The relationship between primary and memory cytotoxic T lymphocyte (CTL) responses, and the factors influencing entry into memory, are poorly understood. Here we address this in the context of Epstein-Barr virus (EBV), a persistent human herpesvirus in which memory CTL responses in long-term virus carriers are highly focused on epitopes preferentially drawn from just three of the eight available virus latent proteins, EBNAs 3A, 3B, and 3C. To determine whether this unusual level of focusing is a consequence of long-term virus challenge, we carried out a detailed analysis of EBV antigen/epitope specificities in the primary virus-induced CTL response in 10 infectious mononucleosis (IM) patients of different HLA types. Primary effectors, studied in ex vivo assays and by limiting dilution cloning in vitro, were again highly skewed toward a small number of viral epitopes, almost all derived from the EBNA3 proteins, with CTL to the immunodominant epitope accounting for at least 1% of the circulating CD8+ IM T cell pool. This is the first unequivocal demonstration of an EBV-specific CD8+ CTL response in IM. Prospective studies on individual patients showed that, whereas all of the EBV reactivities found in CTL memory had been detectable earlier during primary infection, the memory population was not simply a scaled down version of the primary response. In particular (a) differences in the relative frequencies of CTL to immunodominant versus subdominant epitopes appeared to be much less marked in memory than in primary populations, and (b) we found at least one clear example in which a significant virus-specific reactivity within the primary response was never detectable in memory.