In anesthetized or decerebrate animals, negative pressure applied to the upper airway selectively activates the hypoglossal nerve compared with the phrenic nerve. Conversely, positive pressure reduces hypoglossal nerve activity out of proportion to any change in the phrenic neurogram. We have tested the hypothesis that analogous pressure changes applied to awake humans would selectively inhibit or activate genioglossal electromyographic (EMGge) activity relative to diaphragmatic electromyographic activity (EMGdi). We studied seven normal subjects in a head-out body plethysmograph. Pressure at the mouth was either atmospheric, +10 cmH2O, or -10 cmH2O, and lung volume was held constant by applying an identical pressure to the body surface. Thus the transmural pressure distorting the respiratory system was applied only to the upper airway. Subjects breathed CO2-enriched (2-3%) room air to stimulate phasic respiratory EMGge activity. We found that -10 cmH2O pressure applied selectively to the upper airway resulted in a 49% enhancement of peak-integrated EMGge activity, but EMGdi activity remained at control levels. Positive pressure did not result in any changes in EMGge or EMGdi activity. Neither pressure resulted in significant changes in the magnitude or pattern of ventilation. We conclude that reflex mechanisms maintaining upper airway patency are demonstrable in awake humans and probably have an important role in moment-to-moment modulation of upper airway muscle activity in normal awake humans.