The classic experiments of Doherty, Zinkernagel, and Blanden revealed the requirements for recognition by cytotoxic T lymphocytes (CTL) induced in response to infection by lymphocytic choriomeningitis virus (1). 51Chro_ mium labelled target cells were lysed by CTL only if they were infected with LCMV and also shared class-I alleles of the major histocompatibility complex (MHC) with the donor of the CTL (2). This observation was soon generalized to the responses to many additional viruses and other intracellular parasites, the products of minor histocompatibility loci, tumor antigens, and haptenated cell surfaces; it was then enshrined in the term" MHC-restricted" recognition (3, 4). Although the CTL responding to a variety of virus infections could be shown to differentiate between viruses, what remained unclear for several years was the nature of the antigens recognized by class I-restricted CTL and how they might interact with an MHC molecule (3, 4). The evidence prior to the use of recombinant DNA techniques tended to favor the intuitive idea that CTL recognized foreign glycoproteins inserted alongside MHC molecules in the membrane of the target cell (5, 6-10). This view is illustrated by a recent review by Klein (11). The evidence accumulated during the last four years is not compatible with this concept and suggests that class I-restricted T cells recognize the degraded fragments of proteins that have passed through the cytoplasm of the target cell (12, 13, 14, 15, 16, 17) either as a result of new synthesis or by entry from outside through the membrane of an endosome or