The major focus of cancer immunology has shifted away from arguments about the validity of the immunosurveillance theory of cancer to the more basic question of tumor-specific antigens. Despite vast effort aimed at demonstrations of such antigens, their existence in the generality of cancer remains unproved. Serological analysis of 3 tumor types, mouse leukemia and sarcoma and human malignant melanoma, has received the most attention, and a rudimentary classification of the surface antigens expressed by these tumors has begun to emerge. The prime candidates for antigens that can be considered tumor specific are the few instances of Class 1 antigens that have now been serologically defined on mouse and human tumors. These antigens show an absolute restriction to individual tumors and are not demonstrable on any other normal or malignant cell type. Biochemical and genetic characterizations of Class 1 antigens represent an essential next step in an evaluation of the significance of these antigens. The surprising features of the thymus leukemia (TL) antigens of the mouse provide insight into the genetic origin of another key class of tumor antigens, i.e., those with characteristic properties of both differentiation and tumor-specific antigens. In normal mice, TL antigens are restricted to cells in the thymus, and strains differ with regard to expression versus nonexpression of TL antigens. Genetic information for TL is universal in mice, however, as leukemias that develop in mice normally lacking TL are found to express TL. What is clear from the past two decades of research in cancer immunology is that a far more detailed knowledge of surface antigens of tumor cells will be necessary before we can begin to assess the possibility of immunological control of cancer.