Measurement of the ventilatory response of the newborn to carbon dioxide using the rebreathing method usually involves raising the concentration of oxygen in the inspired gas in order to prevent hypoxaemia developing during the test. Described here is a modified rebreathing method which avoids the need to use hyperoxic gas mixtures. Subjects rebreathed 5-7% carbon dioxide in air for 1.5-3 min; hypoxaemia was prevented by the addition of oxygen to the rebreathing circuit at the rate of approximately 7 ml/min per kg body weight. Arterial oxygen saturation (measured by a pulse oximeter) was maintained at approximately 95% throughout the tests. The technique was successfully used to study the ventilatory responses of eight full-term and 4 pre-term babies during behaviorally defined rapid-eye movement and quiet sleep. The reproducibility of the ventilatory response measured by this method was evaluated by repeated testing of each subject. This revealed considerable variability in the magnitude of the response from test to test. Some of the factors which might contribute to this variability are discussed.