The objectives of this study were to determine predictors of consistent condom use in heterosexual relationships of young adults who use hard drugs in a neighbourhood with widespread drug-use-connected HIV. We interviewed 196 18-24 year olds who injected drugs or used heroin, cocaine or crack in the prior year and lived in the Bushwick neighbourhood of New York City. Interviews covered sociodemographics, substance use and sexual networks. The unit of analysis is the relationship; the dependent variable measures consistent condom use over the prior 30 days in a given relationship. Consistent condom use was reported in 26% of 377 non-commercial relationships and in all of 22 commercial relationships. Using multiple logistic regression, consistent condom use in non-commercial relationships was more likely in relationships that are not 'very close'; for men (but not women) with peers whose norms are more favourable to condom use; and for subjects who had concurrent sex partners in the last 12 months. In conclusion, we found that: (1) the lack of relationship between the peer norms of drug-using women and their condom use suggests they may have little control over condom use in their relationships-programmes should attempt to empower young women drug users and to develop ways for their peers to influence the men in their lives; (2) epidemiologically, the positive association of concurrency to consistent condom use suggests that condom use may be restricting HIV spread through the community-the presence of consistent condom use in all of the commercial sexual relationships also may restrict HIV spread; (3) prevention efforts should attempt to change peer cultures as a way to develop self-sustaining risk reduction. These changes should include changes in gender roles and power relations.