Explicit learning based on reward prediction error facilitates agile motor adaptations.

Area of Science:

• Motor control and learning
• Computational neuroscience
• Human adaptation

Background:

• Motor learning is driven by sensory prediction error (sensorimotor adaptation) and reward prediction error (reward learning).
• Understanding the interplay between these two error types is crucial for comprehending motor skill acquisition.

Purpose of the Study:

• To investigate the distinct characteristics and interconnections between sensorimotor adaptation and reward learning.
• To differentiate the neural and computational mechanisms underlying adaptation based on sensory versus reward prediction errors.

Main Methods:

• A visuomotor paradigm was employed where participants performed arm movements under perturbed conditions.
• Subjects received either sensory prediction error, end-point error, or binary reward feedback.
• A computational model was developed to distinguish between sensorimotor and reward learning processes.

Main Results:

• Participants adapted to novel perturbations regardless of the error type, converging to similar movement patterns.
• Sensorimotor adaptation induced significant aftereffects, whereas reward learning resulted in smaller aftereffects, indicating less impact on the internal model.
• Adaptation to randomly changing perturbations was faster with reward learning (explicit strategies) compared to sensorimotor adaptation.
• Computational modeling revealed distinct adaptation processes for sensorimotor and reward learning.

Conclusions:

• Sensorimotor adaptation and reward learning are distinct processes influencing motor control differently.
• Sensorimotor adaptation modifies the internal model, leading to pronounced aftereffects.
• Reward learning relies on explicit strategies without significant aftereffects, but enables faster adaptation to changing environments.
• Humans integrate both sensorimotor and reward learning for effective motor task adaptation.