Predictions not commands: active inference in the motor system.

Area of Science:

• Neuroscience
• Motor Control
• Computational Neuroscience

Background:

• Descending motor cortex projections share characteristics with visual cortex backward connections.
• This presents a paradox, as motor commands are not typically considered modulatory.
• Helmholtz's theory of perception as inference provides a framework to resolve this.

Purpose of the Study:

• To resolve the paradox of motor cortex projections exhibiting backward connection features.
• To explain motor system function using principles of active inference.
• To investigate the anatomical and physiological parallels between motor and sensory systems.

Main Methods:

• Functional characterization of the motor system based on predictive processing.
• Comparison of anatomical, laminar, topographic, and physiological features of motor and sensory cortices.
• Application of active inference principles to explain motor control.

Main Results:

• Motor behavior is explained as inference on proprioceptive sensation causes, utilizing active inference.
• Descending projections convey proprioceptive predictions, analogous to predictions in sensory cortex.
• The motor system exhibits a hierarchical structure with asymmetric ascending (driving) and descending (modulatory) connections, mirroring sensory systems.

Conclusions:

• Motor cortex functions via predictive processing, sending predictions rather than direct commands.
• Motor reflexes can be understood as mechanisms for minimizing prediction errors.
• This active inference framework offers a principled explanation for the agranular nature of motor cortex.