Fast cortical oscillations as a potential network mechanism for language processing

Since the 19th century, the study of brain mechanisms of language has depended on available tools: the clinical study of language-impaired patients, with neuropathological correlates, in the style of Broca and Wernicke; imaging techniques including MRI, functional MRI, and MRI tractography; and direct stimulation of, and recording from, the brains of conscious patients performing language tasks. These tasks can be directed or spontaneous and occur typically in the course of evaluations prior to surgery or intraoperatively. The study of brain and language occurs in the context of classical linguistics, with its (relative) distinctions between semantics and syntax, and its requirements for formal analysis in the latter. A consequence has been the effort to parcellate regions of the left hemisphere (of most individuals) in terms of distinct linguistic functions, and to characterize the anatomical connections between parcels: the various fascicles. In parallel, invasive brain recordings of activity at the level of networks or multiple single cells have allowed correlation of localized electrical signals with linguistic parameters. Recently, however, a paradigm shift has begun concerning the proper framework for interpreting language-related brain measurements. Partly this has occurred because of the success of large language models (LLMs), which do not include explicit dependence on formal syntactic/semantic distinctions. As a result, electrical brain measurements are now examined with a focus on interactions between multiple small cortical "modules." In this paper, we examine the cellular physiology underlying the activities of modules and their interactions, with emphasis on the mechanisms and functions of fast brain oscillations.