Thesis Format
Monograph
Degree
Master of Science
Program
Neuroscience
Supervisor
Batterink, Laura J.
2nd Supervisor
Köhler, Stefan
Co-Supervisor
Abstract
Statistical learning (SL) is a fundamental learning mechanism, but its neural underpinnings are not well understood. Modality-specific sensory regions are known to be involved in SL, but some studies suggest that the hippocampus may also support this process. However, direct neural evidence for engagement of the hippocampus in speech-based SL is scarce. In our study, 13 epilepsy patients with intracranial electrodes listened to a continuous speech stream containing embedded trisyllabic words. Neural entrainment at the frequency of words was used as a neural marker of learning, along with implicit and explicit behavioral measures. We found neural evidence of SL mainly in temporal cortex, insula, parietal, and frontal regions, but not the hippocampus. Behaviorally, we found robust evidence of SL on our implicit measure. Our results suggest that SL engages a widespread brain network that may not involve the hippocampus, but further research is needed to elucidate its role in SL.
Summary for Lay Audience
The external world contains a continuous flow of sensory input that we must learn to navigate and understand. As complex as this input may be, it contains structure in the form of regularities that repeat over time. Statistical learning (SL) is a fundamental learning mechanism that supports the extraction of such regularities from the environment. Among many other aspects of perception and cognition, SL is thought to support a crucial step in the process of acquiring a new language: the ability to discover words from continuous speech. One of the regularities in any given language is that syllables within words occur together more often than the syllables that span word boundaries. SL may support the discovery of words by extracting the underlying statistical regularities among the syllables in a language, even in the absence of cues (e.g. pauses or tone changes) that reliably indicate where one word ends and another begins. Despite its importance, the neural underpinnings of SL are not fully understood. Electroencephalography (EEG) studies have capitalized on a phenomenon known as neural entrainment – the synchronization of neural activity with external rhythmic stimuli- to study SL, and they have found that the brain rapidly tracks the underlying statistical regularities in speech. On the other hand, several functional magnetic resonance imaging (fMRI) studies have shown the engagement of sensory brain regions in SL, but some studies suggest that the hippocampus may support this process as well. To further investigate this, we tested a group of epilepsy patients that had undergone the implantation of intracranial depth electrodes as part of their medical treatment. Patients listened to a continuous speech stream that contained four “hidden” trisyllabic words, where neural entrainment at the frequency of the hidden words provided evidence of neural tracking of the underlying statistical regularities in the speech stream. We found evidence of SL primarily in temporal regions and insula, in addition to parietal and frontal cortices, but not in the hippocampus. Our results suggest that SL involves a widespread network of regions, but more research is necessary to determine whether the hippocampus plays a role in this process.
Recommended Citation
Herrera Chaves, Daniela, "Statistical Learning in the Hippocampus and Neocortical Regions: Evidence From Intracranial Neural Entrainment" (2024). Electronic Thesis and Dissertation Repository. 10425.
https://ir.lib.uwo.ca/etd/10425