Aliko, Sarah; (2022) A Model of the Network Architecture of the Brain that Supports Natural Language Processing. Doctoral thesis (Ph.D), UCL (University College Lodon).

Preview

Text Aliko_10153599_Thesis.pdf Download (18MB) | Preview

Abstract

For centuries, neuroscience has proposed models of the neurobiology of language processing that are static and localised to few temporal and inferior frontal regions. Although existing models have offered some insight into the processes underlying lower-level language features, they have largely overlooked how language operates in the real world. Here, we aimed at investigating the network organisation of the brain and how it supports language processing in a naturalistic setting. We hypothesised that the brain is organised in a multiple core-periphery and dynamic modular architecture, with canonical language regions forming high-connectivity hubs. Moreover, we predicted that language processing would be distributed to much of the rest of the brain, allowing it to perform more complex tasks and to share information with other cognitive domains. To test these hypotheses, we collected the Naturalistic Neuroimaging Database of people watching full length movies during functional magnetic resonance imaging. We computed network algorithms to capture the voxel-wise architecture of the brain in individual participants and inspected variations in activity distribution over different stimuli and over more complex language features. Our results confirmed the hypothesis that the brain is organised in a flexible multiple core-periphery architecture with large dynamic communities. Here, language processing was distributed to much of the rest of the brain, together forming multiple communities. Canonical language regions constituted hubs, explaining why they consistently appear in various other neurobiology of language models. Moreover, language processing was supported by other regions such as visual cortex and episodic memory regions, when processing more complex context-specific language features. Overall, our flexible and distributed model of language comprehension and the brain points to additional brain regions and pathways that could be exploited for novel and more individualised therapies for patients suffering from speech impairments.

Download activity - last month

Export as JSONCSVXML

Download activity - last 12 months

Export as XMLCSVJSON

Downloads by country - last 12 months

Export as JSONCSVXML

Archive Staff Only

View Item