Behavioral and neural mechanisms of optimal sensory discrimination

Recent studies show that behaving animals transition between discrete strategies during sensory decision-making and that arousal modulates task performance. However, it remains a challenge to understand (i) the relationships between arousal and task strategy during sensory discrimination and (ii) the neural mechanisms underlying optimal performance states. To address these topics, we combined behavioral, neurophysiological, and computational approaches to uncover the neural mechanisms underlying arousal-dependent sensory discrimination. We implemented two-alternative-choice tasks, prompting mice to categorize auditory or visual stimuli as they traverse a range of arousal levels. We show that mice transition between minutes-long optimal, sub-optimal, and disengaged performance states, and selectively maintain persistent optimal states. We also find that the likelihood of optimal state occupancy exhibits a robust inverted-U relationship with pupil diameter, and that transitions into the optimal state are accompanied by marked decreases in the trial-to-trial variability of multiple arousal measures. These relationships hold in both auditory and visual tasks, though mice performing the visual task have larger optimal pupil sizes than their auditory counterparts, suggesting modality-dependent arousal modulation. To test whether arousal-induced performance modulations occur at early stages of the cortical hierarchy, we analyzed Neuropixels recordings from auditory cortex of passively-behaving mice presented with pure tones. We show that tone frequency is best decoded from population activity during periods of intermediate pupil size, suggesting that optimal performance states extracted from the behavioral task may be underpinned by arousal-regulated neural representations that are formed even in passive settings. Using a circuit model, we then explain how arousal could modulate neural dynamics in a manner consistent with the observed inverted-U relationship between sound discriminability and pupil size. Taken together, our findings show that the regulation of arousal helps to establish optimal neural and behavioral performance states for sensory discrimination.