Robustness to Noise in the Auditory System: A Distributed and Predictable Property

Abstract

Background noise strongly penalizes auditory perception of speech in humans or vocalizations in animals. Despite this, auditory neurons are still able to detect communications sounds against considerable levels of background noise. We collected neuronal recordings in cochlear nucleus (CN), inferior colliculus (IC), auditory thalamus, and primary and secondary auditory cortex in response to vocalizations presented either against a stationary or a chorus noise in anesthetized guinea pigs at three signal-to-noise ratios (SNRs; −10, 0, and 10 dB). We provide evidence that, at each level of the auditory system, five behaviors in noise exist within a continuum, from neurons with high-fidelity representations of the signal, mostly found in IC and thalamus, to neurons with high-fidelity representations of the noise, mostly found in CN for the stationary noise and in similar proportions in each structure for the chorus noise. The two cortical areas displayed fewer robust responses than the IC and thalamus. Furthermore, between 21% and 72% of the neurons (depending on the structure) switch categories from one background noise to another, even if the initial assignment of these neurons to a category was confirmed by a severe bootstrap procedure. Importantly, supervised learning pointed out that assigning a recording to one of the five categories can be predicted up to a maximum of 70% based on both the response to signal alone and noise alone.

Significance Statement In daily situations, humans and animals are faced with various background noises in which they have to detect behaviorally salient signals. Noise resistance is often viewed as an emergent property of cortical networks, but only a few studies have characterized the relative contribution of cortical and subcortical neurons. Our results demonstrate that the neuronal resistance to noise is distributed along the auditory system with a more important fraction of robust neurons in subcortical structures compared with auditory cortex, and is relatively well predictable based on the responses to the signal alone and the noise alone. Our results also suggest that noise-invariant representations of communication sounds coexist with accurate noise representations, which are detected as early as the cochlear nucleus (CN).