pipette
ENEnglish

Neural processing of linguistic speech features and their modulation by selective attention in users of hearing aids and cochlear implants

C. Jehn, A. Schueller, N. K. Vavatzanidis, A. Hahne, T. Reichenbach

Preprint

En palabras de los autores

Understanding speech relies on the neural processing of acoustic and linguistic features, which can both be analyzed using Temporal Response Functions (TRFs). While both types of features have been the focus of numerous studies in typical-hearing individuals, much less is known about the neural processing of continuous speech for people with hearing loss, particularly how hearing-impaired listeners process word-level features in noisy environments, where they typically struggle. We recorded EEG from hearing aid (HA, n=29) and cochlear implant (CI, n=24) users, and an age-matched typical-hearing (TH) control group (n=29), while participants listened to audiobook segments in quiet and with a competing speaker, using their personal hearing devices (CI and HA). A hierarchical encoding framework incrementally added word onset, word frequency, and word surprisal as predictors of the EEG signal. Attentional modulation was assessed by comparing encoding scores for attended versus ignored speech. All three groups showed clear neural responses to word onset, frequency, and surprisal for speech in quiet. Surprisal tracking emerged at 400~ms in TH participants, 500~ms in HA users, and 600~ms in CI users, weakest in the CI group. All three groups exhibited attentional modulation of surprisal tracking. Word frequency yielded significant encoding gains in quiet only for TH and HI listeners, not CI users. We show for the first time that CI users exhibit significant, attention-modulated neural responses to word surprisal. More broadly, surprisal tracking was present across all hearing-impaired groups but weakest in CI users, the cohort with the greatest difficulty understanding speech in noise.

Resultado principalEl resumen no menciona limitaciones.

Apareció: jueves, 19 de noviembre. bioRxiv. Preprint, todavía sin revisión por pares.

DOI: 10.64898/2026.11.12.765375