Previous Motor Actions Outweigh Sensory Information in Sensorimotor Statistical Learning

Area of Science:

• Cognitive Neuroscience
• Decision Making
• Motor Control

Background:

• Humans leverage statistical priors from experience to enhance decision-making.
• The precise mechanisms and representations of prior learning, particularly its dependence on sensorimotor systems, remain unclear.
• Previous research indicates both modality-specific and modality-general learning patterns.

Purpose of the Study:

• To investigate the learning and representation of spatial priors using a saccadic eye movement task.
• To determine whether prior learning is modality-specific (motor actions) or modality-general (visual hints).
• To examine the transferability of sensorimotor priors across different motor response contexts.

Main Methods:

• Utilized a saccadic eye movement task with unsupervised learning through visual hints.
• Employed a model-comparison approach to analyze participants' prior knowledge representation.
• Compared prior knowledge transfer between two distinct motor contexts (target-directed vs. anti-target-directed eye movements).

Main Results:

• Prior knowledge was predominantly represented by participants' previous motor actions, with minimal influence from visual hints.
• Despite similar learning curves and strategies, knowledge transfer between motor contexts was incomplete, evidenced by performance and confidence drops.
• Sensorimotor priors formed over a few trials are preferentially based on motor representations.

Conclusions:

• Human statistical sensorimotor priors are primarily formed from internal representations of previous motor actions.
• Past sensory information has a limited role in forming these short-term priors.
• The findings highlight the action-centric nature of sensorimotor learning and decision-making in humans.