In an ongoing prospective study of homosexual men conducted since November 1982 in Vancouver, we identified 345 men who did not have antibody to human immunodeficiency virus (HIV) at the time of enrolment and for whom results of follow-up serologic testing were available. A total of 66 cases of seroconversion were documented among the 345 men between November 1982 and October 1985. Methods of survival data analysis that take into account the varying durations of follow-up were used to study the epidemiologic features of seroconversion in this group. The probability of seroconversion during the entire observation period was 23.1%. The seroconversion rates remained stable, at 10.5% and 10.0% during the last 2 years of the observation period. Cox regression analysis revealed the following variables to be independently associated with risk of seroconversion: frequent receptive anal intercourse, elevated number of male sexual partners in the year before enrolment, use of illicit drugs, a history of gonorrhea and age less than 30 years in November 1982. Multivariate analysis failed to reveal any role of oral sexual activity in the transmission of HIV. Oral ingestion of semen was not associated with seroconversion in either univariate or multivariate analysis. The observation that younger men were more likely to seroconvert suggests that young homosexual men were less likely than older men to modify their sexual behaviour.