Individual variability in the localization of language, as measured by object-naming, was assessed for left lateral peri-Sylvian cortex with a multi-sample technique of stimulation mapping at a constant current. This study was performed during craniotomy under local anesthesia in 10 patients with medically intractable epilepsy and the usual pattern of left brain dominance for language. A high degree of variability in the exact location of naming was present. Only one area, a small portion of the third frontal convolution immediately in front of the motor strip, showed naming changes in all patients in whom it was sampled. This area is considerably smaller than the classical Broca's area. Elsewhere in language cortex, including all parts of the posterior language area, there was considerable individual variability. Because of this, the classical model of language localization is an inaccurate basis for establishing the risk of aphasia in surgical therapy of dominant hemisphere peri-Sylvian lesions in this and related patient populations. Rather, that risk should be assessed from the individual localization of language, established by the multisample technique of stimulation mapping.