The tracer-to-infant deuterium dilution method for the measurement of milk intake was evaluated in twenty breast-fed and twenty formula-fed infants. The isotope method was compared with conventional direct-weighing techniques. Human milk intake was assessed by 5 d test-weighing. Intakes of formula, supplemental foods, and water were determined by pre- and post-weighing of feeding bottles. An oral dose of 200 mg 2H2O/kg body-weight was given to each infant, and urine was sampled daily for 14 d. 2H enrichment of the urine was measured by gas-isotope-ratio mass spectrometry. Milk intakes estimated from the deuterium dilution method were consistently higher than those from direct-weighing; the mean difference between methods was 106 (SD 47) g/d or 14% for the breast-fed group and 70 (SD 155) g/d or 8% for the formula-fed group. Estimates of intake for some infants varied substantially between the two methods of measurement. When the estimated values of human milk intake were corrected for environmental water influx and insensible water loss during breast-feeding, the relative bias decreased to 5%. Correction of the estimated values of formula intake for environmental water influx decreased the relative bias to 1-2%. The acceptability of the deuterium dilution method to determine milk intake depends on the goals and the tolerance for error in group and individual intake estimates of a given study.