8–12 Sept 2026
Europe/Vienna timezone

Auditory Nerve Fiber Loss And Vowel Discrimination: Evidence From Machine Learning Analysis Of Single-Unit Recordings

FA2026/912
9 Sept 2026, 16:20
20m
Saal 12A (Messe Congress Graz)

Saal 12A

Messe Congress Graz

Speaker

Daniil Kiselev (Institute for Neuroscience Montpellier)

Description

Traditionally, vowel encoding in the auditory nerve relies on type I spiral ganglion neurons (SGNs) with characteristic frequency (CF) close to global spectral peaks (formants, F). This study investigates contributions of subpopulations of auditory nerve fibers to coding spectral and temporal features of vowels under normal conditions and following kainate-induced cochlear synaptopathy. Four synthetic vowels with varied spectral contrast were used to stimulate the SGNs: /i/, /y/ (F2>1.5 kHz), /o/, and /u/ (F2<1.5 kHz), presented at 70 and 50 dB SPL. Using machine learning methods on single-fiber recordings of auditory nerve fibers’ responses from Mongolian gerbils, we demonstrate that, phenotype-independent progressive ablation of up to 90% only reduced discrimination performance by 5%. On the contrary, loss of ANFs of low spontaneous firing rate (<18 spikes/s; low-SR) and high-CF (>8 kHz) fibers drastically degrades classifier’s performance (by 35% and 30%, respectively). Furthermore, kainate-induced increase of mean population SR by up to 50 spikes/s led to 25% drop in performance even without depletion of ANFs. These findings highlight the functional significance of low-SR and high-CF fibers in vowel coding in the auditory nerve and demonstrate the detrimental effects of their progressive loss on vowel discrimination.

Authors

Daniil Kiselev (Institute for Neuroscience Montpellier) Etienne Gaudrain (CRNL, CNRS UMR5292, Inserm U1028, Université Lyon 1) Sarah Verhulst (Ghent University) Jean-Luc Puel (University of Montpellier) Jérôme Bourien (University of Montpellier)

Presentation materials

There are no materials yet.