Poster D21, Friday, August 17, 4:45 – 6:30 pm, Room 2000AB
Hierarchical Processing of Degraded Speech: A Functional Near-Infrared Spectroscopy Study
Bradley E. White1,2,3, Clifton Langdon1,2,3;1Language and Educational Neuroscience Laboratory, 2PhD in Educational Neuroscience Program, 3Gallaudet University
INTRODUCTION: Behavioral and neuroimaging research suggests that processing degraded acoustic information creates a cascading effect on the mechanisms underlying speech comprehension, suggesting that our cognitive resources are limited and causing a trade-off between effort and comprehension (Peelle, 2018). Here, using a plausibility judgment task and functional near-infrared spectroscopy (fNIRS), we aim to dissociate motivated listening and its modulation of language processing in response to increased demands on executive functioning. Hypotheses. We hypothesize that the processing of degraded speech is hierarchical in that hemodynamic activity in the prefrontal cortex is dependent on the amount of speech information preserved in the signal. Under this hypothesis, we predict the least amount of activation of prefrontal cortical areas for clear speech, with increases for distorted speech with a hearing aid simulation, and the greatest amount of activation for a cochlear implant simulation using 8-channel noise vocoded speech. Alternatively, we hypothesize that the processing of degraded speech is categorical in that hemodynamic activity in the prefrontal cortex may exhibit differences in clear versus degraded speech, but no significant linear relationships exists between hemodynamic activity and preserved speech features in the signal. METHODOLOGY: Participants. Monolingual, English-speaking adult participants (N=4; ages 24 to 37) with clinically-defined typical hearing, characterized by audiometric three-frequency pure-tone averages (i.e., 500 Hz, 1 kHz, and 2 kHz) ≤25 dB hearing level, and no reported history of hearing loss. Procedures. Participants complete a battery of language and cognitive assessments. The fNIRS task presents participants with 288 sentences for a plausibility judgment task. The sentences vary linguistically (i.e., simple subject-relative and complex object-relative clause structures) and acoustically (i.e., clear speech, moderately distorted hearing aid simulation, and severely distorted 8 channel noise-vocoded speech). RESULTS: Group-level fNIRS results show a positive, linear relationship between neural recruitment and increasing acoustic distortion for complex object-relative clauses (OS) compared to simple subject-relative clauses (SO). Specifically, the condition of greatest distortion (8 channel noise-vocoded speech) shows the greatest amount of modulation by linguistic complexity (OS>SO) with greater involvement of left temporal and frontal regions. The lesser-distorted condition (hearing aid simulation) reveals right temporofrontal sites modulated by OS>SO contrast. In the clear, undistorted condition, OS>SO contrast does not reveal significant modulation of neural activation. These results indicate support for the hypothesis that cognitive load to maintain traces in verbal working memory varies as a function of the speech information in the signal. CONCLUSION: Differentiating between these two hypotheses advances our understanding of how auditory degradations modulate language and associated processes (e.g., attention, short-term verbal working memory), and specifically the relationship of this modulation to the amount of speech features preserved in the signal. This work has potential to impact further research on auditory and spoken language processing in hearing aid and cochlear implant users.
Topic Area: Perception: Auditory