Perceptual learning improves discrimination but does not reduce distortions in appearance.

Area of Science:

• Cognitive Neuroscience
• Psychology
• Computational Neuroscience

Background:

• Human perceptual sensitivity often improves with training, a process termed perceptual learning.
• Perceptual appearance, the subjective sense of stimulus magnitude, is another key perceptual dimension.
• The relationship between training-induced sensitivity improvements and appearance accuracy remains unclear.

Purpose of the Study:

• To investigate whether training-induced improvements in perceptual sensitivity are accompanied by more accurate perceptual appearance.
• To examine how training affects both discrimination (sensitivity) and estimation (appearance) of near-horizontal motion directions.
• To explore the underlying mechanisms of perceptual learning using computational modeling.

Main Methods:

• Participants performed discrimination and estimation tasks before and after training.
• Training groups included discrimination task training, estimation task training, and a control group.
• A computational observer model was developed to analyze perceptual learning mechanisms.

Main Results:

• Observers trained in either discrimination or estimation showed improved discrimination accuracy.
• Estimation repulsion did not decrease post-training; it persisted or increased.
• Distortions in perception were found to be exacerbated after perceptual learning.

Conclusions:

• Perceptual learning can enhance distinctions between categories, potentially increasing perceptual distortions.
• Computational modeling suggests learning increases the precision of neural representations.
• Findings indicate that perceptual learning may enhance category distinctions, impacting real-world perception.