Schedule of Events | Search Abstracts | Invited Symposia | Symposia | Poster Sessions | Data Blitz

Neural Encoding of Acoustic and Linguistic Speech Features Reveals Correlates of Understanding

Poster Session D - Monday, March 9, 2026, 8:00 – 10:00 am PDT, Fairview/Kitsilano Ballroom

Alexis Deighton MacIntyre¹ (), Tobias Goehring^1,2, Matt Davis¹; ¹MRC Cognition and Brain Sciences Unit, University of Cambridge, ²Department of Otorhinolaryngology, Head and Neck Surgery, University of Zurich and University Hospital Zurich

Listeners rely on multiple acoustic and linguistic cues to extract meaning from speech, yet the neural basis of this process under degraded listening conditions remains unclear. Using continuous EEG, we recorded neural activity from 38 young adults as they listened to a naturalistic story. Speech cues were manipulated along two axes: acoustic spectral clarity and linguistic intelligibility, with both intelligible and non-intelligible languages produced by a bilingual speaker. We modelled brain responses to spectro-temporal descriptors, such as spectral flux, and linguistic features, such as word onsets, using banded regression and permutation analyses to control for correlated predictors. Four low-level acoustic properties—the amplitude envelope, peaks in the envelope derivative, spectral variation, and spectral flux—accounted for distinct neural variance. When linguistic features were fitted to this model, robust word onset encoding emerged exclusively for intelligible speech, with more mixed results for phoneme, syllable, and lexical surprisal features. Group analyses revealed condition-specific effects reflecting interactions between spectral clarity and language intelligibility. A composite neural index summarising these effects strongly predicted self-reported comprehension of acoustically degraded but intelligible speech (ρ = 0.90, p < .001). The results show that cortical encoding of speech relies on multiple cues whose relative contributions depend on signal quality and speech understanding. Importantly, flexibility in speech encoding across contexts may provide a more informative neural correlate than the absolute magnitude of the response. These data link brain dynamics to subjective understanding during natural speech perception, shedding light on cognitive-sensory interactions and indicating potential clinical applications.

Topic Area: PERCEPTION & ACTION: Audition