| dc.description.abstract | The human ability to understand speech is underpinned by a hierarchical auditory system whose successive stages process increasingly complex attributes of the auditory input. To produce categorical speech perception, it has been suggested that this system must elicit consistent neural responses to speech tokens (e.g., syllables, phonemes) despite variations in their acoustics. This is an intermediate stage of the speech processing hierarchy, followed by lexical and semantical analyses which allow the extraction of concepts from speech sounds.
Although speech is a very important and unique aspect of humans, the cortical mechanisms that allow its efficient processing remain unclear. One of the issues is methodological, with research in auditory neuroscience often constrained to utilising artificial speech stimuli such as isolated syllables or words. Also, important aspects such as the temporal cortical dynamics have been ignored by studies that use technologies such as functional MRI. This thesis investigates the neural underpinnings of speech perception using scalp electro- and magneto-encephalography (EEG, MEG), with the aim of developing methodologies capable of isolating neural activity at different stages of the speech processing hierarchy by analysing spatio-temporal cortical dynamics in response to different features of speech.
Previous research revealed that cortical activity is entrained to the low frequency amplitude fluctuations of speech in time (i.e., amplitude envelope). However, the neural underpinnings of this phenomenon remain unclear. Here, further insights on this topic are revealed by studying the responses to continuous natural speech. In the first study (Chapter 3), participants were presented with natural speech from an audio-book while non-invasive EEG signals were recorded. This chapter demonstrates that EEG signals are sensitive to the temporal tracking of categorical phonological features of speech and provides a novel analysis framework to quantitatively investigate this phenomenon.
The study in Chapter 4 aimed to further assess this framework and, specifically, its ability to isolate cortical responses to phonetic features from the ones in response to speech acoustics. This involved implementing a perceptual pop-out paradigm that, by providing or not providing prior predictive knowledge on the upcoming stimuli, allowed for the comparison between two conditions consisting of the same stimulus but different perceived clarity. As a result, this study introduced a novel cortical measure of phoneme-level speech processing that is modulated by the perceived clarity of the incoming stimuli. The effects of prior predictive knowledge on the cortical responses to speech are further investigated in Chapter 5, in which a similar pop-out paradigm and source-space MEG signals were used to provide new insights on the neural substrates of these effects and, specifically, on the interactions between and within selected cortical sites in temporal and frontal areas.
The other important question of this thesis regards the applicability of the novel framework in Chapter 3 in the study of language development and specific speech and language impairments. This kind of research can involve working with particular cohorts (e.g., clinical populations, infants), which may be unable to undertake long EEG experiments, and one issue relating to the framework described in the previous chapters is whether it would work with small amounts of data. Chapter 6 aims to address this by introducing an extension of the original approach that allows the reduction of the recording time down to 10 minutes. Finally, this improved approach is applied to investigate speech perception in children with dyslexia and the results support theories stating that this developmental deficit, whose symptoms involve mainly a person?s reading skills, is related with language processing and, in particular, with the low frequency cortical tracking of phonological features. | en |
