The treatment of cancer with tumor vaccines has been a goal of physicians and scientists ever since effective immunization against infectious disease with vaccines was developed. In the past, major tumor antigens had not been molecularly characterized. Recent advances are, however, beginning to define potential molecular targets and strategies and this had evolved with the principle that T-cell mediated responses are a key target for approaches to cancer immunization. In addition, these antigens are not truly foreign and tumour antigens fit more with a self/altered self paradigm, compared to a non-self paradigm for antigens recognized in infectious diseases. Potential antigens include the glycolipids and glycoproteins (e.g. gangliosides), the developmental antigens (e.g. MAGE, tyrosinase, melan-A and gp75) and mutant oncogene products (e.g. p53, ras, and HER-2/neu). Innovations for construction of cancer vaccines are emerging from these advances in molecular immunology and cancer biology. While vaccines against infectious agents are models for vaccine development, there are clearly distinct considerations and problems associated with cancer vaccines. One of the focal issues in designing active cancer immunotherapy is that cancer cells are derived from normal host cells. Thus, the antigenic profile of cancer cells closely mimics that of normal cells. How the immune system identifies and destroys cancer cells is therefore crucial. Clearly, the ultimate goal of tumor vaccine design is the generation of antigen-specific vaccines. The recent success identifying molecularly defined tumor antigens opens up potentially novel strategies for this approach. Vaccine possibilities include purified proteins and glycolipids, peptides, cDNA expressed in various vectors, and a range of immune adjuvants. The molecular and structural definition of tumor antigens provides an opportunity for cautious optimism that we are entering an era when we will soon begin to recapitulate the success of immunization against infectious disease.