Gender-based violence and gender inequality are increasingly cited as important determinants of women's HIV risk; yet empirical research on possible connections remains limited. No study on women has yet assessed gender-based violence as a risk factor for HIV after adjustment for women's own high-risk behaviours, although these are known to be associated with experience of violence. We did a cross-sectional study of 1366 women presenting for antenatal care at four health centres in Soweto, South Africa, who accepted routine antenatal HIV testing. Private face-to-face interviews were done in local languages and included assessement of sociodemographic characteristics, experience of gender-based violence, the South African adaptation of the Sexual Relationship Power Scale (SRPS), and risk behaviours including multiple, concurrent, and casual male partners, and transactional sex. After adjustment for age and current relationship status and women's risk behaviour, intimate partner violence (odds ratio 1.48, 95% CI 1.15-1.89) and high levels of male control in a woman's current relationship as measured by the SRPS (1.52, 1.13-2.04) were associated with HIV seropositivity. Child sexual assault, forced first intercourse, and adult sexual assault by non-partners were not associated with HIV serostatus. Women with violent or controlling male partners are at increased risk of HIV infection. We postulate that abusive men are more likely to have HIV and impose risky sexual practices on partners. Research on connections between social constructions of masculinity, intimate partner violence, male dominance in relationships, and HIV risk behaviours in men, as well as effective interventions, are urgently needed.