This review solidifies a new concept that common and rare types of human cancers harbor a variety of tumor-specific mutant proteins that may be recognized as tumor-specific antigens. These mutant proteins are encoded by oncogenes or suppressor genes that have undergone structural mutations resulting from point mutations, chromosomal translocations, internal deletions and viral insertional mutagenesis; several of these changes result in fusion proteins. While there is no evidence that immunosurveillance against these mutant proteins can prevent the development of primary cancers without prior immunization of the host, such tumor-specific molecules might be important for diagnosis and as targets for specific immunotherapy once the cancer has developed or even as targets for preventive cancer vaccines. Evidence further supports the notion that cytolytic or helper T cells are exquisitely selective in recognizing intracellular mutant proteins, and tumor-specific T cell clones presently available may become useful for identifying previously unrecognized tumor-specific mutations. Many tumor-specific mutant proteins clearly play a causative role in the establishment of malignant behavior, whereas other carcinogen-induced changes have at least immunological relevance. In any case strong evidence in mouse and man indicates that a single malignant cell can express multiple independent antigenic target sites. Such multiplicity may allow a multi-pronged immune attack that substantially decreases the chance of tumor escape. Future work must explore whether immune responses to tumor-specific mutant proteins can lead to immunological tumor rejection and explore the possibility of chemically engineering tumor mutant peptides to be highly immunogenic, even in hosts that have previously failed to respond to the tumor.