The incidence of the acquired immune deficiency syndrome (AIDS) among persons infected with human T-lymphotropic virus type III (HTLV-III) was evaluated prospectively among 725 persons who were at high risk of AIDS and had enrolled before October 1982 in cohort studies of homosexual men, parenteral drug users, and hemophiliacs. A total of 276 (38.1 percent) of the subjects were either HTLV-III seropositive at enrollment or developed HTLV-III antibodies subsequently. AIDS had developed in 28 (10.1 percent) of the seropositive subjects before August 1985. By actuarial survival calculations, the 3-year incidence of AIDS among all HTLV-III seropositive subjects was 34.2 percent in the cohort of homosexual men in Manhattan, New York, and 14.9 percent (range 8.0 to 17.2 percent) in the four other cohorts. Out of 117 subjects followed for a mean of 31 months after documented seroconversion, five (all hemophiliacs) developed AIDS 28 to 62 months after the estimated date of seroconversion, supporting the hypothesis that there is a long latency between acquisition of viral infection and the development of clinical AIDS. This long latency could account for the significantly higher AIDS incidence in the New York cohort compared with other cohorts if the virus entered the New York homosexual population before it entered the populations from which the other cohorts were drawn. However, risk of AIDS development in different populations may also depend on the presence of as yet unidentified cofactors.