A double-blind study was conducted to determine between-laboratory variability in the doubly labeled water method for measurement of total energy expenditure in humans, and to compare the accuracy and precision of three widely-used procedures for calculating rates of carbon dioxide production from the original isotope data. Eighteen laboratories from five countries participated in the study. All laboratories were provided with five water standards containing varying amounts of 2H and 18O, and in addition 11 laboratories were provided with urine and dose specimens from one (six laboratories) or two (five laboratories) healthy elderly subjects of normal height and weight undergoing a calorimetric validation of the doubly labeled water method. The data from the five water standards were analyzed to predict between-laboratory variability in the doubly labeled water technique in all laboratories. In addition, data from the subjects were analyzed using the "slope-intercept", "2-point" and "modified" methods of calculation. The results confirm that the doubly labeled water method can be an accurate technique for the measurement of energy expenditure in adult human subjects in some laboratories. However, there was substantial between-laboratory variability in the results and some laboratories returned physiologically impossible results. There was no significant effect of calculation procedure on the accuracy of the technique in this limited comparison, although the slope-intercept procedure appeared to be more susceptible to analytical error than the other procedures. The isotope standards analyzed by participants in this study will be made available to other investigators on request.