Neural integration of acoustic statistics enables detecting acoustic targets in noise

Abstract Sound detection amidst noise presents an important challenge in audition. Many naturally occurring sounds (rain, wind) can be described and predicted only statistically, so-called sound textures. Previous research has demonstrated the human ability to leverage this statistical predictability for sound recognition, but the neural mechanisms remain elusive. We trained mice to detect vocalizations embedded in sound textures with different statistical predictability, while recording and optogenetically modulating the neural activity in the auditory cortex. Mice showed improved performance and neural encoding if they could sample the statistics longer per trial. Textures with more exploitable structure, specifically higher cross-frequency correlations improved behavioral performance as well as neural representation of background and vocalization. Activating parvalbumin-positive (PV) interneurons had an asymmetric effect, improving detection and neural encoding of vocalizations for low correlations, and impoverishing them for high cross-frequency correlations. In summary, mice exploit stimulus statistics to improve sound detection in naturalistic background noise, reflected in behavioral performance and neural activity, relying on PV interneurons for temporal integration. Highlights Mice integrate statistical information indicated by behavior and neural activity Encoding of background sounds stays stable in A1, while vocalizations are enhanced High cross-frequency correlations improve target detection and neural encoding Activating PV cells improves detection of sounds with low cross-frequency correlations Competing Interest Statement The authors have declared no competing interest. Footnotes We noticed that through a coding error a subset of stimuli was accidentally excluded in the computation of Figure 5F, leading to an imbalance in the included conditions. We have fixed this error, which did not lead to a qualitative change, but increased the quantitative difference in the vocalization decoding in high CFC vs low CFC. We have therefore replaced the reconstruction result for the vocalizations with the more specific SVM decoding result for panel 5E and updated figure 5F, as well as the associated results text, figure caption and some minor changes in the discussion.