Abstracts

Rationale: Electrocortical stimulation mapping is frequently used to identify language cortex prior to resective surgery involving the language dominant hemisphere. Object naming is the most frequently utilized task to identify language cortex. Typically, once a cortical site is identified as positive for naming, no further steps are taken to determine the underlying cause of naming failure. Naming is a complex process, involving multiple mechanisms that culminate in the identification of the word that best matches a particular meaning (i.e., semantics), and the retrieval and realization of word sounds (i.e., phonology). Thus, naming could be disrupted due to impairment to either of these linguistic subprocesses, which are likely mediated by different brain regions. We reasoned that the type of error produced during stimulation might provide insight regarding the linguistic nature of the naming site, and hypothesized an anatomical dissociation between sites at which stimulation would produce semantic versus phonological errors. Methods: Subjects were 70 epilepsy patients with seizures arising from the left (dominant) perisylvian region (mean age = 34.0, (SD= 11.5), mean age of epilepsy onset 19.8 (SD = 13.4), mean education = 13.8 (SD= 2.8), mean Full Scale IQ =94.8 (SD=14.4)) who underwent extra- (n =48) or intra-operative (n = 22) language mapping at Columbia University Medical Center. Naming sites were identified using visual object naming or auditory description naming tasks. Semantic errors were defined as words related in meaning to the target word, e.g., "crayon" for "pencil." Phonological errors were defined as word or nonwords containing phonological substitutions for the target word, e.g., "fen" for "pen." Electrode locations were documented using digital photography and schematic diagrams. For extra-operative mapping, in addition to skull X-rays, electrode location was confirmed with post-operative MRI and/or CT scans. Sites were plotted on a group composite schematic. Chi square analysis assessed the topographical distribution of sites associated with semantic versus phonological errors. Results: Of 159 semantic errors, 58 occurred with stimulation on the superior temporal gyrus (STG), whereas, 101 occurred with stimulation along the middle temporal gyrus (MTG). In contrast, of 29 phonological errors, 21 occurred with STG stimulation, whereas only 8 occurred with MTG stimulation, resulting in a significant double dissociation in the topography of semantic and phonological errors (x² = 13.0, P=0.003). Conclusions: Our results suggest that stimulation along the STG primarily disrupts phonological processing, whereas MTG stimulation primarily disrupts semantic processing. This pattern is consistent with a dual stream model of speech processing, and suggests that naming sites identified by cortical stimulation might not be functionally equivalent. This potentially raises questions regarding resectability of specific naming sites, and may carry implications for remediation of postoperative naming deficits.