We tested the hypothesis that features of upper airway and tracheal geometry can be inferred from acoustic reflection data recorded at the mouth. In six subjects we computed inferences of airway cross-sectional area vs. distance and compared them with measurements obtained from orthogonal radiographic projections of the trachea. The acoustic data show local area maxima at the uvula and hypopharynx and local minima at the oropharynx and the glottis. With subjects breathing air the inferred tracheal areas markedly exceeded the radiographic measurements. With subjects breathing 80% He-20% O2 there was good intrasubject agreement between acoustic and radiographic data in spite of large intersubject variability. The average coefficient of variation of tracheal area determinations for five trials in all subjects was 0.16. These studies suggest that features of airway geometry between the mouth and carina can be determined accurately and noninvasively in individual subjects from high-frequency reflection data measured at the mouth.