1. We utilized a chronically tracheostomized, unanaesthetized dog model to study the reflex effects on inspiratory motor output of low-amplitude, high-frequency pressure oscillations (HFPOs) applied to the isolated upper airway (UA) during stable non-rapid eye movement (NREM) sleep. 2. HFPOs (30 Hz and +/-2 to +/-4 cmH2O) were applied via a piston pump during eupnoea, inspiratory resistive loading and tracheal occlusion. 3. When applied to the patent UA during expiration, and especially during late expiration, HFPOs prolonged expiratory time (TE) and tonically activated the genioglossus muscle EMG. When applied to the patent UA during inspiration, HFPOs caused tonic activation of the genioglossus muscle EMG and inhibition of inspiratory motor output by either: (a) a shortening of inspiratory time (TI), as inspiration was terminated coincident with the onset of HFPOs; or (b) a prolonged TI accompanied by a decreased rate of rise of diaphragm EMG and rate of fall of tracheal pressure. These effects of HFPOs were observed during eupnoea and inspiratory resistive loading, but were maximal during tracheal occlusion where the additional inhibitory effects of lung inflation reflexes were minimized. 4. During eupnoea, topical anaesthesia of the UA abolished the HFPO-induced prolongation of TE, suggesting that the response was mediated primarily by mechanoreceptors close to the mucosal surface; whereas the TE-prolonging effects of a sustained square wave of negative pressure (range, -4.0 to -14.9 cmH2O) sufficient to close the airway were preserved following anaesthesia. 5. These results demonstrate that high-frequency, low-amplitude oscillatory pressure waves in the UA, similar to those found in snoring, produce reflex inhibition of inspiratory motor output. This reflex may help maintain UA patency by decreasing the collapsing pressure generated by the inspiratory pump muscles and transmitted to the UA.