Knowledge of the nature of immune reactivity to human tumors has increased dramatically in the past decade. Prior to this, there was significant doubt as to whether tumor antigens actually existed on growing human cancers, and if so, there was considerable skepticism about whether immunotherapies could ever be used to affect growing tumors. The development of reproducible techniques for generating human lymphocytes with reactivity against tumor antigens, the demonstration that the recognition of tumor antigens obeys the same immunologic laws of MHC restriction that characterize immunologic reactions against other antigens, and the demonstration that T-cell transfer therapies--as well as immunotherapies using immune stimulants such as IL-2--could cause the regression of established growing cancers in humans have transformed the study of human tumor immunology from a speculative science to one with concrete applications to the treatment of patients with cancer. The cloning of the genes encoding cancer antigens and the molecular identification of the encoded proteins has opened new possibilities for the development of cancer treatments. Although several human cancer antigens have been identified, only a small number of histologic types of tumors have been studied and these investigations need to be extended to identify putative tumor antigens expressed on common epithelial neoplasms. There is considerable uncertainty surrounding the mechanism by which immune reactions to these cancer antigens are generated in humans. Thus far, all of the cancer rejection antigens recognized by TIL, as well as additional antigens described by others, represent antigens that are encoded by normal nonmutated genes that are expressed on selected normal tissues as well as on tumors. An understanding of the mechanisms by which the tumor can break tolerance to these normal antigens represents a significant challenge to immunologic studies, the answer to which may provide additional insight into methods for enhancing antitumor immune responses.