The clinical experience that spontaneous anti-melanoma immune reactivity can occur has stimulated the search for methods to induce this in patients diagnosed with melanoma. Non-specific approaches using a variety of immune stimulants such as BCG or cytokines have met with limited success, as have vaccines derived from tumour cells. More recently, melanoma antigens have been identified that can act as specific targets for immune recognition. Cell surface glycolipids such as the gangliosides GM2 and GD3, can be targeted by antibodies. This has provided the basis for clinical trials with ganglioside vaccines and monoclonal antibody infusions. Antigens recognized by cytotoxic lymphocytes have also been described in the last 5 years. These are peptide antigens derived from intracellular proteins which are present on the cell surface in association with HLA molecules. These antigens include MAGE 1 and 3, tyrosinase, MelanA/MART-1 and gp100. Clinical trials with these have commenced and novel treatment strategies are being developed. Since tumours can be typed for specific antigens and specific immune responses can be measured, the reasons for treatment success or failure can be analysed more effectively than in the past. For example, the emergence of antigen-negative tumour variants can be assessed. This should enable a more systematic approach for developing new immunotherapies for melanoma.