The continuous increase in the number of acquired immunodeficiency syndrome (AIDS) cases for whom no effective therapy is currently possible mandates attempts at developing primary prevention by a vaccine. Two basic unknowns are considered important: the identification of virus-exposed, protected individuals; and the isolation of the antigen which contains epitopes which induce a protective response. Although almost all individuals exposed to human T-cell leukemia-lymphoma virus type III (HTLV-III) develop antibody, most of these do not have neutralizing antibody. The antigen which can induce the response is the major external glycoprotein, which is highly glycosylated (Mr 120,000). Based on past attempts at developing vaccines against retroviruses, the most feasible configuration will be the glycoprotein linked to its transmembrane protein and assembled into micelles or rosettes by hydrophobic bonding. Any virus preparation containing nucleic acids could be considered less safe. An advanced version of such a viral subunit presentation is matrices composed of immunostimulating complexes. This format could also be useful for the inoculation of sequence determined synthetic peptides or genetically engineered readout products of the viral envelope (env) gene. Potential problems exist in that there is extensive heterogeneity among various HTLV-III isolates, particularly in the env gene. This fact and the known relationship of HTLV-III to some lentiviruses suggest that functional antigenic variation could be encountered. The methodology of developing a vaccine against the retroviruses causing AIDS should also be helpful in designing vaccine strategies against human leukemia and lymphomas caused by other members of this virus family.