Speaker
Description
Animal studies suggest cochlear synaptopathy predominantly affects high-threshold, low spontaneous rate auditory nerve fibers. This pathology has been proposed as a contributor to speech-in-noise deficits despite normal audiometric thresholds. A candidate explanation is that remaining fibers compensate by shifting encoding to alternative frequency channels, but this compensation depends on channels that noise can saturate.We test this computationally by quantifying how selective fiber loss reshapes phoneme level neural representations, and by evaluating perceptual consequences using a neural vocoder and a speech foundation model for transcription.Over 10,000 phonemes from the TIMIT corpus were passed through a model of the auditory nerve and the inferior colliculus, across five fiber-loss configurations and four noise conditions. In quiet, midbrain processing compensated most peripheral distortion, leaving few phonemes affected. In noise, compensation collapsed: most phonemes were distorted at the inferior colliculus, and the increase in speech errors relative to normal hearing nearly tripled. The most vulnerable category shifted from stops to voiced fricatives, because low spontaneous rate fibers provide precise timing in quiet and dynamic range in noise.Because these fibers also drive the olivocochlear efferent reflex, their loss may compound the deficit through reduced noise protection, revealing deficits that shift with listening conditions and escape the audiogram.