To determine sociodemographic risk factors associated with HIV-1 infection in a rural Ugandan population. A population-based survey. All adult residents (aged > or = 13 years) in a cluster of 15 neighbouring villages of the Masaka District of south-west Uganda were invited to participate in a sociodemographic and serological survey. Questions relating to sexual behaviour were asked separately in an accompanying case-control study. Socioeconomic data and an unambiguous HIV-1 serostatus were obtained by house-to-house survey for 3809 (72%) of the adult population. The association between serostatus and the following variables were analysed: age, sex, marital status, tribe, religion, education, occupational group, place and frequency of travel and recent history of sexually transmitted disease. Women aged 13-21 years were at a much higher risk than men of the same age [odds ratio (OR), 8.6; 95% confidence interval (CI), 3.0-24.5]. Married people aged < 25 years were twice as likely to be infected as those who were not currently married (OR, 2.3; 95% CI, 1.5-3.7). In contrast, in those aged > or = 25 years, women were at a lower risk than men (OR, 0.72; 95% CI, 0.52-0.98) as were those who were currently married compared with those who were not (OR, 0.47; 95% CI, 0.34-0.64). In both age groups those with a history of a recent genital ulcer were approximately three times more likely to be infected. Muslims had lower risks than non-Muslims (OR, 0.58 for both age groups). The people most at risk of HIV-1 infection in this rural Ugandan population are young married women who had, presumably, commenced sexual activity recently.