We studied the effect of supplemental nocturnal oxygen on blood gases in 15 patients with severe but stable chronic obstructive lung disease (ratio of forced expired volume in one second to forced vital capacity, 37.2 +/- 1.8 [mean +/- S.E.] per cent of predicted; arterial oxygen tension, 50.7 +/- 1.4 mm Hg; and arterial carbon dioxide tension [PCO2], 53.1 +/- 1.5 mm Hg). Sleep variables and measures of gas exchange were determined on two consecutive nights; on the first night the subjects breathed supplemental oxygen, and on the second they breathed room air. Transcutaneous PCO2 was measured with an infrared sensor, and arterial oxygen saturation with an ear oximeter. Breathing of supplemental oxygen sufficient to keep oxygen saturation at or above 90 per cent was associated with only small increases (less than 6 mm Hg) in PCO2 throughout sleep, as compared with values while subjects were breathing room air. The increase in PCO2 occurred early in the night and was not progressive. Only three patients, who were found to have obstructive sleep apnea in addition to obstructive lung disease, had larger increases in PCO2 during sleep and reported morning headaches. We conclude that nocturnal oxygen does not induce clinically important increases in PCO2 during sleep in patients with stable obstructive lung disease and therefore can safely be used to prevent the dangerous consequences of hypoxia.