40 mice were trained to choose among three potentially rewarding actions (nose-poking into three ports). The reward probabilities associated with the three actions were shifted twice during the experiment for each animal, with extensive training between each shift of reward contingencies. Each file within the “behaviour” folder contains the timeline of events such as trial initiations, port entries and exits, visual and auditory cues and reward deliveries from one experiment session, indexed in the file name by animal ID and data and time of the session.

Photometry signals of dLight levels were recorded from 5 out of the 40 mice during most of their sessions with behavioural data. Each file within the “recordings” folder contains the raw signal amplitudes data throughout each session, indexed in the file name by animal ID, brain region and hemisphere of the fibre placement and date and time of the session.

The code for parsing, extracting and processing the raw data files which produced the processed data ready for subsequent analysis is available as a separate dataset and at https://github.com/Bogacz-Group/DLS_APE. More details on data collection and processing methods will be published in a future research article.