Neural substrate of dynamic Bayesian inference in the cerebral cortex.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Cognitive Neuroscience

Background:

• Dynamic Bayesian inference is essential for inferring environmental states with limited sensory information.
• The posterior parietal cortex (PPC) and posteromedial cortex (PM) are implicated in complex cognitive functions.
• Understanding neural mechanisms of prediction and updating is key to deciphering brain function.

Purpose of the Study:

• To investigate the role of mouse PPC and PM in implementing dynamic Bayesian inference.
• To examine how these brain regions perform state prediction and evidence updating during a goal-reaching task.
• To elucidate the neural basis of mental simulation in the cerebral cortex.

Main Methods:

• Optical imaging of neuronal activity in mouse PPC and PM (layers 2, 3, 5) within an acoustic virtual-reality system.
• A goal-reaching task involving anticipatory licking in response to auditory cues.
• PPC silencing experiments to assess its causal role.
• Probabilistic population decoding to analyze neural representations of goal distance.

Main Results:

• Mice exhibited anticipatory licking, demonstrating prediction of reward location, which was impaired by PPC silencing.
• Neurons in PPC and PM represented goal distances, particularly during periods of sound omission (prediction).
• Neural prediction accuracy improved with the presentation of cue sounds, indicating effective updating based on sensory evidence.

Conclusions:

• The PPC and PM are critical for dynamic Bayesian inference, supporting both prediction and updating of hidden environmental states.
• These findings demonstrate how the cerebral cortex utilizes action-dependent dynamic models for mental simulation.
• The study provides insights into the neural computations underlying adaptive behavior in uncertain environments.