Speech understanding is a key function of human audition with important roles in learning, professional, and social functions. Ageing is associated with increased difficulties with speech recognition, particularly in noisy backgrounds, even in the absence of clinical hearing loss. A growing body of work has identified increased compensatory use of cognitive resources, collectively known as listening effort (LE), as a key change within the ageing and hearing-impaired populations. Additionally, individuals with mild cognitive impairment (MCI) and Alzheimer’s Disease (AD), whose cognitive deficits compromise their ability to efficiently utilize LE mechanisms for coping with distracting information, exhibit substantial deficits in perception of speech in the presence of competing talkers. However, the neural underpinnings of these behavioral deficits in MCI remain poorly understood, including the relative extent to which neurobiological changes in the brain due to AD affect lower- level processing of speech acoustics versus higher-level mechanisms involved in linguistic processing. Moreover, because cortical processing of continuous speech in people with MCI due to AD has been virtually unexplored, it is unknown whether such measures could have utility as a neural biomarker for early, affordable diagnostics of AD, as well as for tracking of AD progression. The present proposal aims to address both of these key goals by collecting the first data set of continuous speech processing under varying listening effort demands in people with MCI due to AD and matched controls. In Aim 1, we will address the question of how neurobiological changes due to AD influence cortical processing of acoustic and linguistic features of speech in MCI. In a sample of participants with MCI due to AD with AD biomarkers, as well as matched controls, we will measure non- invasive electroencephalographic (EEG) responses to continuous narrative audiobooks presented either in isolation (Low LE condition) or dichotically in the presence of a distracting secondary audiobook (High LE condition). Model-based temporal response function (TRF) analyses will subsequently be used to derive responses to a range of acoustic and linguistic features of speech to assess how MCI status and LE demands interact in cortical processing of speech. In Aim 2, we will test whether cortical responses to speech can be reliably used to, 1), classify the MCI (vs. control) status of individual participants, and 2), predict speech comprehension performance and scores on cognitive battery tests (e.g., working memory capacity, attentional inhibition scores). These goals will be achieved by building cross-validated classification and regression models to identify which features (acoustics vs. linguistic), conditions (clean vs dichotic speech), and types of neural measures (e.g., feature tracking strength vs. response latencies) are most reliable for classifying MCI status and predicting behavioral scores. This w...