A majority of CTL-recognized antigens of human melanomas have turned out to be unmutated differentiation antigens expressed on melanomas and melanocytes alike. Additional, relatively cancer-specific antigens which are expressed on a significant proportion of melanomas, but not on most normal tissues, have also been shown to be recognized by CTLs. Finally, CTLs have been shown to detect mutated oncoproteins which are expressed in a wide spectrum of cancers but not in normal tissues. These observations have given rise to a view that human cancers share a number of antigens, which can form the basis for their immunotherapy. In contrast, cancer antigens of mice and rats (which are detected by tumor rejection assays in vivo, rather than CTLs generated in vitro) have been observed, generally, to be individually distinct, i.e. unique for each individual cancer. No convincing examples of shared cancer antigens exist in animal models of cancer. This dichotomy between the nature of cancer antigens of humans ('shared'), and of rodents ('unique') is addressed, and hopefully, resolved in this article.