This article describes a novel approach to HIV vaccine design that is, as yet, unproven and still in preliminary development. In rhesus macaques infected with simian immunodeficiency virus (SIV), we have identified particular cellular immune responses that select for viral variants during primary infection. We speculate that the detection of viral variants with altered amino acids in CTL epitopes implies the successful clearance of cells harboring wild-type virus. Here, we present our rationale suggesting why such potent early CTL responses that exert an antiviral effect may be particularly attractive targets for induction by candidate vaccines. Conventional wisdom suggests that regions of the virus that are structurally and functionally important will generally be well-conserved both among clades and within an infected host. Amino acid replacements within these well-conserved regions should be difficult for the virus to accommodate. Therefore, these regions are traditionally considered ideal targets for vaccine induced immune responses because they are refractory to CTL escape mutations. Many examples of these regions have been identified in both HIV-1 and SIV(mac) (J. Immunol. 162 (1999) 3727; J. Virol. 67 (1993) 438) and have been included in candidate vaccine formulations. Human clinical trials testing these vaccines are currently underway. Our proposed method of vaccination with CTL epitopes that escape explores an alternative hypothesis. Rather than engendering responses to regions of the virus that do not escape, we reason that vaccination needs to accelerate the development of the initial immune responses that effectively select for amino acid variants during acute infection. By examining CTL escape during the acute phase, we will identify CTL responses that the virus cannot tolerate and incorporate these responses into vaccines.