Non-stationary Salience Processing During Perceptual Training in Humans

Area of Science:

• Cognitive Neuroscience
• Perceptual Learning
• Computational Vision

Background:

• Perceptual learning enhances sensory task performance.
• Task difficulty influences learning generalization, but its effect on learning curves and discrimination efficiency is unclear.
• Understanding how changing task difficulty shapes learning trajectories is crucial for optimizing training protocols.

Purpose of the Study:

• Investigate the impact of sustained changes in task difficulty on visual perceptual learning.
• Compare learning curves and efficiency under increasing versus decreasing stimulus similarity training.
• Elucidate the underlying computational mechanisms driving distinct learning trajectories.

Main Methods:

• Developed monocular visual training programs with increasing (SIMinc) and decreasing (SIMdec) stimulus similarities.
• Administered 10-day training protocols to human participants.
• Applied associative and non-stationary sequential sampling models to analyze learning curves and decision-making processes.

Main Results:

• Significant performance improvements observed in all participants after 10 days of training, with near-complete transfer to untrained eyes.
• Distinct learning curves emerged between SIMinc and SIMdec groups, best explained by an associative learning model incorporating stimulus salience.
• The SIMinc group exhibited a faster evidence accumulation rate compared to the SIMdec group, as revealed by a non-stationary sequential sampling model.

Conclusions:

• Different learning trajectories, shaped by training paradigms, modulate attentional salience processing.
• Distinct stimulus processing efficiencies arise from varying learning curves.
• The interplay between learning trajectories and attentional guidance is fundamental to visual perception and decision-making.