Although melanoma accounts for only 4% of skin cancers, it is responsible for 80% of deaths from skin cancers and its incidence in Caucasians has been increasing steadily during the last 30 years. So far, no treatment - except surgery at the earliest stages of the disease - has been shown to significantly improve survival. New treatments are thus clearly needed. The interest of immunologists for melanoma is based on particular features of this tumor. Rare spontaneous regressions have been described, which are possibly mediated by immune responses. Moreover, melanoma cell lines are relatively easy to obtain, providing essential tools for laboratory studies. The first melanoma vaccines involved inoculations of patients with autologous or allogeneic melanoma cells, as well as a variety of immunological adjuvants. Since the beginning of the nineties, the identification of antigens recognized on human tumors by autologous T lymphocytes has opened the way for new vaccination strategies involving molecularly defined tumor antigens. An important group of antigens recognized by T lymphocytes is encoded by “cancer-germ line genes”, which are expressed in tumors of various histological types, but are silent in normal tissues, with the exception of testis germinal cells and placental trophoblast. Since the latter do not express the HLA molecules required to present these antigens to the T lymphocytes, cancer-germ line genes encoded antigens are only present on tumors, which should limit the risk of generating autoimmune diseases as a consequence of vaccination. Therefore, these widely shared tumor specific antigens should represent good targets for the development of cancer vaccines. Our clinical research program of therapeutic vaccinations focuses on antigens encoded by the MAGE family of cancer-germ line genes. Most of the patients included in our phase I/II immunization trials had measurable metastatic melanoma. Several MAGE peptides as well as a recombinant MAGE-3 protein have been tested, while several additional trials are ongoing, including immunizations with a recombinant poxvirus coding for 2 MAGE epitopes. No major toxicity was reported. Tumor regressions have been observed in a minority of patients, mainly those who had regional or distant metastases without visceral involvement. Some of these regressions have been complete and long lasting. Although the rate of objective tumor response observed is low, it is clearly higher than the rate of spontaneous tumor regression observed in melanoma. Other immunization modes against T-cell defined epitopes are currently being explored by several groups in human clinical trials. Vaccines include peptides presented by class I or class II HLA molecules, proteins given alone or mixed with immunological adjuvants or cytokines, recombinant viral or bacterial vectors, dendritic cells and DNA encoding the antigen. Adoptive transfer of T lymphocytes selected for their capacity to recognize defined epitopes presented by the tumor represents another type of approach aiming at the destruction of the tumor by the immune system. It is difficult to predict whether and when therapeutic vaccination against cancer will reach an efficacy that will be sufficient for a standard cancer treatment. Provided their low toxicity, these vaccines should be tested in an adjuvant setting, at earlier stages of the disease.