Measurement of the pulmonary excretion of trace gases has been used as a simple means of assessing metabolic reactions. End alveolar trace gas concentration, rather than excretory rate, is usually measured. However, the reproducibility of this measurement has received little attention. In 17 healthy subjects, duplicate collections of alveolar air were obtained within 1 minute of each other using a commercially available alveolar air sampler. The concentrations of hydrogen, methane, carbon monoxide, and carbon dioxide were measured. When the subject received no instruction on how to expire into the device, a difference of 28% +/- 19% (1SD) was found between duplicate determinations of hydrogen. Instructing the subjects to avoid hyperventilation or to inspire maximally and exhale immediately resulted in only minor reduction in variability. However, a maximal inspiration held for 15 seconds before exhalation reduced the difference to a mean of 9.6% +/- 8.0%, less than half that observed with the other expiratory techniques. Percentage difference of methane measurements with the four different expiratory techniques yielded results comparable to those obtained for hydrogen. In contrast, percentage differences for carbon monoxide measurements were similar for all expiratory techniques. When normalized to a PCO2 of 5%, the variability of hydrogen measurements with the breath-holding technique was reduced to 6.8% +/- 4.7%, a value significantly lower than that obtained with the other expiratory methods. This study suggests that attention to the expiratory technique could improve the accuracy of tests using breath hydrogen measurements.