The major conceptual problem for HIV vaccine development has been the lack of information on immune responses known to correlate with protection against HIV infection in humans. In this regard, studies on the natural history of HIV infection and AIDS, especially of people with apparent resistance to HIV infection and of patients with HIV infection who have long term survival without disease progression, may provide important information for vaccine development. In addition, a major concern for the development of broadly effective vaccines has been the extensive genetic variability which is characteristic of HIV. In spite of these unknowns, the first generation of HIV candidate vaccines has been developed and evaluated. HIV candidate vaccines based on the subunit recombinant envelope concept (gp120 or gp160) have been shown to protect chimpanzees from HIV infection on challenge, and have now been evaluated in humans in phase I and phase II trials. These products are well tolerated, and capable of inducing neutralising antibodies, but not cytotoxic T lymphocytes. A second vaccine concept, currently in phase I trials, is based on live recombinant vectors, especially using poxvirus vectors followed by boosting with subunit recombinant envelope vaccines. This concept is theoretically very attractive because preliminary data suggest that these vaccines induce both humoral and cell-mediated immunity. However, no published information is available on the ability of live recombinant vector vaccines to protect chimpanzees from HIV infection. The next step in HIV vaccine development is to proceed carefully to expanded phase II and phase III trials to assess the protective efficacy of these candidate vaccines in humans. These trials will be extremely complex from the logistical, scientific and ethical points of view, and will require close collaboration between clinical, basic science and behavioural researchers, national and international organisations, and the pharmaceutical industry.