Repeated inspiratory occlusions in humans elicit respiratory-related cortical potentials, the respiratory counterpart of somatosensory-evoked potentials. These potentials comprise early components (stimulus detection) and late components (cognitive processing). They are considered as the summation of several afferent activities from various part of the respiratory system. This study assesses the role of the upper airway as a determinant of the early and late components of the potentials, taking advantage of the presence of a tracheotomy in patients totally or partially deafferented. Eight patients who could breathe either through the mouth or through a tracheotomy orifice (whole upper airway bypassed) were studied (4 quadriplegic patients with phrenic pacing, 4 patients with various sources of inspiratory pump dysfunction). Respiratory-related evoked potentials were recorded in CZ-C3 and CZ-C4. They were consistently present after mouth occlusions, with a first positive P1 and a first negative N1 components of normal latencies (P1: 40.4 +/- 6.1 ms in CZ-C3 and 47.6 +/- 7.6 ms in CZ-C4; N1: 84.4 +/- 27.1 ms in CZ-C3 and 90.2 +/- 17.4 ms in CZ-C4) and amplitudes. Tracheal occlusions did not evoke any cortical activity. Therefore, in patients with inspiratory pump dysfunction, the activation of upper airway afferents is sufficient to produce the early components of the respiratory-related evoked cortical potentials. Per contra, in this setting, pulmonary afferents do not suffice to evoke these components.