We report a method of measuring the static pressure-volume properties of the upper airway in unanesthetised humans. In 10 normal subjects the pressure in the upper airway was measured during aspiration of air in 10 ml decrements from 30-40 ml above resting volume to 30-40 ml below resting volume, with the glottis closed voluntarily and the oro-pharyngeal muscles relaxed. The pressure-volume properties of the upper airway were well described by the pressure-volume rate upper airway elastance (UAE, cmH2O/ml). The mean (+/- 95% CI) UAE measured at near total lung capacity and with the neck in neutral position for all runs in all subjects was 0.22 +/- 0.06 cmH2O/ml. When UAE was measured repeatedly on different days the greatest change observed was from a mean value of 0.17 cmH2O/ml on one day to 0.25 cmH2O/ml on another, and the standard deviation of the means from different days ranged from 0.01 to 0.04. Neither hypoxia (SaO2 80-85%) nor hyperoxic hypercapnia (PETCO2 65-75 mmHg) had any influence on UAE. Maximal extension of the neck did not affect UAE, but flexion of the neck was associated with a rise in UAE from a mean of 0.23 +/- 0.05 cmH2O/ml to a mean of 0.53 +/- 0.14 cmH2O/ml. UAE was not affected by the lung volume at which measurements were made. We conclude that upper airway pressure-volume properties can be measured non-invasively in man, and that in their linearity, reproducibility and lack of sensitivity to hypoxia and hypercapnia they resemble those of experimental animals.