Respiratory sinus arrhythmia (RSA) is used as a noninvasive measure of vagal cardiac input, but its causative mechanisms in humans remain undetermined. We compared the RSA of five lung-denervated double-lung transplant patients with intact hearts to six normal (N) control subjects, five heart-denervated patients, and two liver transplant patients at matched tidal volumes (VT's) and breathing frequencies. In N and liver transplant subjects, RSA was significant during eupnea and increased two- to threefold with increasing VT and inspiratory effort. In heart- and lung-denervated subjects, RSA at eupnea was significant but was only 53% of that in N subjects and was not respondent to changing VT, inspiratory effort, or breathing frequency. We also compared the RSA of N subjects during voluntary (active) and passive positive pressure ventilation at normocapnia. RSA was reduced from 11 +/- 2.2 beats/min during active ventilation to 5.4 +/- 0.8 beats/min during PPV. We conclude that vagal feedback from pulmonary stretch receptors is obligatory for the generation of a neurally mediated RSA in awake humans at normal and raised levels of VT and respiratory motor output. In intact humans, we also hypothesize an important effect for nonpulmonary central and/or peripheral modulation of RSA. It is likely that the key mechanisms for neurally mediated RSA in unanesthetized humans are mutually dependent.