We investigated the mechanism of hyperoxic-induced hypercapnia in 17 stable patients with moderate to severe chronic obstructive pulmonary disease (mean FEV1 = 0.95 L and FVC = 2.43 L). Ventilatory and mouth occlusion pressure (P0.1) responses to hypercapnia and hypoxia were measured with standard rebreathing techniques. In a randomized, single-blind fashion, we studied the effect of 15 min of hyperoxia or air on transcutaneous carbon dioxide (PtcCO2), CO2 production (VCO2), total minute ventilation (VE), and calculated dead space to tidal volume ratio (VD/VT). With O2, the PtcCO2 (p less than 0.01) and VD/VT (p less than 0.02) increased. The change in PtcCO2 with O2 was not significantly related to the indices of respiratory drive, nor to the baseline PtcCO2 or SaO2, but was related to the FEV1 (p less than 0.05). The O2 caused a slight decrease in mean VE and mean VCO2, but the effects in individual patients were variable. Both substantial increases or decreases in VE (delta VE) occurred, but these were accompanied by changes in VCO2 (delta VCO2) in the same direction. The effect of changes in VE on PaCO2 is shown to be almost completely cancelled by the concomitant changes in VCO2. Thus, the major portion of the change in PaCO2 was due to changes in VD/VT. We conclude that hyperoxic-induced hypercapnia is primarily due to impairment in gas exchange rather than to depression of ventilation. A reduced FEV1 appears to be a significant risk factor, whereas indices of respiratory drive are not likely to play a major role.