The clinical, electrophysiological and haemodynamic effects of precentral gyrus stimulation (PGS) as a treatment of refractory post-stroke pain were studied in 2 patients. The first patient had a right hemibody pain secondary to a left parietal infarct sparing the thalamus, while the second patient had left lower limb pain developed after a right mesencephalic infarct. In both cases, spontaneous pain was associated with hyperpathia, allodynia and hypoaesthesia in the painful territory involving both lemniscal and extra-lemniscal sensory modalities in patient 1, extra-lemniscal sensory modality only in patient 2. Both patients were treated with electrical PGS by means of a 4-pole electrode, the central sulcus being per-operatively located using the phase-reversal of the N20 wave of somatosensory evoked potentials. No sensory side effect, abnormal movement or epileptic seizure were observed during PGS. The analgesic effects were somatotopically distributed according to the localization of electrode on motor cortex. A satisfactory long-lasting pain control (60-70% on visual analog scale) as well as attenuation of nociceptive reflexes were obtained during PGS in the first patient. Pain relief was less marked and only transient (2 months) in patient 2, in spite of a similar operative procedure. In this patient, in whom PGS eventually evoked painful dysethesiae, no attenuation of nociceptive RIII reflex could be evidenced during PGS. Cerebral blood flow (CBF) was studied using emission tomography (PET) with O-labeled water. The sites of CBF increase during PGS were the same in both patients, namely the thalamus ipsilateral to PGS, cingulate gyrus, orbito-frontal cortex and brainstem. CBF increase in brainstem structures was greater and lasted longer in patient 1 while patient 2 showed a greater CBF increase in orbito-frontal and cingular regions. Our results suggest that PGS-induced analgesia is somatotopically mediated and does not require the integrity of somatosensory cortex and lemniscal system. PGS analgesic efficacy may be mainly related to increased synaptic activity in the thalamus and brainstem while changes in cingulate gyrus and orbito-frontal cortex may be rather related to attentional and/or emotional processes. The inhibitory control on pain would involve thalamic and/or brainstem relays on descending pathways down to the spinal cord segments, leading to a depression of nociceptive reflexes. Painful dysesthesiae during stimulation have to be distinguished from other innocuous sensory side effects, since they may compromise PGS efficacy.