To assess the prevalence and the sociodemographic and behavioral correlates of anal sex in a cohort of HIV-seronegative U.S. women at high risk of HIV exposure, we administered a risk assessment using audio computer-assisted self-interview (A-CASI). Of 1268 sexually active women, 432 (32%) reported anal sex in the previous 6 months. Compared with women who did not report anal sex, those who did had more unprotected vaginal sex (median of 11 versus 7 episodes; p <. 001) and a higher proportion of unprotected sexual (vaginal plus anal) episodes (median of 0.90 versus 0.81; p =.01). Anal sex was reported by higher proportions of women who did not always use condoms, who used crack in the past year, who were </=35 years of age, with no formal education beyond high school, who had had a diagnosed sexually transmitted disease in the previous year, a primary male sex partner, and a male sex partner with a history of injecting use. Women were more likely to report anal sex by A-CASI than during interviewer-administered risk assessments (odds ratio [OR], 9.00; 95% confidence interval [CI], 1.14-71.0). A less biased method of ascertainment may account for the large proportion of women reporting anal sex. Given increased vaginal risk among women reporting anal sex, the relative importance of anal sex in heterosexual transmission merits further study. Behavioral and biomedical prevention strategies effective for anal as well as vaginal sex are needed.