Investigating the neurophysiology of auditory attention in multi-talker scenarios: from sustained attention to attention switching

Abstract

Despite the complexity of our auditory world and the limitations of listeners' cognitive and perceptual resources, the human ability to use and comprehend speech is remarkably effective. Its robustness makes it possible to sustain attention on a person of interest among competing speakers, to redirect the focus of attention at will, and to rapidly reorient attention on unexpected auditory events. As such, speech processing abilities are tightly intertwined with the ability to maintain and shift attention. The neural cortical mechanisms underpinning attention to speech have been explored with invasive and non-invasive technologies, establishing the existence of a selective enhanced representation of the speaker of interest (target) in cortical signals, along with a weaker encoding of interfering speakers outside the focus of attention (maskers). However, several questions remain unanswered, especially regarding the limits of the attentional filter on different levels of speech information (from acoustics to abstract meaning), and the ability to flexibly reorient the focus of attention on multiple speakers alternatively. Furthermore, there is even less clarity on how those cortical mechanisms may operate across ages and hearing abilities. In this thesis, I investigate the neural basis of attention to speech across different populations and attention scenarios. Electroencephalography (EEG) recordings of the listeners' brain signals were related to various speech features through methods based on lagged linear regression, yielding measures of neural tracking of competing speakers in multi-talker scenes characterised by different attentional demands. Firstly, this work offers novel insights into the neural encoding of competing talkers in listeners with hearing impairment, focussing on the level of phonological information. The study presented in Chapter 3 indicates that neural tracking of speech phonological information is increased for the target relative to the masker speaker, although phoneme onsets of the masker are more neurally represented than those of the target. This mechanism may represent a neural basis for the increased distractibility in multi-talker scenes in listeners with hearing impairment. The studies presented in Chapter 4 and Chapter 5, on the other hand, investigate the neural underpinnings of flexible attention reorienting between two competing speakers, with a multi-talker scenario requiring the listener to perform frequent switches of attention. Through this dynamic attention paradigm, the experiment in Chapter 4 offers a novel framework to interpret the attention switch in terms of the disengagement and engagement of attention between competing speakers, exploring their temporal dynamics and establishing a transient simultaneous tracking of the two streams while a switch of attention unfolds. That study also finds a robust reduction in brain activity within the alpha frequency band (8-12 Hz) after the switching cue, and that the trough of the alpha decrement follows the neural attention switch. Furthermore, we investigate the impact of previous context on the neural tracking of the current speech block, comparing four alternative models on their ability to explain the EEG data, and revealing that a model that resets the context at every switching cue has the highest fit to the neural signals. Finally, in Chapter 5, we extend our attention switching paradigm by including frequent changes of topic within the same stream after a switch of attention, challenging the reliance on previous context, and we test in on two cohorts of younger and older listeners. The results suggest that neural tracking of speech is highly impacted by attention switching cues, decreasing over the course of each trial, in particular for older listeners. Neural responses to the speech acoustics are also impacted by age, with differences in the latency and amplitude of N1 and P2 components. Furthermore, the attention switching cue introduces a P300-like detection response, which also displays delayed temporal dynamics in older listeners. Similarly to the study in Chapter 4, also here an alpha activity decrement follows the attention switching cue, with similar dynamics for younger and older listeners.