In Parkinson's disease, functional imaging studies during limb motor tasks reveal cerebral activation abnormalities that can be reversed by subthalamic nucleus (STN) stimulation. The effect of STN stimulation on parkinsonian dysarthria has not, however, been investigated using PET. The aim of the present study was to evaluate the effect of STN stimulation on regional cerebral blood flow (rCBF) during speech production and silent articulation in patients with Parkinson's disease. Ten Parkinson's disease patients surgically implanted bilaterally in the STN and with significant improvement of their dysarthria induced by STN stimulation were included. Ten healthy control subjects also participated in this study. Control subjects performed six sessions of [15O]H2O-PET scanning corresponding to three duplicated conditions externally cued by an auditory signal. The conditions were: (i) rest; (ii) production of a short, simple sentence; and (iii) silent articulation of the same sentence. Parkinson's disease patients carried out the six PET sessions twice, i.e., in the ON and OFF STN stimulation states. PET data analysis was performed using statistical parametric mapping (SPM99). In control subjects, speech production (SP) compared with rest was associated with increased rCBF bilaterally in the primary motor cortex (M1) corresponding to the orofacial somatotopy, the supplementary motor area (SMA), the associative auditory cortex and the cerebellar hemispheres. Silent articulation (SA) compared with rest induced a bilateral rCBF increase restricted to the orofacial M1 and cerebellar hemispheres. In Parkinson's disease patients in the OFF stimulation condition, during both SP and SA there was a lack of activation in the right orofacial M1 and in the cerebellum, abnormal increased rCBF in the right superior premotor cortex, and overactivation of the SMA. There was also an abnormal, increased rCBF in the dorsolateral prefrontal cortex (DLPFC) only during SP and increased rCBF in the left insula only during SA. In Parkinson's disease patients ON stimulation, for both SP and SA the activation pattern appeared similar to that in control subjects. In conclusion, our results suggest that parkinsonian dysarthria is associated with altered recruitment of the main motor cerebral regions (orofacial M1, cerebellum), and increased involvement of the premotor and prefrontal cortices (DLPFC, SMA, superior premotor cortex). These abnormal activations are different from those reported during hand motor tasks. They could be a compensatory mechanism, but might also arise directly as part of the pathophysiology of Parkinson's disease. STN stimulation tends to reverse these abnormal activations, which is consistent with the observed improvement of Parkinson's disease dysarthria.