Assessment of eloquent functions during brain mapping usually relies on testing reading, speech, and comprehension to uncover transient deficits during electrical stimulation. These tests stem from findings predicted by the Geschwind-Wernicke hypothesis of receptive and expressive cortices connected by white matter tracts. Later work, however, has emphasized cortical mechanisms of language function. The authors report two cases that demonstrate that conduction aphasia is cortically mediated and can be inadequately assessed if not specifically evaluated during brain mapping. To determine the distribution of language on the dominant cortex, electrical cortical stimulation was performed in two cases by using implanted subdural electrodes during brain mapping before epilepsy surgery. A transient isolated deficit in repetition of language was reported during stimulation of the posterior portion of the dominant superior temporal gyrus in one patient and during stimulation of the supramarginal gyrus in the other patient. These cases demonstrate a localization of language repetition to the posterior perisylvian cortex. Brain mapping of this region should include assessment of verbal repetition to avoid potential deficits resembling conduction aphasia.