Generalization and multirate models of motor adaptation.

Area of Science:

• Motor control
• Computational neuroscience
• Human movement science

Background:

• Movement adaptation exhibits generalization, assessed via post-adaptation or trial-by-trial measures.
• The distinct mechanisms underlying these generalization types remain underexplored.
• Understanding these differences is crucial for interpreting neural representations of movement.

Purpose of the Study:

• Develop a computational framework to identify when a two-state model fits adaptation data.
• Investigate how two distinct generalization measures relate to fast and slow adaptation processes.
• Explore implications for neural representations of movement direction.

Main Methods:

• Utilized statistical model selection to differentiate single-process from two-process models in single-target learning.
• Constructed a two-state model for multi-target motor adaptation.
• Systematically varied learning and retention coefficients to analyze model performance.

Main Results:

• The computational framework successfully discriminated between single- and two-process models.
• Post-adaptation generalization primarily reflects the slow adaptation process.
• Trial-by-trial generalization is most sensitive to the fast adaptation process.

Conclusions:

• Movement adaptation can be characterized by distinct fast and slow processes.
• The two generalization measures (post-adaptation and trial-by-trial) probe these separate processes.
• These findings suggest different neural representations underlie fast and slow adaptation components.