This review discusses human tumor antigens recognized by T lymphocytes and their potential for cancer immunotherapy. It focuses on non-viral antigens. The initial studies on mouse tumors identified two mechanisms for generating tumor-specific antigens: point mutations and transcriptional activation of genes not expressed in normal tissues. Three families of genes—MAGE, BAGE, and GAGE—encode highly specific tumor antigens, expressed in various tumors but rarely in others. These genes are expressed in testis and placenta. CTLs derived from a melanoma patient recognized antigens encoded by these genes. These antigens are expressed in over 60% of melanoma patients, and their expression in testis suggests they may not cause harmful side effects. Another source of tumor-specific antigens is the activation of an intron in the N-acetyl-glucosaminyltransferase V gene, which produces a new open reading frame. Some CTLs recognize mucin, a surface protein with repeated amino acid sequences, in tumors where it is underglycosylated. This recognition is HLA-unrestricted and has been observed in myeloma cells. Differentiation antigens, such as tyrosinase, Melan-A/Mart-1, gp100, and gp75, are expressed on melanocytes and melanomas. These antigens are recognized by CTLs, and their recognition is HLA-restricted. However, the effectiveness of these antigens in immunotherapy is uncertain. Point mutations can generate antigens recognized by autologous CTLs. For example, a mutation in cyclin-dependent kinase 4 increases its oncogenic potential and is recognized by CTLs. Another mutation in the β-catenin gene produces a new antigenic peptide. Ubiquitous antigens, expressed in many tissues, may not be suitable for immunotherapy unless they are overexpressed in tumors. "Reverse immunology" involves identifying antigens from known HLA-binding peptides. This approach has been used to identify antigens from mutated oncogenes. The review concludes that antigens identified by direct immunology (from tumor cell stimulation) may be more effective for immunotherapy than those identified by reverse immunology. Clinical trials are needed to determine the effectiveness of these antigens. Some antigens are in early clinical trials, and future research may lead to more effective cancer immunotherapies.This review discusses human tumor antigens recognized by T lymphocytes and their potential for cancer immunotherapy. It focuses on non-viral antigens. The initial studies on mouse tumors identified two mechanisms for generating tumor-specific antigens: point mutations and transcriptional activation of genes not expressed in normal tissues. Three families of genes—MAGE, BAGE, and GAGE—encode highly specific tumor antigens, expressed in various tumors but rarely in others. These genes are expressed in testis and placenta. CTLs derived from a melanoma patient recognized antigens encoded by these genes. These antigens are expressed in over 60% of melanoma patients, and their expression in testis suggests they may not cause harmful side effects. Another source of tumor-specific antigens is the activation of an intron in the N-acetyl-glucosaminyltransferase V gene, which produces a new open reading frame. Some CTLs recognize mucin, a surface protein with repeated amino acid sequences, in tumors where it is underglycosylated. This recognition is HLA-unrestricted and has been observed in myeloma cells. Differentiation antigens, such as tyrosinase, Melan-A/Mart-1, gp100, and gp75, are expressed on melanocytes and melanomas. These antigens are recognized by CTLs, and their recognition is HLA-restricted. However, the effectiveness of these antigens in immunotherapy is uncertain. Point mutations can generate antigens recognized by autologous CTLs. For example, a mutation in cyclin-dependent kinase 4 increases its oncogenic potential and is recognized by CTLs. Another mutation in the β-catenin gene produces a new antigenic peptide. Ubiquitous antigens, expressed in many tissues, may not be suitable for immunotherapy unless they are overexpressed in tumors. "Reverse immunology" involves identifying antigens from known HLA-binding peptides. This approach has been used to identify antigens from mutated oncogenes. The review concludes that antigens identified by direct immunology (from tumor cell stimulation) may be more effective for immunotherapy than those identified by reverse immunology. Clinical trials are needed to determine the effectiveness of these antigens. Some antigens are in early clinical trials, and future research may lead to more effective cancer immunotherapies.