Nine hundred fifty-five of 1,384 (69 per cent) gay and bisexual men enrolled in a prospective study of the natural history of human immunodeficiency virus (HIV) infection who reported engaging in anal intercourse in the past six months were surveyed about condom use practices for both insertive (IAI) and receptive anal intercourse (RAI). The following results were obtained: 23 per cent of the men reported that they always used condoms for IAI and 21 per cent for RAI; 32 per cent sometimes used condoms for IAI; 28 per cent sometimes used condoms for RAI; 45 per cent never used condoms for IAI; and 50 per cent never used condoms for RAI. Multiple logistic regression analysis revealed that the following variables were associated with both insertive and receptive condom use: condom acceptability; a history of multiple and/or anonymous partners in the past six months, and the number of partners with whom one is "high" (drugs/alcohol) during sex. Knowledge of positive HIV serostatus was more strongly associated with receptive than with insertive use. Condom use is a relatively complex health-related behavior, and condom promotion programs should not limit themselves to stressing the dangers of unprotected intercourse.