This article investigates a new acoustic device to assess the behaviour of the upper airway in patients with OSA. Currently there is no simple non-invasive method to perform such measurements. As such this paper describes the device in probing the patency of the airway during sleep and increasing the efficiency of diagnosing OSA. OBJECTIVE OSA is a common disorder resulting in health and economic burdens. Currently identifying OSA in patients involves expensive techniques that require overnight studies in a laboratory setting with qualified staff. This paper tests a new acoustic device (AirwayClear (AC)) for assessing upper airway patency in human subjects with OSA. We hypothesize that obstructive apnoeas would be detected equally well with AC and polysomnography (PSG). METHODS Twenty-three patients with severe OSA underwent an overnight CPAP titration study. We introduced pseudorandom noise (600-1200 Hz) using AC to the patient's nasal mask during 1 h of subtherapeutic CPAP. AC determined a measure of airway patency based on the level of pseudorandom noise reaching a sternal notch sensor. The ability of AC to detect obstructive respiratory events was compared with standard PSG. RESULTS Three hundred and twenty-two obstructive events (obstructive and mixed apnoeas) were identified by PSG. AC scored 80% as complete obstructions and 16% as partial obstructions. Conversely, AC detected 281 complete obstructions. PSG recognized 84% as apnoeas and scored 11% as hypopnoeas. Of the 204 hypopnoeas identified with PSG, AC indicated the airway was partially or completely obstructed in 69% of patients. A Bland-Altman analysis for the apnoeas from the two measures showed a mean difference of 2.3 events/h and 95% confidence intervals of +/-15.5 events/h. CONCLUSIONS We conclude that AC is able to track airway patency and to identify airway closure in patients with OSA.