1. The ventilatory response to isoxic square-wave challenges in end-tidal PCO2 was investigated at three levels of end-tidal PO2 (PET, O2) in nine healthy male subjects. 2. Twenty-seven responses against a background of mild hypoxia (PET, O2 approximately 10 kPa), sixty-seven against a background of normoxia (PET, O2 approximately 14.5 kPa) and seventy-six against a background of hyperoxia (PET, O2 approximately 70 kPa) were collected. 3. The breath-to-breath data were partitioned into a fast and a slow ventilatory component using a two-compartment model. 4. In the normoxic and hypoxic experiments the CO2 sensitivity of the fast component averaged to about 30 and 40% of the total CO2 sensitivity, respectively. In the hyperoxic experiments three subjects had no fast component in their response while in three others the CO2 sensitivity of the fast component averaged to about 24% of the total CO2 sensitivity. In the remaining three subjects the presence of a fast component was doubtful. 5. We argue that the fast component is due to the peripheral chemoreflex loop and the slow component to the central chemoreflex loop. 6. The central CO2 sensitivity and the apnoeic threshold (extrapolated end-tidal CO2 at zero ventilation in the steady state) were 15% smaller in hyperoxia than those in normoxia and hypoxia. In normoxia and mild hypoxia the central CO2 sensitivities were not significantly different. 7. We argue, that apart from peripheral oxygen-carbon dioxide interaction, there is evidence for central oxygen-carbon dioxide interaction in human subjects. 8. We conclude that in general there is a contribution to ventilation of the peripheral chemoreceptors during hyperoxia in man.