Objectives To measure the extent of dilution of helium-oxygen (heliox) by room air when given via high concentration reservoir mask to spontaneously breathing subjects. Substantial dilution of heliox by room air under these circumstances might alter its physical properties sufficiently to negate any potential clinical benefit in obstructive respiratory failure. Design Healthy volunteers breathing different concentrations of helium in oxygen via two different masks in a randomised crossover design. Setting Operating theatre in a university hospital. Patients and participants Six healthy volunteers. Interventions The concentrations of helium, nitrogen and oxygen were measured in the trachea of each volunteer using a mass spectrometer during normal breathing, hyperventilation and hypoventilation. Measurements and results During normal breathing of Heliox21 (79% helium) via a standard non-rebreathe reservoir mask, within subject median percentage tracheal helium was 37.2% (range 29.3–52.2%) and nitrogen was 41.7% (27.4–49.4%). Air entrainment was affected by changes in breathing pattern: tracheal nitrogen concentration was greater during hyperventilation (55.4%; range 49.4–63.5%) and less during hypoventilation (33.1%; range 24.6–39.6%, p  = 0.043). Tracheal nitrogen could be almost completely abolished by administering heliox via a tightly fitting cushioned facemask, even during hyperventilation (2.2%; range 0.6–6.1%, p  = 0.028). Conclusions Heliox administration via a standard high-concentration reservoir mask leads to significant dilution by room air. For the full potential benefits of heliox to be realised in spontaneously breathing patients, it should be administered via a system that achieves a gas tight seal, with no leaks between the delivery device and the surroundings.