Dyspnoea is a frequent and devastating symptom of advanced cancer. The purpose of this prospective, double-blind, crossover trial was to assess the effects of supplemental oxygen on the intensity of dyspnoea. 14 patients with hypoxaemic dyspnoea due to advanced cancer were randomised to receive either oxygen or air; the gases were delivered at 5 L/min by mask. After 5 min of stable oxygen saturation (pulse oximetry), patients were crossed over to receive the other treatment. The crossover was repeated twice. Dyspnoea was assessed with a visual analogue scale (O = no dyspnoea, 100 = worst dyspnoea). Mean difference in dyspnoea visual analogue scale between air and oxygen treatment was 20.5 (95% confidence interval 13.5 to 27.6). 12 patients consistently preferred oxygen to air; similarly, the investigator consistently chose oxygen for the same 12 patients. In a global rating questionnaire, patients reported little or no benefit during the air phase compared with moderate to much benefit during the oxygen phase. We conclude that oxygen is beneficial to patients with hypoxia and dyspnoea at rest.