Cerebral hierarchies: predictive processing, precision and the pulvinar.

Area of Science:

• Computational neuroscience
• Neuroanatomy
• Neurophysiology

Background:

• The brain is increasingly understood as an organ of inference, operating via hierarchical predictive coding.
• Current models require further elaboration on how inferences are dynamically coordinated and contextualized.

Purpose of the Study:

• To explore the computational principles underlying the orchestration of neuronal inferences.
• To investigate the role of modulatory gain control in coordinating message passing for optimized computational goals.

Main Methods:

• Extending hierarchical predictive coding frameworks.
• Formal analysis and simulations of feature-ground segregation.
• Investigating neurobiological substrates of precision-engineered dynamics.

Main Results:

• Neuronal computations are shaped by modulatory gain control, selecting and coordinating prediction error signals.
• A hierarchical anatomy of extrinsic connections comprises distinct driving (content encoding) and modulatory (context setting) classes.
• Simulations demonstrate the implications of this distinction for feature-ground segregation.

Conclusions:

• Modulatory gain control is essential for dynamically coordinating neuronal communication and optimizing computational goals.
• The distinction between driving and modulatory connections provides a framework for understanding brain circuitry.
• Precision-engineered dynamics, particularly involving the pulvinar and attention, are critical for contextualizing neuronal representations.