After more than a decade of intensive research, the precise role of human immunodeficiency virus (HIV)-specific cytotoxic T lymphocytes (CTL) in determining the course of the infection remains open to argument. It is established that HIV-specific CTL appear early in the infection and are temporally associated with the clearance of culturable virus from the blood; that CTL are generally detectable at very high levels throughout the asymptomatic phase and decline at the time of progression to AIDS; and that CTL-mediated killing is sufficiently fast to prevent production of new virions by HIV-infected cells. However, viral turnover is high throughout the course of the infection, and infected individuals progress inexorably to disease in spite of the CTL response. In order to address the question of whether CTL play an active part in influencing the course of HIV infection, one approach has been to seek evidence for CTL-mediated selection pressure on the virus. Several clear examples of CTL epitope-specific mutations selected to fixation are described. We argue that CTL escape is a common event which occurs at all stages of the infection. Detailed longitudinal studies are required to detect CTL escape and to understand the complexities contributed by factors such as a polyvalent CTL response and the presence of epitope variants which antagonise the CTL response. In conclusion, there is strong evidence of a dynamic process in which CTL impose important selection constraints upon HIV from which the virus attempts to escape; ultimately, at the time of disease progression, the tenuous control of CTL over the virus is lost.