Mapping shape to visuomotor mapping: learning and generalisation of sensorimotor behaviour based on contextual...

Area of Science:

• Cognitive Neuroscience
• Motor Control
• Human-Computer Interaction

Background:

• Humans exhibit dual-adaptation, learning and storing multiple visuomotor mappings.
• Context cues facilitate switching between learned visuomotor mappings.
• Understanding cue-mapping association and generalization remains limited.

Purpose of the Study:

• Investigate associative learning between context cues and visuomotor mappings.
• Determine how visuomotor learning generalizes to novel conditions.
• Compare explicit rule extraction versus implicit Bayesian learning models.

Main Methods:

• Participants performed a rapid target-pointing task with shape-dependent visual feedback offsets.
• Target shapes varied along a continuum from spiky to circular.
• Post-training generalization was tested on novel shapes without feedback.

Main Results:

• Learning visuomotor associations requires explicit cue awareness.
• Generalization favored mappings of the most similar trained shape, supporting a Bayesian categorization model.
• Learning slowed with increased training pairs, consistent with the Bayesian model.

Conclusions:

• A Bayesian learning model accurately predicts generalization in dual-adaptation tasks.
• Visuomotor learning generalizes via categorization based on stimulus similarity.
• This model offers a potential mechanism for simultaneously learning multiple visuomotor mappings.