Hyperventilation is of little clinical relevance unless it causes symptoms. These are often non-specific. Their threshold for onset and relation to steady level of arterial (or its equivalent, end-tidal PCO2; PETCO2) are uncertain, and it has been suggested that they may relate better to the rate of fall of PCO2 than to the absolute level. We investigated this in nine normal subjects, who breathed to and fro through a pneumotachograph into an open circuit in which the concentration of CO2 could be varied. Tidal volume, respiratory frequency and ventilation was measured on-line by a Compaq computer, and PETCO2 at the mouth was measured by capnograph. Subjects overbreathed at a fixed rate and depth until symptoms consisting of dizziness, paraesthesiae and light headedness occurred. Then, without their knowledge and while they continued to overbreathe, inspired CO2 was increased to restore PETCO2 to normal and abolish symptoms, and was then withdrawn again over either approximately 0.1, 2.5 or 5 min until symptoms were again reported. The PETCO2 at this point was noted. The three protocols were performed in each subject in a random order and the same symptoms were reported each time. When averaged across all subjects, symptoms occurred at mean PETCO2 values of 20.3, 19.2 and 18.6 mmHg (2.71, 2.56 and 2.48 kPa), respectively. These were not significantly different, and it can be concluded that there was no influence of rate of fall of PCO2 on threshold for symptoms. Chest pain only occurred in one subject and may have a different mechanism.