Excess female over male mortality during childhood, well known in the northern Indian subcontinent, is particularly marked in rural Bangladesh. While the determinants of this phenomenon and the respective roles of cultural and economic factors are still debated, little data exist on cause-specific mortality, to identify the specific causes of death producing this differential. In 1986-1987 in Matlab, a study area under intensive demographic surveillance in rural Bangladesh, female children aged 1 to 4 years had a risk of dying 1.8 times higher than male children (95% confidence interval: 1.5-2.1). The causes of death which contributed the most to this excess female mortality were severe malnutrition and diarrhoeal diseases. The risks of dying were 2.5 and 2.1 higher for female than for male children for these two causes, respectively. Possible mechanisms are examined using data on incidence of selected diseases and admission rates to curative facilities. There was no gender difference in incidence of severe diarrhoeal diseases, but female children with diarrhoea were taken to the hospital significantly less often than male children. In contrast, there was a higher incidence of severe malnutrition in female than male children, and a lower rate of hospital admission. The data suggest that gender differentials in mortality may not be as much affected by preventive measures against diarrhoea as by efforts to provide equivalent curative services to female and male children.