Abstract
Learning is an essential mechanism that adjusts neuronal processing to adapt animal behavior to novel contexts. In a recent study, we have shown that learning a visual task is associated with modifications of the neuronal responses in the mouse primary visual cortex (V1) that favor the representation of stimuli relevant to the task but distort the space of orientation representation. Here, we investigated how this distortion of the orientation representation space is established during the time course of training. Our previous results showed that representations of rewarded and non-rewarded orientations were more accurate and stable across trials in trained mice. This improvement was due to a distortion of the orientation representation space such that stimuli flanking the task-relevant orientations were represented as the task stimuli themselves. This suggested that flanking orientations were generalized as the task cues by the trained mice. We designed a new experiment to determine the time course of this distortion in trained mice. We imaged L2/3 of V1 while training for and after disengaging from the task. We found that the distortion was still present a few days after the animals were disengaged but decayed over days. The disappearance of the distortions in the representation of orientation did not impair the mouse performance when they were asked to perform the Go/NoGo orientation discrimination task again. Our results suggest that the population activity in V1 returns to an unbiased state once expertise is reached, and that this return does not influence the behavioral performance.