Oxygen is the commonest drug prescribed in hospitals. The inhaled concentration is altered by the administered oxygen flow rate, the characteristics of the delivery device and the patient's respiratory pattern. Using healthy volunteers we measured the inspired oxygen concentration achieved with different devices both at rest and when the breathing pattern of respiratory failure was simulated by binding the subjects' chests until the forced expiratory volume in 1 s was reduced by > 50% and the respiratory rate was > 25 breaths.min(-1). With this respiratory pattern, there was a statistically significant fall in F(I)o(2) while administering oxygen via a Hudson mask at 4 l.min(-1) (23.8% (95% CI 17.4-30.3%) reduction), 12 l.min(-1), humidified (17.8% (95% CI 8.8-26.7%) reduction) and 24 l.min(-1), humidified (12.2% (95% CI 5.0-19.3%) reduction). There was no statistically significant change with a nonrebreathing (reservoir) mask at 15 and 110 l.min(-1) or with a Vapotherm 2000i at 40 l.min(-1), humidified, via nasal prongs. We conclude that the F(I)o(2) delivered by high flow devices is unaffected when the breathing pattern of respiratory failure is simulated. The F(I)o(2) achieved at rest by a nonrebreathing mask (0.68) is less than that often quoted in the literature.