To assess the effects and site of action of a sub-anaesthetic concentration of isoflurane, desflurane and sevoflurane (0.1 minimum alveolar concentration (MAC)) on respiratory control, we measured the ventilatory response to square wave changes in PE1CO2 against a background of normoxia. Using the computer steered "end-tidal forcing system", 2 min of steady state ventilation were followed by a step increase in PE1CO2 (1-1.5 kPa). This level was maintained for 8 min, followed by a step decrease to the original value for another 8 min. Each hypercapnic response was separated into a fast, peripheral component and a slow, central component, characterized by a time constant, carbon dioxide sensitivity, time delay and off-set. We studied 25 healthy volunteers; they performed 2-3 studies without and 2-3 studies during inhalation of the anaesthetic agent. Level of sedation was scored using a subjective seven-point scale from 0 (= alert and awake) to 6 (unrousable). In the isoflurane (16 subjects, 33 control, 37 drug studies) and sevoflurane (15 subjects, 40 control, 41 drug studies) studies, peripheral carbon dioxide sensitivity was reduced by approximately 45% and approximately 27% (ANOVA, P < 0.05 vs control), respectively, without affecting central carbon dioxide sensitivity or apnoeic threshold. In the desflurane study (16 subjects, 36 control, 37 drug studies), no significant effect was observed for any of the variables measured. A significant relation was observed between sedation score and change from control in central carbon dioxide sensitivities in the isoflurane and desflurane studies and in the change in the ratio peripheral carbon dioxide sensitivity over total carbon dioxide sensitivity in the sevoflurane studies. At the highest level of sedation observed (score 3-arousal state comparable with "light sleep"--in three subjects) these latter variables differed significantly from those in the other observed sedation levels (scores 1 and 2-a state of drowsiness). We conclude that 0.1 MAC of isoflurane and sevoflurane depressed the peripheral chemoreflex loop, without affecting the central chemoreflex loop. Desflurane at the same MAC showed no effect on peripheral and central carbon dioxide sensitivity. When the level of sedation was considered, our data suggested that at levels of sedation comparable with sleep, a depressive effect of all three anaesthetics was observed on the central chemoreflex loop.