Event

Committee:

Professor Jonathan Z. Simon (Chair)

Professor Betash Babadi

Professor Shihab Shamma

Date/time: Tuesday, May 5th at 11:00 AM

Location: AVW 1146

Title: Understanding Neural Processes of Auditory Stream Segregation: Impacts from Auditory-Cognitive Training and Bilingualism

Abstract: 

Understanding speech in noisy environments is a common challenge among older listeners and multilinguals. Behavioral measures show that these individuals perform poorly on speech-in-noise tasks compared to younger or monolingual adults, and clinical assessments often use similar tasks to evaluate hearing-in-noise performance. Yet, age and multilingualism may affect different aspects of the auditory processing hierarchy. Therefore, assessments using behavioral outcomes alone is not enough to evaluate and/or assess listening difficulties. Analyzing neural processes, with respect to auditory features, may provide better insight into the behavioral measures observed; though, this analyses remains largely unexplored in older and non-native English listeners. In this thesis proposal, we aim to understand the time-locked neural processes of auditory stream segregation among older and non-native English listeners and how these processes may improve using auditory-cognitive training. Firstly, we analyze neural differences in auditory stream segregation among native and non-native English speakers. Because traditional speech-in-noise tasks are recall-based tasks, additional language knowledge may enhance either informational masking or language proficiency may impact recollection. In this project, we use a non-linguistic auditory perception task, stochastic figure-ground (SFG). The SFG task targets the same neural processes necessary for auditory stream segregation but reduces the confounds of language proficiency and working memory. The detection-based task uses simple tones manipulated such that the tones can be divided into two different auditory streams, a temporally coherent stream and random background noise. Using magnetoencephalography (MEG) and electroencephalography (EEG), we analyze the neural differences of the temporally coherent stream between monolinguals and bilinguals. We obtain a generalized, time-locked neural response using the temporal response function (TRF) and compare differences using generalized additive mixed models. Preliminary results indicate that differences may emerge around the 4th to 6th temporally coherent tones but resolve afterwards. Secondly, we introduce a method of analyzing training performance in staircase adaptive procedures. In clinical assessments or research experiments, staircase adaptive procedures are used to obtain speech reception thresholds, a volume threshold at which a person can accurately recall words a certain percent of the time (e.g., recall words correctly 70% of the time). However, with cognitively-intensive tasks, the progression of estimating this threshold may provide more valuable insight than the threshold value alone. We use segmented regression and k-means clustering to characterize temporal patterns of the stair-case procedure. In this analysis, we show that characterizing temporal patterns of auditory-cognitive training performance may help identify events of learning or fatigue. This result highlights the importance of keeping temporal dynamics when assessing training performance. Lastly, we analyze the neural and behavioral differences among younger and older adults, and how these differences in older adults change after undergoing auditory-cognitive training. Older adults often report higher listening effort while listening in noisy conditions, despite having clinically normal hearing thresholds. Auditory-cognitive training has shown promising behavioral evidence for speech-in-noise listening. However, how auditory-cognitive training impacts cortical neural processes is not well understood. In this project, we use MEG to obtain high temporal resolution of time-locked neural responses, non-invasively. To analyze the time-locked responses, we use encoding and decoding models using the TRF framework to obtain a generalizable response relative to the stimulus feature. Results show that younger adults have an attentional contrast of cortical representation during competing-speaker listening. However, older adults do not. We show that auditory-cognitive training improves attentional contrast in older adults; moreover, the degree of attentional contrast improvement may be predicted by pre-training SFG performance. These projects show how different neural and behavioral processes may still result in the same behavioral outcome and emphasize the importance of interpreting un-averaged data. Together, this work explores the neural mechanisms among listeners who have difficulties in speech-in-noise understanding and provides meaningful insight to further our knowledge of auditory stream segregation.