We investigated the role of somatosensory feedback on cardioventilatory responses to rhythmic exercise in five men. In a double-blind, placebo-controlled design, subjects performed the same leg cycling exercise (50/100/150/325 ± 19 W, 3 min each) under placebo conditions (interspinous saline, L(3)-L(4)) and with lumbar intrathecal fentanyl impairing central projection of spinal opioid receptor-sensitive muscle afferents. Quadriceps strength was similar before and after fentanyl administration. To evaluate whether a cephalad migration of fentanyl affected cardioventilatory control centers in the brain stem, we compared resting ventilatory responses to hypercapnia (HCVR) and cardioventilatory responses to arm vs. leg cycling exercise after each injection. Similar HCVR and minor effects of fentanyl on cardioventilatory responses to arm exercise excluded direct medullary effects of fentanyl. Central command during leg exercise was estimated via quadriceps electromyogram. No differences between conditions were found in resting heart rate (HR), ventilation [minute ventilation (VE)], or mean arterial pressure (MAP). Quadriceps electromyogram, O(2) consumption (VO(2)), and plasma lactate were similar in both conditions at the four steady-state workloads. Compared with placebo, a substantial hypoventilation during fentanyl exercise was indicated by the 8-17% reduction in VE/CO(2) production (VCO(2)) secondary to a reduced breathing frequency, leading to average increases of 4-7 Torr in end-tidal PCO(2) (P < 0.001) and a reduced hemoglobin saturation (-3 ± 1%; P < 0.05) at the heaviest workload (∼90% maximal VO(2)) with fentanyl. HR was reduced 2-8%, MAP 8-13%, and ratings of perceived exertion by 13% during fentanyl vs. placebo exercise (P < 0.05). These findings demonstrate the essential contribution of muscle afferent feedback to the ventilatory, cardiovascular, and perceptual responses to rhythmic exercise in humans, even in the presence of unaltered contributions from other major inputs to cardioventilatory control.