Multi-level integration for predictive inference: mPFC connectivity in action anticipation across representational...

Area of Science:

• Neuroscience
• Cognitive Science
• Motor Control

Background:

• Action anticipation is often explained by the action observation network (AON) and dual visual streams.
• This sensorimotor view neglects the inference needed for predicting outcomes from incomplete information.
• The interaction between sensorimotor systems and the medial prefrontal cortex (mPFC) for inference under uncertainty is not well understood.

Purpose of the Study:

• To investigate the connectivity between the mPFC and sensorimotor systems.
• To determine how different representational profiles influence this architecture.
• To understand the neural mechanisms of action anticipation and predictive inference.

Main Methods:

• Novice participants underwent observation-based or execution-based training for table tennis serve anticipation.
• fMRI scanning was used during anticipation tasks with point-light displays and full-body videos.
• Behavioral modeling, neural activation, and dynamic causal modeling (DCM) were analyzed.

Main Results:

• DCM revealed an invariant architecture where mPFC integrates AON-based action representations and visual stream feature information for predictive inference.
• Representational profiles modulated this architecture: perception-dominant representations prioritized context, while motor-dominant representations focused on kinematics.
• Experiential learning influenced information weighting and modulatory dynamics, with distinct mPFC engagement based on stimulus richness and representation type.

Conclusions:

• Action anticipation utilizes a universal multi-level connectivity architecture involving the mPFC for cognitive inference.
• Experiential learning, through training, shapes how information is weighted and processed within this system.
• The interplay between sensorimotor networks and the mPFC allows for flexible and adaptive action anticipation.