The performance of the Hudson Multi-Vent oxygen mask was assessed by attaching it to a model of a head and generating varying inspiratory flows of oxygen through the mouth and nasal orifices. The oxygen concentration that was delivered by the mask varied with the inspiratory flow rate at all oxygen concentrations. This effect was proportionately similar at all levels of set oxygen concentration: about 50% of the increment of inspired oxygen was lost as the inspiratory flow rate increased from 20 L/min to 100 L/min. The performance of the mask was not affected by the route (oral or nasal) of inspiratory flow. Rotation of the mask in the coronal plane produced marked changes in delivered oxygen concentration; with the oxygen concentration set at 50%, and the inspiratory flow rate at 40 L/min, the delivered oxygen concentration was 45% when the mask was parallel to the coronal plane of the model, was 41% when it was angled 10 degrees towards the face, and was 27% when it was angled 10 degrees away from the face. The measurement of oxygen concentrations at different points inside the mask confirmed that the oxygen-enriched gas is directed in a localized stream in the centre of the mask, and that air is entrained through the large ports that are sited at the sides of the mask to satisfy inspiratory flow. The measurement of inspired oxygen concentrations in 12 hospital inpatients confirmed that the model reflected accurately the performance of the mask when used in a routine way. We conclude that, in clinical use, the Hudson Multi-Vent mask is unlikely to deliver the set oxygen concentration, and that the delivered oxygen concentration is not predictable.