A bench study using an artificial lung model was performed to evaluate the snoring detection sensitivity of six (commercially available) auto-nasal continuous positive airway pressure (NCPAP) devices.Snoring was simulated by a loudspeaker connected to the lung model and abruptly activated during 1 s of each inspiratory period to induce pressure oscillation. The oscillation frequencies chosen were 30, 60, 90, and 120 Hz. For each frequency, the amplitude of the pressure oscillation produced by the loudspeaker was adjusted to find the threshold at which the auto-nCPAP devices detected snoring.Differences in pressure-amplitude thresholds of up to three-fold were found across auto-nCPAP devices. A randomized clinical study to compare the effects of the least sensitive (Virtuoso LX; Respironics, Nantes, France) and one of the most sensitive, (Goodknight 418A; Malinckrodt, Nancy, France) devices, in two groups of six patients with obstructive sleep apnoea syndrome was then conducted. Goodknight 418A was more sensitive than Virtuoso LX for detecting snoring (mean¡SD 92¡11% versus 50¡39% respectively, p=0.03).To conclude, striking differences exist between auto-nasal continuous positive airway pressure devices in sensitivity for detecting snoring.