The ventilatory response to mild-to-moderate exercise in humans is isocapnic, or 'error-free'. It has been suggested that this response is learned over many repetitions of exercise through the process of minimising any deviations from normal in the blood gas tensions, as sensed by the chemoreceptors. However, relatively limited training programmes in humans have failed to produce any convincing evidence that forcibly altering the blood gas tensions during repeated periods of exercise alters the subsequent steady-state ventilatory response to exercise. In this study, 8 healthy young subjects were exposed, over a 7-day training period, to a total of 70 repeated bouts of exercise paired with a simultaneous airway CO2 load to stimulate the chemoreceptors (Protocol EX+CO2). The ventilatory response to exercise was measured before and after training to determine whether it had been modified. Two further training protocols were undertaken as controls. One employed repeated exercise without an airway CO2 load, and the other employed repeated airway CO2 loading without exercise. On the first and second days following training with Protocol EX+CO2, end-tidal PCO2 was regulated at a lower level during steady-state exercise than following training with the control protocols and than before training (reduction in end- tidal PCO2 = 1.32±0.36 Torr, mean±SE, ANOVA, p<0.05). In contrast to previous studies, this finding demonstrates that the steady-state ventilatory response to exercise can be modified by a prior period of altered chemoreception during exercise. This suggests that ventilation is matched to metabolic rate during exercise by a mechanism that involves learning and memory.