Total body water was measured by H218O stable isotope dilution in two groups: in premature infants without complications, who were studied from 8 d of age until discharge; and in Peruvian subjects aged 6–36 mo, who were in the long-term convalescent stage of recovery from malnutrition. Results indicated that reliable total body water estimates can be obtained from sample volumes as small as 50μl of urine or plasma using a gas-isotope-ratio mass spectrometer equipped with an automated purification inlet system. Results from 21 studies in 10 Peruvian infants indicated substantially completed isotope equilibration in plasma by 2 h after the dose; total body water estimates from the 2-h samples averaged 98.7%(±4.1) of 6-h values. Samples obtained at 4-h postdose gave total body water estimates that averaged 99.0% (± 2.9) of the 6-h value, showing essentially complete equilibration and reduced variability. Total body water estimates from urine samples collected 3–5 h postdose were closely correlated with 6-h, plasma-based total body water values in both premature and older infants; however, some reduction in variability was observed when urine collection was extended to 5–7 h, at which time urine-based estimates averaged 98.8%(±2.0) and 100.7% (±3.1) of plasma-based values for prematures and older Peruvian infants, respectively. The correlation between 5–7 h urine-based estimates of total body water with plasma-based values was r = 0.96 for 30 studies in prematures and r = 0.99 for 57 studies in older Peruvian infants. Data points adhered closely to the lines of identity in both study groups. These results suggest that noninvasive urine sampling techniques can be substituted for plasma sampling in body water studies in infants.