A serological study of hepatitis A was carried out in low-income areas scheduled for a major sanitation programme in Rio de Janeiro, Brazil. Blood spots were collected by finger puncture and transported on filter paper, and total antibodies to hepatitis A virus were detected by ELISA. Households were also interviewed to collect information on their environmental conditions and socio-economic status. A generalized linear model using a complementary log–log function was fitted to the data, using the logarithm of age as an explanatory variable to derive adjusted rate ratios (RR). The risk of infection was greater among households with 2–3 members per room (RR = 1·4; 95% CI = 1·04–1·8) or more than three per room (RR = 1·5; 95% CI = 1·2–2·0). People living on hilltops (RR = 1·5; 95% CI = 1·02–2·2), near to open sewers (RR = 1·2; 95% CI = 1·03–1·5) or lacking a kitchen (RR = 1·4; 95% CI = 1·08–1·9) were also at greater risk than others. The number of taps and water-using fittings in the house was associated with a protective effect (RR = 0·9 for each tap; 95% CI = 0·9–0·98). A significant protective association was found with maternal education but not with gender or household income. The results do not suggest a strong association with water quality. Ownership of a ceramic water filter was associated with a protective effect on the margin of significance, but the practice of boiling drinking-water was not, nor was the type of water source used. The results suggest that that the risk of infection with hepatitis A is determined by environmental variables in the domestic and public domains.