The failure of HIV-1 to escape at some cytotoxic T-lymphocyte (CTL) epitopes has generally been explained in terms of viral fitness costs or ineffective or attenuated CTL responses. Relatively little attention has been paid to the evolutionary time required for escape mutants to be detected. This time is significantly affected by selection, mutation rates, the presence of other advantageous mutations, and the effective population size of HIV-1 in vivo (typically estimated to be approximately 10(3) in chronically infected patients, though one study has estimated it to be approximately 10(5)). Here, we use a forward simulator with experimentally estimated HIV-1 parameters to show that these delays can be substantial. For an effective population size of 10(3), even highly advantageous mutants (s = 0.5) may not be detected for a couple of years in chronically infected patients, while moderately advantageous escape mutants (s = 0.1) may not be detected for up to 10 years. Even with an effective population size of 10(5), a moderately advantageous escape mutant (s = 0.1) may not be detected in the population within 2 years if it has to compete with other selectively advantageous mutants. Stochastic evolutionary forces, therefore, in addition to viral fitness costs and ineffective or attenuated CTL responses, must be taken into account when assessing the selection of CTL escape mutations.