We studied 18 patients with obstructive sleep apnea (OSA). Each subject slept while breathing through the nose with a specially designed valveless breathing circuit. Low levels of continuous positive airway pressure (CPAP) applied through the nose (2.5-15.0 cmH2O) prevented OSA and allowed long periods of stable stage III/IV sleep and rapid-eye-movement (REM) sleep. Externally applied complete nasal occlusion while the upper airway was patent resulted in upper airway closure during inspiration which was identified by a sudden deviation of nasal pressure from tracheal or esophageal pressure. The level of upper airway closing pressure (UACP) did not change throughout the occlusion test, suggesting that upper airway dilator muscles do not respond to asphyxia during sleep. The upper airway was more collapsible during stage I/II non-rapid-eye-movement (NREM) and REM sleep compared with stage III/IV NREM sleep. The pooled mean UACP was 3.1 +/- 0.4 cmH2O in stage I/II NREM, 4.2 +/- 0.2 cmH2O in stage III/IV NREM, and 2.4 +/- 0.2 cmH2O in REM sleep. Nasal occlusion at successively higher levels of CPAP did not alter the level of UACP in stage I/II NREM and REM sleep but resulted in the upper airway becoming more stable in stage III/IV NREM sleep, suggesting a reflex which augments the tone of upper airway dilator muscles.