Quantifying the Influence of Lexical Surprisal on Acoustic Speech Encoding While Controlling for Within‐Speaker Variability
European Journal of Neuroscience · 2026-07 · 64 (1)
DOI: 10.1111/ejn.70569 ↗Kurz gefasst
Fachartikel in der Zeitschrift European Journal of Neuroscience, online erschienen am 02.07.2026. Verfasst von Shyanthony R Synigal und 2 weiteren Autor:innen. Beteiligte Einrichtung: Department of Neuroscience, University of Rochester, Rochester, New York, USA. Hauptthemen laut MEDLINE: Speech Perception, Speech Acoustics, Brain. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 76 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: Del Monte Institute for Neuroscience, Irish Research Council, Simons Foundation Autism Research Initiative.
Abstract (Originalsprache)
There is substantial support for the idea that the listening brain makes predictions about upcoming speech and that these predictions are integrated with sensory input to influence perception. For example, the early auditory encoding of words appears to vary based on how those words semantically relate to their preceding context, suggesting that top‐down information might feed back to affect acoustic speech processing. However, the way in which speakers enunciate words can vary based on how well those words fit with their preceding context. This presents a potential confound to the interpretation of top‐down prediction in the listener. In this study, we address this possibility by assessing the influence of probability‐based predictions (word surprisal) on electroencephalographic (EEG) indices of acoustic speech processing while controlling for variations in speaker dynamics. We analyzed EEG from 14 adults who undertook a perceptual pop‐out task in which prior information enhanced the comprehensibility of degraded speech while acoustic information was held constant. Behavioral results confirmed the manipulation's effectiveness and were mirrored in the neural indices of word surprisal processing. Importantly, a positive relationship between word surprisal and EEG tracking of word acoustics emerged for degraded speech when prior information rendered it intelligible, but was absent when it was unintelligible, despite identical acoustic input across conditions. The difference in neural effects between conditions also correlated with the corresponding difference in behavioral pop‐out. These findings support the claim that top‐down word predictability influences the acoustic encoding of natural speech, independent of variations in speaker enunciation.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Department of Neuroscience, University of Rochester, Rochester, New York, USA.
- Del Monte Institute for Neuroscience, University of Rochester, Rochester, New York, USA.
- School of Engineering, Trinity Centre for Bioengineering and Trinity College Institute of Neuroscience, Trinity College Dublin, Dublin, Ireland.
- Department of Biomedical Engineering, University of Rochester, Rochester, New York, USA.
- Center for Visual Science, University of Rochester, Rochester, New York, USA.
Förderung
- Del Monte Institute for Neuroscience
- Irish Research Council
- Simons Foundation Autism Research Initiative
Zitieren
APA
Synigal, S. R., Broderick, M. P., & Lalor, E. C. (2026). Quantifying the Influence of Lexical Surprisal on Acoustic Speech Encoding While Controlling for Within‐Speaker Variability. European Journal of Neuroscience, 64(1). https://doi.org/10.1111/ejn.70569
Vancouver
Synigal SR, Broderick MP, Lalor EC. Quantifying the Influence of Lexical Surprisal on Acoustic Speech Encoding While Controlling for Within‐Speaker Variability. European Journal of Neuroscience. 2026;64(1). doi:10.1111/ejn.70569
BibTeX
@article{Synigal2026,
author = {Synigal, Shyanthony R and Broderick, Michael P and Lalor, Edmund C},
title = {Quantifying the Influence of Lexical Surprisal on Acoustic Speech Encoding While Controlling for Within‐Speaker Variability},
journal = {European Journal of Neuroscience},
year = {2026},
volume = {64},
number = {1},
doi = {10.1111/ejn.70569}
}
EUVIMED