Error-driven learning in statistical summary perception.

Area of Science:

• Cognitive Psychology
• Neuroscience
• Human Perception

Background:

• Humans effectively process statistical summary features of object groups, like center-of-mass.
• Acquisition and refinement of summary perception through experience remain underexplored.
• Understanding how experience shapes statistical perception is crucial for explaining complex interactions.

Purpose of the Study:

• To investigate if training and task feedback enhance the perception of statistical summary features.
• To determine the specific types of feedback and interaction that improve centroid estimation.
• To explore the role of sensory-prediction error in perceptual learning.

Main Methods:

• Participants practiced estimating object centroids on a touchscreen.
• Training involved varied feedback types (e.g., vector error) and interaction methods (e.g., reaching movements).
• Perceptual discrimination of centroids was tested before and after training.

Main Results:

• Vector error feedback significantly improved centroid perceptual discrimination.
• This improvement was observed only when vector error feedback was paired with reaching movements.
• Other feedback types or interaction methods did not yield significant improvements.

Conclusions:

• Sensory-prediction error, signaled by vector feedback during action, is critical for refining perceptual representations of statistical features.
• Motor interaction is essential for effective perceptual learning of object group properties.
• This study highlights a novel mechanism for how experience tunes our ability to perceive collective object properties.