Invariant cortical representations of sounds in ambient noise

In our everyday lives, sounds of interest like speech, alarms and other foreground sounds are effortlessly listened to in noisy background environments. This perceptual capability implies that internal representations of sounds are robust and unaffected by the acoustic background. This process, called noise invariance, emerges throughout cortical stages. In particular previous works highlighted noise invariant representation of sounds throughout human auditory cortex (Kell & McDermott. 2019. Nat. Comm.). In this study and many others, sounds that are familiar to humans have been the main focus of investigation. Here our working hypothesis is that noise invariance in cortical representations is shaped by familiarity and exposure to sounds. This implies that: (i) cortical responses are noise invariant for specific relevant categories of sounds (like speech for humans, or vocalizations for other animals), and (ii) that noise invariance develops across development and passive exposure to environmental backgrounds (like city noise for humans, or forest background for other animals).

The primary objective of this project is to explore the degree of noise invariance in sound representations across the auditory cortex. The proposal combines high-resolution functional UltraSound (fUS) neuroimaging in awake ferrets (Bimbard et al, 2018. eLIFE) with computational analysis (dimensionality reduction, multi-voxel pattern analysis). The experiments will explore how sound encoding changes across cortical fields of the auditory cortex. More specifically we want to compare cortical responses to sounds of different ecological relevance across background noise. We will probe if invariance to noise in neural representations is more pronounced for ecologically relevant sounds (for instance conspecific communication sounds). Follow-up research will address whether exposure to background noise during development is a necessary condition for noise invariant sound encoding (Homma et al, 2020. Cell reports).

Please contact Yves Boubenec (boubenec@ens.fr) for additional information, including ongoing manuscript of the laboratory related to this project. The internship can be taylor-cut to your particular interests and backgrounds, from pure data analysis of existing datasets to animal behavior and training on new tasks. This internship can be indiscriminately conducted in English or in French.