A better understanding of the mechanisms involved in human higher cortical functions requires a detailed knowledge of neuronal connectivity between functional cortical regions. Currently no good method for tracking in vivo neuronal connectivity exists. We investigated the inter-areal connections in vivo in the human language system using a new method, which we termed 'cortico-cortical evoked potentials' (CCEPs). Eight patients with epilepsy (age 13-42 years) underwent invasive monitoring with subdural electrodes for epilepsy surgery. Six patients had language dominance on the side of grid implantation and two had bilateral language representation by the intracarotid amobarbital test. Conventional cortical electrical stimulation was performed to identify the anterior and posterior language areas. Single pulse electrical stimuli were delivered to the anterior language (eight patients), posterior language (four patients) or face motor (two patients) area, and CCEPs were obtained by averaging electrocorticograms (ECoGs) recorded from the perisylvian and extrasylvian basal temporal language areas time-locked to the stimulus. The subjects were not asked to perform any tasks during the study. Stimulation at the anterior language area elicited CCEPs in the lateral temporo-parietal area (seven of eight patients) in the middle and posterior part of the superior temporal gyrus, the adjacent part of the middle temporal gyrus and the supramarginal gyrus. CCEPs were recorded in 3-21 electrodes per patient. CCEPs occurred at or around the particular electrodes in the posterior language area which, when stimulated, produced speech arrest. Similar early and late CCEPs were obtained from the basal temporal area by stimulating the anterior language area (three of three patients). In contrast, stimulation of the adjacent face motor area did not elicit CCEPs in language areas but rather in the postcentral gyrus. Stimulation of the posterior language area produced CCEPs in the anterior language (three of four patients) as well as in the basal temporal area (one of two patients). These CCEPs were less well defined. These findings suggest that perisylvian and extrasylvian language areas participate in the language system as components of a network by means of feed-forward and feed-back projections. Different from the classical Wernicke-Geschwind model, the present study revealed a bidirectional connection between Broca's and Wernicke's areas probably through the arcuate fasciculus and/or the cortico-subcortico-cortical pathway. CCEPs were recorded from a larger area than the posterior language area identified by electrical stimulation. This suggests the existence of a rather broad neuronal network surrounding the previously recognized core region of this area.
