Effective cortico-cortical connectivity in auditory and speech processing

The thesis aimed to update the traditional understanding of the speech chain with recent proposals on communicative behaviour from theoretical and computational neuroscience. For a generative brain that engages in active (Bayesian) inference, speech perception itself is considered as a form of predictive processing. Experiments presented in this thesis were designed to answer the questions of ‘what’, ‘how’, and ‘where’: 1. What constitutes an auditory prediction error, and is it selective to a specific sound type? 2. How is surprise minimisation implemented in hierarchical cortical networks for auditory and speech perception? 3. Where – at what level in cognition and linguistic knowledge – could predictions for speech perception come from? Study 1 answered the first question – of ‘what’ – by collecting passive Mismatch Negativity responses to speech and non-speech sounds. This was recorded as subjects ignored the auditory stimuli. It was hypothesised that if speech-specific auditory prediction errors existed, certain aspects of MMN – a change detection response – would be selective to sound type and to speech category. Source-analyses of the same EEG data in Study 1 provided clues to the question of ‘how’. To address ‘where’, Study 2 considered attentional modulations to the effect of categorical processing recorded in passive brain responses, and tested for a prelexical locus of speech-specific prediction and prior knowledge. Auditory-evoked brain responses were recorded in both Ignore and Attend conditions to acoustically identical stimuli from three subject groups of varying levels of familiarity to a certain speech category. Stimuli manipulation in an accompanying behavioural discrimination task allowed for the monitoring of top-down lexical influences. Findings from this project will provide a way to reframe existing theoretical debates in speech processing on topics such as speech-specific auditory processing (against non-speech sounds), categorical perception, and pre-lexical abstraction within spoken word recognition