A hypothetical model of the surface (SU) glycoproteins of human immunodeficiency virus (HIV) and other retroviruses is proposed. The model is based on repetition of a limited number of sequence motifs conserved within the virus family; similarities in biological, immunological, or genetic properties; as well as the tendency for regions of dissimilar sequence to share protein structures predicted by computer algorithms. It is proposed that the protein consists of three structural and functional domains interspersed by relatively conserved interdomain regions. For each retrovirus, these amino-terminal, central, and carboxy-terminal domains may play different roles in binding, postbinding events, and the immune response to viral infection.