Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Altered hierarchical speech tracking in schizophrenia during naturalistic listening
Poster Session E, Friday, October 2, 11:00 am - 1:00 pm, Wangari Maathai
Shyanthony Synigal1, Wen Li1, Michael Elacqua1, Madison Dougherty2, Haley Dennis1, Tsion Eshetu1, Steven Silverstein1, Judy Thompson1, Edmund Lalor1,2; 1University of Rochester Medical Center, 2University of Rochester
Speech perception relies on the integration of information across multiple representational levels, from acoustic cues to higher-level linguistic representations. While impairments in speech and language processing are well documented in schizophrenia (SZ), it remains unclear how these differences manifest across hierarchical levels during naturalistic listening. In particular, it is unknown how much different levels contribute to neural responses, and whether effects observed at higher levels reflect independent changes or are related to earlier stages of processing. In the present study, we used electroencephalography (EEG) and temporal response function (TRF) modeling to investigate how individuals with SZ and matched healthy controls encode continuous, natural speech. Participants listened to narrative speech while EEG was recorded and answered comprehension questions to ensure engagement. Neural responses were modeled using predictors that capture multiple levels of speech representation, including low-level acoustic features, phonetic features, and lexical surprisal. This analysis allowed us to quantify the contribution of each feature space to neural responses and account for shared variance across levels, providing a framework to dissociate hierarchical speech encoding. Preliminary results indicate that individuals with SZ exhibit altered speech tracking across representational levels. Specifically, we observe group differences in acoustic encoding, alongside alterations in responses associated with lexical surprisal. These findings suggest that speech processing differences in SZ are not confined to a single level of representation but reflect changes across multiple levels in the processing hierarchy. Ongoing analyses are examining whether these neural measures relate to symptom profiles. Together, these results demonstrate the utility of TRF modeling for dissociating hierarchical speech representations in clinical populations. By characterizing how acoustic, phonetic, and lexical information are encoded during naturalistic listening, this work provides insight into how speech processing is altered across representational levels in schizophrenia.
Topic Areas: Speech Perception,