T helper (Th) lymphocytes mediate critical effector and regulatory functions in infectious, allergic, or autoimmune diseases. Th cells possess clonal receptors that recognize antigenic peptides that are complexed with self-molecules of the major histocompatibility complex (MHC) on the surface of antigen presenting cells. An organism's repertoire of T cell receptors must be broad enough to recognize any possible microbial antigen. At the same time, tissue destruction resulting from the attack of autoreactive T lymphocytes that recognize self-peptides must be avoided. It was therefore believed that the immune system could distinguish between self and non-self antigens. This hypothesis was supported by several lines of evidence, including the seemingly exquisite specificity of immune responses. What, then, triggers autoaggressive attacks by the immune system? Clinical and epidemiological observations strongly suggest a link between infection and autoimmunity. A popular hypothesis considers autoimmunity as a side effect of antimicrobial immune responses. Cross-reactive T cells, capable of recognizing both microbial and self-peptides, have been prime suspects as instigators of autoimmunity ever since computerized data base searches revealed astonishing sequence homologies between microbial and self-peptides. Here we review recent data that show a previously unexpected degeneracy of antigen recognition by T cells. It has become clear that each individual T cell receptor can recognize a large number of different ligands. Furthermore, structural criteria rather than sequence homology dictate the antigen recognition process. Thus, the idea that cross-reactivity per se would cause autoimmune disease is most likely too simple. Instead, a variety of different molecular mechanisms dictate the immunological outcome of ligand recognition by T cells.