Based on experimental and clinical data, it appears that human T cells are capable of mediating tumor cell destruction and, thus, are potentially important for immunotherapy of cancer. To date, neither the mechanisms responsible for T cell-mediated tumor cell destruction in vivo nor in vitro correlates of clinical responses in cancer patients treated with immunotherapy have been defined. Nevertheless, substantial evidence for the presence in cancer patients of specific autotumor (AuTu) responses mediated by T lymphocytes has accumulated. T cells recognize tumor-associated antigens (Ags) by means of clonally distributed T cell receptors (TCR). Molecular analysis of the preferential use of the TCR V beta gene families for recognition of Ags (tumor-associated peptides) by circulating or tumor-infiltrating T cells indicates that clones of AuTu-reactive effector T cells are present in patients with cancer. Recent advances in the characterization of tumor peptides bound to the MHC class I or class II molecules, definition of allele-specific consensus motifs, and availability of computer programs for modeling of T cell Ag interactions now allow for identification of specific T cell-reactive tumor peptide epitopes from proteins with known amino acid sequences. Also, the tumor-bearing host appears to be able to discriminate between tolerance to self, antitumor responses, and autoimmune phenomena. This type of a regulatory mechanism has not been precisely defined, but it might be surmised that immunosuppression, which is commonly seen in the tumor microenvironment, may result from attempts by the host to dampen or control self-reactivity rather than from tumor-induced down-regulation of T cell responses to escape immune surveillance. To progress toward molecular immunotherapy of cancer and to overcome tumor-host-induced immunosuppression in the tumor microenvironment, a better understanding of T cell-tumor interactions is necessary. It might be possible in the future to select for therapy clones of human T cells expressing the desired TCR and reacting with specific tumor peptides. It might also be feasible to reduce or eliminate tumor-mediated immunosuppression of T cell responses. Such selective molecular interventions in human cancers will depend on current advances in the definition of tumor Ag epitopes which elicit strong and sustained T cell responses.