Breast-milk intakes in infants can be estimated from measurements of water turnover rates after the oral administration of small doses of deuterium oxide (2H2O) to the infant or its mother. However, when the former method is used it is not possible to distinguish water intakes from milk and those from other sources. Giving a 2H2O dose to the mother avoids this problem. The theoretical basis for these techniques is examined and useful compromises between the requirements of theory and practice are suggested. Results obtained for breast-milk intakes of Gambian and Papua New Guinean children are used to illustrate the potential value of the procedures.