NEURONAL DYNAMICS DURING INITIAL LEARNING VS EXPERT PERFORMANCE OF A SENSORIMOTOR TRANSFORMATION

Learning operates across multiple timescales, from rapid initial association to long-term consolidation. To dissect these processes, we developed a rapid learning paradigm where mice pre-trained in an auditory-detection task learned a novel whisker-reward association within their first exposure session, with performance improving over subsequent days. One cohort learned whisker-reward associations while for the other cohort the whisker stimulus was not associated to reward. We recorded neuronal activity using up to five simultaneous Neuropixels probes targeting motor cortex, somatosensory areas, prefrontal cortex, striatum, thalamus, and superior colliculus in over 80 mice during initial learning and in over 30 expert mice. To map sensory representations, passive whisker and auditory stimuli were presented at the beginning and end of each recording session.
Comparing passive trials before and after the initial learning session, we observed decreased whisker representations in non-rewarded mice across the caudate putamen, secondary whisker somatosensory and motor cortices, whereas rewarded mice maintained more stable responses. Comparing expert to initial sessions revealed consolidation of the learning of the new association. Expert rewarded mice demonstrated elevated whisker-evoked neuronal responses in the passive trials at the start of the expert session across primary whisker somatosensory cortex, caudate putamen, and superior colliculus, suggesting consolidation of initial representations. In contrast, expert non-rewarded mice overall showed lower neuronal responses to the whisker stimulus.
Ongoing analyses investigate how population dynamics and inter-area functional connectivity evolve from rapid learning to consolidated expertise, aiming to reveal how distributed brain networks support reward-based sensorimotor learning across timescales.