Terri Scott¹, Peter Brunner², Gerwin Schalk³, Nancy Kanwisher4, Evelina Fedorenko5
¹Boston University, ²Wadsworth Center, ³Wadsworth Center, 4MIT, 5Massachusetts General Hospital
Language is a core human faculty that allows us to convey and understand a seemingly infinite number of abstract ideas using a limited repertoire of sounds and words. Using functional magnetic resonance imagining (fMRI), it has been shown that a set of brain regions in the frontal and temporal cortices, more robustly present in the left hemisphere, are selectively engaged during language tasks. However, the time-course of linguistic processing – critical to eventually understanding the computations performed at different processing stages – has not yet been characterized for sentence-level comprehension. We here use electrocorticography (ECoG) to probe the temporal structure of language processing. We recorded electrical activity from patients with intractable epilepsy while they read sentences and nonword lists, each containing eight items, presented one at a time at a fixed presentation rate. We took advantage of the periodic nature of the stimulus presentation to model the component of the high-gamma amplitude that was locked to the onset of each word in the sentence condition in electrodes that showed higher responses to sentences than nonwords. By modeling the Fourier (sine and cosine) basis functions of this amplitude modulation, the phase of the time-locked modulation was estimated and converted to an onset time of response in each electrode. The earliest time-locked responses occurred in mi and posterior temporal regions and later responses were found in the anterior temporal and frontal regions. Given these latency differences, we can now begin to ask what information gets extracted from the linguistic signal at different processing stages.
Keywords: Language, Electrocorticography