A recently developed CO2 pulse technique was used to test for ventilatory sensitivity to CO2 in four normal men following 2 min voluntary hyperventilation down to an end-tidal CO2 tension (PETCO2) of 20 mmHg (2.7 kPa). Pure CO2 was injected into the inspiratory limb of a breathing circuit at 0.4 l.min-1 for 30 s and any small ventilatory response was detected against background noise by ensemble-averaging of multiple runs. Following hyperventilation, ventilation was initially often above control and apnoea was not seen. In one subject, the ventilatory response to the CO2 pulse was barely detectable either before or after hyperventilation. In another subject, there was a response to pulses given before hyperventilation and 3 and 5.5 min after hyperventilation but not 30 s after hyperventilation when PETCO2 was about 25 mmHg (3.3 kPa) and rising. In the two remaining subjects ventilatory responses were seen to CO2 pulses started 30 s after hyperventilation, although PETCO2 following the pulse remained some 5 mmHg (0.7 kPa) below baseline. We conclude that in some subjects the PETCO2 threshold lies well below the normal PETCO2. The technique is tedious for the experimental subject because of the large number of repetitions required and, therefore, unsuitable for a study on a large number of subjects.