Minimizing Precision-Weighted Sensory Prediction Errors via Memory Formation and Switching in Motor Adaptation

Area of Science:

• Neuroscience
• Motor Control
• Machine Learning

Background:

• Humans learn internal models to predict sensory consequences of actions.
• Sensory prediction errors challenge existing models, requiring updates or new model creation.
• Distinguishing between body and environmental changes is crucial for adaptation.

Purpose of the Study:

• Propose a decision-making process for selecting and updating internal models during motor adaptation.
• Investigate how prediction errors, weighted by precision, guide model selection.
• Explain adaptation phenomena like aftereffects, savings, and error-clamp data.

Main Methods:

• Computational simulations of a proposed decision-making model.
• Visuomotor adaptation experiment with human participants.
• Analysis of model predictions against experimental data.

Main Results:

• The model accurately predicts short aftereffects and large savings for significant perturbations.
• It explains gradual decay in motor memories for small perturbations versus initial lack of decay for large ones (error-clamp).
• Individual differences in adaptation are linked to preferential updating of body or perturbation models.

Conclusions:

• Motor adaptation involves a general learning principle of creating new memories when existing ones fail to predict sensory data accurately.
• The proposed model offers a unified framework for understanding diverse motor adaptation behaviors.
• This work provides insights into CNS strategies for handling unexpected sensory outcomes and individual learning variations.