MULTIPLE PATHWAYS TO SUPRATHRESHOLD SPEECH IN NOISE DEFICIT IN HUMAN LISTENERS

Threshold audiometry, which measures the audibility of sounds in quiet, is currently the foundation of clinical hearing evaluation and patient management. Yet, despite using clinically prescribed state-of-the-art hearing aids that can restore audibility in quiet, patients with sensorineural hearing loss (SNHL) experience difficulty understanding speech in noisy backgrounds (e.g. cocktail party-like situations). This is likely because the amplification provided by modern hearing aids while restoring audibility in quiet, cannot compensate for the degradation in neural coding of speech in noise resulting from a range of non-linear changes in cochlear function that occur due to hearing damage. Furthermore, in addition to robust neural coding, the efficacy of cognitive processes such as selective attention also influences speech understanding outcomes. While much is known about how audibility affects speech understanding outcomes, little is known about suprathreshold deficits in SNHL. Unfortunately, direct measurements of the physiological changes in human inner ears are not possible due to ethical constraints. Here, I use noninvasive tools to characterize the effects of two less-familiar forms of SNHL: cochlear synaptopathy (CS; chapter 2) and distorted tonotopy (DT; chapters 3 and 4). Results from our experiments in Chapter 2 showed that age-related CS degrades envelope coding even in the absence of audiometric hearing loss and that these effects can be quantified using non-invasive electroencephalography (EEG)-based envelope-following response (EFRs) metrics. To date, DT has been only studied in laboratory-controlled animal models. In chapters 3 and 4, I combined psychophysical tuning curves, EFRs, and speech-in-noise measurements to characterize the effects of DT. Our results suggest that low-frequency noise produces a strong masking effect on the coding of speech in individuals with SNHL and that an index of DT (tip-to-tail ratio) obtained from psychophysical tuning curves can account for a significant portion of the large individual variability in listening outcomes among hearing-aid users, over and beyond audibility. Lastly, I propose a machine-learning framework to study the effect of attentional control on speech-in-noise outcomes (chapter 5). Specifically, I introduced a machine-learning model to assess how attentional control influences speech-in-noise understanding, using EEG to predict listening performance outcomes based on prestimulus neural activity. This design allows for examining the influence of top-down executive function on listening outcomes separately from the peripheral effects of SNHL. The results from our study suggest prestimulus EEG can predict subsequent listening outcomes and the changes in alpha rhythm may be used as a neural correlate of attention.