Some gay men who have unprotected anal intercourse avoid ejaculation-they practise 'withdrawal'. Using data collected in 1997 from a sample of Sydney gay men (n=625), we explored the relation between men's practice of ejaculation and their use of condoms. We also investigated whether men who had unprotected withdrawal but not unprotected ejaculation were more likely to think unprotected withdrawal was safe, liked condoms less, liked anal intercourse more, or were more sexually adventurous. Considering separately insertive and receptive anal intercourse with regular and with casual partners, we found that the majority of men who practised unprotected withdrawal also practised unprotected ejaculation. Of those whose only unprotected sex was withdrawal ('true withdrawers'), most never used condoms (they did not also have protected sex with ejaculation). True withdrawers were compared with men who had unprotected ejaculation, who always used condoms, who had no anal sex and who had no partners. Those who were true withdrawers with casual partners were more likely to believe withdrawal was safe; no group effects were found with regular partners. No significant differences in condom attitudes were found. True withdrawers with regular partners liked anal intercourse less than other men, but true withdrawers with casual partners were indistinguishable from those who had unprotected ejaculation. True withdrawers did not differ in sexual adventurousness from other men who had anal intercourse. Most withdrawers avoided anal sex with ejaculation rather than use condoms. Converting them into reliable condom users may be a considerable challenge for health promotion.