What Was I Thinking? Using Multi-Voxel Pattern Analysis to Identify Shared Neural Representations Between External and Internal Speech
Year: 2017
Type: dissertation
Abstract: The phenomenological experience of hearing one’s own voice inside one’s head – inner speech – is a fundamental yet mysterious aspect of cognition. Compared to speech perception and production, relatively little is known about the neural representations that underlie inner speech. This is partly due to its “hidden” nature; functional neuroimaging offers one way to observe the neural substrates of inner speech. In particular, multi-voxel pattern analysis (MVPA) is well suited to investigate its representational nature using functional magnetic resonance imaging (fMRI) data. In this dissertation, I use MVPA and fMRI to investigate the extent to which auditory and motor representations of speech are co-opted during inner speech. In Chapter 2, I establish that MVPA can be used to identify phonological features (voicing, place of articulation, and manner of articulation) during speech perception. In Chapter 3, this work is extended to include silent speech production (“speech miming”). Together, these data show that phonological features are represented throughout the bilateral auditory cortices during perception and that certain phonological features are robustly represented in the speech-motor system during miming. In Chapter 4, I use a technique known as multivariate cross-decoding to demonstrate that inner speech recapitulates neural patterns of activation observed during miming in both the auditory and motor cortices. In Chapter 5, the concept of inner speech is extended to include a verbal working memory (WM) task and MVPA is used to track the neural time-course of auditory and motor representations. The results from Chapter 4 are replicated and it is further shown that neural patterns associated with speech perception can also be reactivated during verbal WM. The neural representations shared between perception and inner speech are active early in the WM task, while the representations shared between miming and inner speech remain active throughout the WM task. This work suggests that MVPA can detect the underlying representations associated with inner speech and that multiple types of representations interact to support the task at hand. Further, it demonstrates that the patterns of activation present in the auditory and motor systems during miming provide a reliable model from which to predict inner speech.
Source: TSpace (University of Toronto)
Author: Jessica S McQuiggan
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W2778189107