1. Respiratory sensation during exercise is generally considered to be related to respiratory mechanical factors which may be manifest as an abnormal relationship between the force applied to the lungs and chest wall and the resulting motion (if any); that is, a 'length-tension' inappropriateness (Campbell & Howell, 1963). This suggests that there should be a direct correlation between ventilation (VE) and the associated intensity of the perceived sensation, such that the sensation associated with a particular level of VE should remain essentially constant regardless of the source of respiratory stimulation. 2. In order to establish whether certain respiratory stimuli might be 'dyspnoeagenic' (i.e. capable of evoking an intensity of respiratory sensation out of proportion to their influence on VE), we investigated the influence of both peripheral chemoreflex activation (induced by isocapnic hypoxia) and central chemoreflex activation (induced by hypercapnic hyperoxia) on the intensity of respiratory sensation in seven healthy adults during moderate cycle ergometer exercise (i.e. below the lactate threshold, theta 1ac). 3. In each test, an 'isopnoea' was established for which a particular level of VE was sustained over a prolonged period (approximately 30 min) while the proportional contributions to the ventilatory drive from either exercise and the peripheral chemoreflex or from exercise and the central chemoreflex were slowly altered to new stable levels, without the subject's knowledge, VE, tidal volume, inspiratory and expiratory durations, mean inspiratory flow, and end-tidal PCO2 and PO2 (PET,CO2, PET,O2) were monitored breath-by-breath. The intensity of respiratory sensation was rated with a visual analogue scale. 4. Isopnoeic ratings of respiratory sensation were systematically greater for peripheral chemoreflex activation by isocapnic hypoxia during exercise at 50% theta 1ac (for which the degree of peripheral chemoreflex activation, estimated by hyperoxic transition or 'Dejours' testing, averaged approximately 23% of the total VE), compared to 90% theta 1ac during isocapnic hyperoxia. Ratings during exercise at 50% theta 1ac for central chemoreflex activation by hypercapnic hyperoxia were not systematically different from 90% theta 1ac during isocapnic hyperoxia, however. 5. As VE was stable throughout each isopnoea and the MVV (maximum voluntary ventilation) was uninfluenced by the test condition, the dyspnoea index (VE x 100/MVV) was not affected. Breathing pattern was also unaffected. 6. We conclude that in normal subjects exercising moderately, activation of the peripheral chemoreceptors by isocapnic hypoxia evokes an intensity of respiratory sensation which is out of proportion to that evoked by an isopnoeic stimulation of the central chemoreceptors with hypercapnic hyperoxia at the same level of exercise.(ABSTRACT TRUNCATED AT 400 WORDS)