A case-control study, aimed at identifying factors associated with rotavirus diarrhoea cases presenting to health facilities, was conducted in children from low-income and middle-low-income families in Brazil. Cases were 390 children with diarrhoea and rotavirus in stools; controls were 1674 children without diarrhoea presenting to the same facilities. Data were collected by questionnaire and observations during home visits. Explanatory variables were grouped according to a conceptual model of causation. The ORs by non-conditional logistic regression and population-attributable fractions were calculated. Socioeconomic factors contributed a third of cases, followed by contact with diarrhoea cases and by not being breast fed. In cases aged <1 year, not being breast fed was the main determinant, followed by socioeconomic factors, and crowding and contact outside the home; in older children, socioeconomic factors followed by contact inside and outside the home were the main determinants. Environmental and sanitation variables were not associated with diarrhoea in the final model, and socioeconomic factors were only partly mediated by proximal variables. Transmission of rotavirus appears to be mostly by person-to-person contact, and shows marked social differentials not explained by the biological factors studied. The rotavirus vaccine is unlikely to protect against the full range of circulating genotypes of rotavirus, and understanding rotavirus epidemiology remains essential to the development of control policies.