Temporal masking and rollover in the neural code for speech with and without hearing loss

Abstract

Natural sounds, such as speech, are complex time-varying waveforms containing information critical to how we communicate with each other and navigate the external world. Hearing loss results in a breakdown of this information and causes distortions in the neural code. As a result, perception of complex sounds such as speech is compromised. This problem is further complicated by the fact that sound intensity varies in natural settings, both in quiet and in noisy backgrounds. Somewhat paradoxically, despite increased audibility at high sound intensities, perception and discrimination of speech is actually diminished, especially in the presence of background noise. This phenomenon is known as rollover of speech and its neural basis is poorly understood in both normal-hearing listeners and hearing-impaired listeners. Here we performed in-vivo electrophysiology in awake and anaesthetized Mongolian gerbils (Meriones Unguiculatus) to investigate how hearing loss affects the neural encoding of speech. We presented 22 Vowel-Consonant-Vowel (VCV) syllables to the gerbil and recorded neural responses from the inferior colliculus (IC). We used a k-nearest neighbor neural classifier to investigate whether IC neurons could discriminate between different consonants in normal hearing (NH) and noise-exposed hearing-loss (HL) animals. We found that neural correlates of perceptual rollover were present in the IC and that performance in discrimination decreased when VCVs were presented in background noise when compared to in quiet. Furthermore, we found that forward masking played a prominent role in shaping neural responses and discrimination between various consonants in NH and HL animals. These results suggest there is a critical trade-off in listening between audibility and rollover mediated by temporal masking.