Background: We have shown that minute ventilation [VE] and end-tidal CO2 tension [PETCO2] were determined by the interaction between the properties of controller and plant. During exercise, the controller shifted to the direction of decreased PETCO2, so as to compensate for the shift of plant accompanying increased metabolism. This effectively fixes PETCO2 in the normal range, with the expense of exercise hyperpnea. We examined how athletes are trained to reduce this exercise hyperpnea. Methods: In 6 trained (Tr) and 6 untrained (UT) healthy males, to characterize the controller, we induced hypercapnia by changing inspiratory CO2 fraction and measured the linear PETCO2-VE relation (VE = S·(PETCO2 - B)). To characterize the plant, we made subjects alter VE and measured the hyperbolic VE-PETCO2 relation (PETCO2 = A / VE + C). We characterized these relations both at rest and during light exercise. Results: Physical conditioning did not affect characteristics of either controller or plant during rest. Exercise decreased B in UT, while not in Tr. During exercise, slope S slightly increased in Tr. The hyperbolic plant property shifted right and upward during exercise as predicted by increased metabolism. Though constant C was slightly lower in Tr than that in UT (p<0.05), this does not contribute much to changes in VE. The VE during exercise in Tr was by 22 % lower than that in UT (P<0.01). Conclusion: The attenuation of exercise hyperpnea induced by regular exercise training results mainly from the adaptation of the controller, the lower sensitization of chemoreflex controller. Strenuous regular exercise training almost abolishes the exercise-induced shift of central ventilation controller.