To gain an insight into the origin of the phase I ventilatory response to exercise (ph I) in humans, pulmonary ventilation (VE) and end-tidal partial pressures of oxygen and carbon dioxide (PETO2 and PETCO2, respectively) were measured breath-by-breath in six male subjects during constant-intensity exercise on the cycle ergometer at 50, 100 and 150 W, with eupnoeic normocapnia (N) or hyperpnoeic hypocapnia (H) established prior to the exercise test. Cardiac output (Qc) was also determined beat-by-beat by impedance cardiography on eight subjects during moderate exercise (50 W), and the CO2 flow to the lungs (Qc.Cv-CO2 where Cv-CO2 is concentration of CO2 in mixed veneous blood) was estimated with a time resolution of one breathing cycle. In N, the initial abrupt increase of VE during ph I (delta VE approximately 18 1.min-1 above rest) was followed by a transient fall. When PETCO2 started to increase (and PETO2 decreased) VE increased again (phase II ventilatory response, ph II). In H, during ph I delta VE was similar to that of N. By contrast, during ph II delta VE kept gradually decreasing and started to increase only when PETCO2 had returned to approximately 40 mmHg (5.3 kPa). Thus, as a result of the prevailing initial conditions (N or H) a temporal shift of the time-course of VE during ph II became apparent. No correlation was found between CO2 flow to the lungs and VE during ph I. These results are interpreted as suggesting that an increased CO2 flow to the lungs does not constitute an important factor for the initial hyperventilatory response to exercise. They are rather compatible with a neural origin of ph I, and would support the "neurohumoral" theory of ventilatory control during exercise.