An anonymous self-administered questionnaire was distributed to (1) male members of the Danish Gay and Lesbian Association (2) through a gay magazine and (3) to readers of a gay pornographic magazine. For the purpose of this study sexual practices were classified into three categories taking into account the HIV-status of the respondent and his partner(s): safe sex (mutual masturbation, sex with condoms, sex without condoms between two HIV-positives), potentially safe sex (oral-genital sex without condoms irrespective of HIV-status, anal-genital sex without condoms between two HIV-negatives), unsafe sex (anal-genital sex without condoms between discordant partners or partners of unknown HIV status). Of the 2058 respondents 29.7% had had unsafe sex in the last 12 months. Multivariate analysis by recursive graphical models showed that sexual practice was directly related (that is conditionally dependent given the rest of the variables) to having a steady partner. Among men without a steady partner sexual practice was also directly related to age and number of partners showing an increase in unsafe sex with number of partners and a decrease with age. Thus of the men 16-19 years of age 43.5% had had unsafe sex irrespective of number of partners vs 5.7% of men older than 44 years and with one to two partners. Sexual practice was not directly related to any other demographic or psychosocial factor in the study. The adopted classification of sexual practice preclude that the high occurrence of unsafe sex could be explained by unsafe sex taking place among partners of concordant HIV status. The results emphasize the need for further preventive efforts to reduce transmission of HIV among homosexual men.