Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Lexical access in degraded speech: ERP evidence from noise-vocoded audiovisual words
Poster Session B, Wednesday, September 30, 4:30 - 6:30 pm, Wangari Maathai
This poster is part of the Sandbox Series.
Summer Bolla1, Sharon Coffey-Corina1, Elizabeth Pierotti1, David P. Corina1; 1UC Davis Center for Mind and Brain
Introduction: Deaf individuals who use cochlear implants to hear face distinct challenges when processing spoken language. Cochlear implants (CIs) provide access to sound through direct electrical stimulation of the auditory nerve, but the resulting signal substantially reduces spectral detail compared to typical acoustic hearing. One widely used method for studying the auditory experience of CI users is noise-vocoded speech, which replaces spectral information in the speech signal with modulated noise while preserving temporal cues. Noise-vocoding is a well-validated method for simulating the perceptual degradation characteristic of CIs. Beyond acoustic limitations, many CI users also experience reduced or delayed language exposure prior to implantation, potentially contributing to developmental differences in linguistic processing. A recent ERP study by Pierotti (2024) examined audiovisual word processing in cochlear implanted and typically hearing children across semantic and phonological conditions. While N400 responses to semantic information were similar across groups, CI users showed distinct ERP patterns in the 200–350 ms window for phonological conditions, suggesting differences in how phonological information is accessed or integrated. The present study investigates whether such differences arise primarily from degraded perceptual input or from broader developmental factors. By presenting noise-vocoded and clear speech to typically hearing adults, this project isolates the perceptual component to better understand how reduced spectral detail influences lexical and phonological access. Methods: ERP data was gathered from typically hearing college students (n = 28, data collection in progress) presented with the same audiovisual paradigm as the previous experiment in either 16 channel noise-vocoded or clear speech. Following a picture prime, the participants saw a video of a speaker saying a word across different conditions: a match (e.g. CAKE – CAKE), unrelated (e.g. CAKE – SEAL), or rhyme (e.g. CAKE – RAKE). Participants made a match/mismatch judgement via button press. EEG data was processed at the midline sites Pz, Fz and Cz, and mean amplitudes were extracted for the 200-350 and 400-600ms time windows. Results: Preliminary ERP results are based on data from 16 participants (8 vocode, 8 clear speech). Visual inspections of grand-averaged waveforms revealed broadly similar ERP morphologies across groups in both phonological and semantic time windows. Linear mixed effects models were fit for both the 200-350ms and 400-600 ms windows with mean amplitude as the dependent variable and fixed effects of group x condition averaged across midline sites. In the 200-350ms window a main effect of group (p = 0.046) emerged, with the vocode group showing a larger negative mean amplitude. In the 400-600ms window, a main effect of condition was observed, such that rhyme (p = 0.033) and unrelated (p = 0.022) trials elicited higher negative mean amplitudes than match trials across groups. Conclusion: These preliminary findings suggest that noise-vocoded speech modulates the processing of phonological information without substantially affecting semantic processing in typically hearing adults. The observed pattern partially aligns with the previous work on CI using children, suggesting that the degraded perceptual quality of the implants may account for some of the neurological patterns seen in phonological processing of audiovisually presented speech.
Topic Areas: Disorders: Developmental, Speech Perception