The role of expected informativity on linguistic and memory encoding

About This Grant

Project Summary/Abstract Prosody is a salient feature of speech that helps listeners anticipate when the most informative parts of a message will occur, facilitating comprehension. Prosodic impairments are linked to language deficits across disorders, yet prosody is largely absent from neural models of language processing. This proposal addresses this gap by investigating how neural tracking of prosody allows listeners to anticipate the expected informativity of upcoming words, engaging cortico-subcortical mechanisms to support linguistic processing and memory encoding. We overcome limitations of prior behavioral and scalp EEG work based on artificial paradigms, by leveraging intracranial EEG (iEEG) and machine learning (ML) to measure the contribution of subcortical structures to naturalistic comprehension. The mentored phase will involve intensive analyses of iEEG data from neurosurgical patients listening to naturalistic stories. Aim 1 tests whether prosodic tracking enhances the encoding of phonetic, semantic and syntactic features throughout the temporal lobe. Aim 2 investigates whether prosodic tracking supports memory encoding by modulating hippocampal theta (2.5–5 Hz). During the independent phase, I will combine naturalistic and factorial designs to orthogonalize prosodic cues and actual informativity (semantic surprisal). Aim 3 will dissociate how the brain tracks prosody to prepare for greater information load, from how it predicts and responds to the content of that information. The proposed work will advance neurolinguistic models by identifying the understudied key role of prosody as a cue to expected informativity. It will also contribute to ongoing debates on whether the hippocampus supports language processing, and how this interacts with its established role in memory encoding. I will achieve my long-term goal of launching an independent research career in speech neuroscience through critical new training during the mentored phase, addressing two key gaps. First, I need to move beyond controlled psycholinguistic paradigms by mastering the use of ML methods to study the processing of naturalistic speech. Second, I need to develop a deeper understanding of systems neuroscience, acquiring new theoretical background and methodological skills for the analysis of high-resolution imaging methods such as iEEG. This project integrates ML tools and iEEG with my prior background in psycholinguistics and scalp EEG, advancing a highly innovative approach that takes advantage of the benefits of multiple methods. I will receive exceptional training by working with Dr. Gwilliams, a pioneer in the application of ML for speech neuroscience, and Dr. Parvizi, a leader in use of iEEG. Stanford is an ideal training environment, as a hub for collaborative research between neuroscientists and clinician-scientists across the country. Joining this collaborative network will ensure I will continue to have access to iEEG or other high-resolution methods at the R00 institution. In sum, this proposal offers a novel neurocognitive framework for how prosody guides processing of the most informative parts of speech, laying the groundwork for understanding how disruptions in this system contribute to communication disorders.