Many clinical and research situations require maintenance of isocapnia, which occurs when alveolar ventilation (V'A) is matched to CO2 production. A simple, passive circuit that minimizes changes in V'A during hyperpnoea was devised. It is comprised of a manifold, with two gas inlets, attached to the intake port of a nonrebreathing circuit or ventilator. The first inlet receives a flow of fresh gas (CO2=0%) equal to the subject's minute ventilation (V'E). During hyperpnoea, the balance of V'E is drawn (inlet 2) from a reservoir containing gas, the carbon dioxide tension (PCO2) approximates that of mixed venous blood and therefore contributes minimally to V'A. Nine normal subjects breathed through the circuit for 4 min at 15-31 times resting levels. End-tidal PCO2 (Pet,CO2) at rest, 0, 1.5 and 3.0 min were (mean+/-SE) 5.1+/-0.1 kPa (38.1+/-1.1 mmHg), 4.9+/-0.1 kPa (36.4+/-1.1 mmHg), 5.0+/-0.2 kPa (37.8+/-1.6 mmHg) and 5.0+/-0.2 kPa (37.6+/-1.4 mmHg) (p=0.53, analysis of variance (ANOVA)), respectively; without the circuit, Pet,CO2 would be expected to have decreased by at least 2.7 kPa (20 mmHg). Six anaesthetized, intubated dogs were first ventilated at control levels and then hyperventilated by stepwise increases in either respiratory frequency (fR) from 10 to 24 min(-1) or tidal volume (VT) from 400 to 1,200 mL. Increases in fR did not significantly affect arterial CO2 tension (Pa,CO2) (p=0.28, ANOVA). Only the highest VT decreased Pa,CO2 from control (-0.5 +/- 0.3 kPa (-3.4 +/- 2.3 mmHg), p<0.05). In conclusion, this circuit effectively minimizes changes in alveolar ventilation and therefore arterial carbon dioxide tension during hyperpnoea.