Objective: To investigate the relationship between HIV seropositivity and bacterial vaginosis (BV) in a population at high risk for sexual acquisition of HIV. Design: A cross-sectional study was conducted among 144 female commercial sex workers in Chiang Mai, Thailand. Methods: The participants were tested for cervical gonorrhea and Chlamydia infection, syphilis, Trichomonas vaginitis, Candida vaginitis, BV, and HIV infection. BV was diagnosed by clinical criteria (pH > 4.5, positive amine test, and presence of clue cells) and using Gram stains. Results: Thirty-three per cent of participants had BV, and 43% were HIV-positive. Using clinical criteria, the association of BV and HIV seropositivity was significant [odds ratio (OR), 2.7; 95% confidence interval (CI), 1.3-5.0]. Although the association between BV and HIV prevalence was not significant using Gram stains alone for diagnosis of BV, an association was found between abnormal vaginal flora and HIV (OR, 2.1; 95% CI, 1.0-4.8). In multiple logistic regression analysis, adjusting for age, number of sexual encounters per week, current condom use, and currently having a sexually transmitted disease (STD), both BV and a history of an STD were independently associated with HIV seropositivity (adjusted OR for BV, 4.0 and 95% CI, 1.7-9.4; adjusted OR for history of an STD, 6.9 and 95% CI, 2.1-22.9). Conclusions: When diagnosed clinically, BV is independently associated with HIV seroprevalence. HIV infection may promote abnormal vaginal flora, or BV may increase susceptibility to sexual transmission of HIV. Alternatively, the association seen here may result from intervening variables; in this case BV may be a marker or a cofactor of HIV transmission.