Both helper and cytotoxic T lymphocytes generally recognize protein antigens not in their intact form, as antibodies do, but on the surface of another cell, after "processing" by that cell to unfold or cleave the protein into fragments and after association of the processed antigen with major histocompatibility complex (MHC) molecules on that cell. This complex process leads to immunodominance of certain segments from the protein, which depends not only on structural features intrinsic to the antigenic segment itself, but also on antigen processing and on the structure of the MHC molecules of the responding individual. We have explored all three of these factors, including the enzymes involved in processing, the way peptides bind to MHC molecules, and structural features such as helical amphipathicity that seem to favour T cell recognition. We have used this information to locate and characterize antigenic sites of proteins of interest for vaccine development, including proteins from the malaria parasite and the AIDS virus, HIV. For HIV, we have identified both helper and cytotoxic T cell sites, coupled a helper site to a B cell site to produce a synthetic immunogen that elicits neutralizing antibodies, and studied the effect of viral sequence variation on cytotoxic T cell recognition and binding of the immunodominant peptide to MHC molecules. This information suggests strategies for the rational design of synthetic or recombinant vaccines.