Generalized prefrontal task-progress sequences efficiently encode task structure

Abstract

The medial prefrontal cortex (mPFC) occupies a central role in executive function including decision-making, rule learning and cognitive flexibility. Here, we investigated how mPFC populations represent the animal's progression along sequential task stages. Freely moving mice performed an odour-guided two-choice task requiring sensory sampling, action selection, navigation, and reward collection. Population activity in the mPFC formed sequences that evolved through cue sampling, movement towards the reward sites, and reward-related epochs. These sequences exhibited a shared temporal organization across trial types, pointing to a generalized encoding of task progress by prefrontal neurons. Artificial neural networks trained on the same task developed task-progress sequences, and experimentally disrupting task-progress tuning in these networks impaired learning efficiency. Our results indicate that mPFC activity organizes behavioural episodes along correlated trajectories and suggest that task-progress tuning offers an efficient solution for the encoding of task space during multi-stage behaviour.