We computed decoding chance level by randomly shuffling laps labels. This revealed that the visual cortex maintained a high sensory discrimination regardless of behavioural engagement, whereas the dentate gyrus required active navigation for effective discrimination.

We employed the CEBRA method for latent embeddings, and built a binary decoder trained to predict which environment was being explored based solely on neuronal activity.

During each session, either two very distinct environments were alternating at random (A-B) or two more similar corridors (A-A’). Animal received a reward when stopping long enough within an uncued location. This location was different in environment A and A’/B, forcing the animal to discriminate.

We used 2-photon imaging to record mice navigating in virtual-reality.
🎮 During active navigation, their position along the track was controlled by the distance run on the wheel.
📺 In passive exposure, instead of converting speed into a position, we broadcasted a replay from another session.