We assessed the factors associated with unprotected anal intercourse in gay men's casual partnerships, using data from a longitudinal cohort study ongoing since October 1992. Participants were recruited through gay community and a variety of other sources and interviewed for the Sydney Men and Sexual Health study. Of 1,037 interviewees at outset, 834 men had casual partners during the six months prior to interview; 609 had any anal intercourse with a casual partner. Of 172 HIV-positive men, 58 (33.7%) engaged in unprotected anal intercourse with casual partners. Of 662 non-positive men, 111 (16.8%) engaged in unprotected anal intercourse with casual partners. Logistic regression analysis revealed that gay men's engagement in unprotected anal intercourse with casual partners (compared with engagement in safe sex--either no anal intercourse or protected anal intercourse) was predicted by positive HIV status, occupational differences (being out of the workforce or in white collar positions rather than in professional or managerial positions), engagement in a wider range of anal practices, believing withdrawal to be safe and less favourable attitudes towards condoms. Additional logistic regressions which, initially, excluded positive men who only engaged in unprotected anal intercourse with positive casual partners (n = 26) and then excluded men who did not engage in anal intercourse (n = 225) yielded quite similar sets of predictors of unprotected anal intercourse with casual partners. However, in both the additional analyses HIV status was not a significant predictor. It is concluded that gay men who engage in unprotected anal intercourse with casual partners do not comprise a singular group, although they commonly engage in a diversity of anal practices and they dislike condoms.