Air-puff stimulation applied to a fingertip is known to exert a location-specific facilitatory effect on the size of the motor evoked potentials elicited in hand muscles by transcranial magnetic stimulation. In order to clarify its nature and the pathway responsible for its generation, we studied 27 patients with discrete lesions in the brain (16, 9 and 2 patients with lesions in the cerebral cortex, thalamus and brainstem, respectively). Facilitation was absent in patients with lesions affecting the primary sensorimotor area, whereas it was preserved in patients with cortical lesions that spared this area. Facilitation was abolished with thalamic lesions that totally destroyed the nucleus ventralis posterolateralis (VPL), but was preserved with lesions that at least partly spared it. Lesions of the spinothalamic tract did not impair facilitation. The size of the N20-P25 component of the somatosensory evoked potential showed a mild correlation with the amount of facilitation. The facilitation is mainly mediated by sensory inputs that ascend the dorsal column and reach the cortex through VPL. These are fed into the primary motor area via the primary sensory area, especially its anterior portion, corresponding to Brodmann areas 3 and 1 (possibly also area 2), without involving other cortical regions. The spinothalamic tract and direct thalamic inputs into the motor cortex do not contribute much to this effect. Some patients could generate voluntary movements despite the absence of the facilitatory effect. The present method will enable us to investigate in humans the function of one of the somatotopically organized sensory feedback input pathways into the motor cortex, and will be useful in monitoring ongoing finger movements during object manipulation.