We have previously observed that, in normal awake subjects passively hyperventilated with intermittent positive-pressure ventilation delivered through nasal access (nIPPV), the glottis could interfere with the ventilation. We report on data obtained in the same subjects during stable sleep. In all cases, the glottis was continuously observed through a fiber-optic bronchoscope, and other indexes were also continuously recorded. Mechanical ventilation was progressively increased up to 30 l/min. We have observed during passive nIPPV in stable sleep that increases in delivered minute ventilation (VEd) resulted in progressive narrowing of the glottic aperture, with increases in inspiratory resistance and progressive reductions in the percentage of the delivered tidal volume effectively reaching the lungs. For a given level of VEd, comparisons showed that the glottis was significantly narrower during sleep than during wakefulness and that the glottis was significantly narrower during stage 2 than during stages 3/4 non-rapid-eye-movement sleep. Moreover, when CO2 is added to the inspired air, glottic aperture increased in five of nine trials without changes in sleep stage. We also observed a significant negative correlation between glottic width and the VED, independent of the CO2 level. We conclude that during nIPPV glottis narrowing results in a decrease in the proportion of the delivered tidal volume reaching the lungs.