This prior found in the neural activity is reflected in the behavior. Higher decoded priors are associated with a greater proportion of rightward choices. (7/8)

Two different modalities (electrophysiology and widefield imaging [WFI]), different mice but same task! Having one shared canonical task allows us to run the same experiment in different modalities. We found very similar prior representations between modalities. (6/8)

We found that prior information is encoded in more than 20% of brain regions which, remarkably, span all levels of processing, from early sensory areas (LGd, VISp) to motor regions (MOs, MOp, GRN) and high level cortical regions (ACCd, ORBvl). (5/8)

At the other extreme, the brain might operate like a very large Bayesian network, in which probabilistic inference is the modus operandi in all brain regions and inference can be performed in all directions. (4/8)

Where is prior information about the state of the world represented in the brain? At one extreme, the brain might combine prior information with sensory evidence in high level decision-making brain regions, right before decisions are turned into actions. (3/8)

In the IBL task, the prior probability that the stimulus appeared on the right side switched in a random and uncued manner between 0.2 and 0.8 in blocks of trials. Mice leverage their prior knowledge of this block structure to improve their performance and obtain more rewards 🧀 (2/8)

The neural representation of movement and feedback/reward consumption were particularly widespread, reflecting brain-wide changes in neural activity. 🧠 (6/8)

The upcoming choice (before the first movement on a trial) was represented across the cortex, basal ganglia, thalamus, midbrain, hindbrain, and cerebellum. 🤸‍♂️(5/8)