array of endogenous peptides will be bound, due to determinant selection. Thus, allorecognition within these groups may result from the recognition of endogenous peptides that are bound by stimulator but not by responder MHC products, seen in a self-restricted manner. In combinations where histotopic similarity does not exist, allorecognition may be best explained by the chance occurrence of a receptor selected for intermediate affinity for thymically expressed MHC molecules having a higher affinity for an allogeneic histotope. Such a receptor would have been deleted in a thymus expressing the allospecificity, but would be perceived as "safe" in the absence of this MHC product.