The Neural Code and Dynamics of the Reading Network

Reading involves complex transformations of word forms, with visual input mapped to lexical, semantic and phonological systems in less than a second. While much has been learned about the neuroanatomy of reading from functional imaging and lesion studies, the dynamic and interactive properties of this system remain largely unknown. We will investigate the rapid computations that allow us to convert from the visual input of a string of letters to a known word with an associated sound and meaning using our established techniques for precise co- localization and analysis of a large population intracranial recordings (75 patients), thus circumventing the sparse sampling problems inherent to human intracranial experiments. In a series of experiments that systematically vary different properties of written words and modulate what kind of linguistic information participants must attend to, we will map the brain's global reading network for words. We will evaluate the neurocomputational architecture across the ventral visual stream that allows us to rapidly identify written words, and probe dynamic interactions with the broader reading network during a variety of behavioral tasks biasing lexical, phonological and semantic processes. We will then use autoregressive models to derive metastable brain states during reading and characterize dynamic network-level interactions during these stages and relate these to observable behavior. elaborate on the roles of nodes of the reading network in word learning, we will track the modulations in the distributed reading network that enable successful word learning. This will involve teaching patients new words and examining the reading network's response changes over a number of days. Critical nodes and transitions in network states derived from recordings will be validated using unifocal and multifocal direct cortical stimulation. To accomplish our goals we have set up a large multicenter collaboration. Our team has proven expertise in all aspects of language, reading, intracranial signal analysis, population level network modeling, and neural networks. This work will dramatically improve our understanding of written language systems and develop new ways to model neural computation. It will greatly enhance our understanding of dyslexia and language disorders following brain injury or degeneration, with our experimental focus on word learning directly informing neurobiological models of language.