Complexity and specificity of experimentally-induced expectations in motion perception

Area of Science:

• Cognitive Science
• Neuroscience
• Psychology

Background:

• Perception is influenced by prior expectations (priors) about the world.
• Previous research demonstrated automatic learning of motion direction priors from statistical regularities in visual stimuli.

Purpose of the Study:

• To investigate the specificity of learned priors, specifically if different motion direction expectations can be learned for stimuli of different colors.
• To determine how varying statistical structures of motion distributions for different colored stimuli affect prior learning.

Main Methods:

• Subjects viewed interleaved moving dot displays of red and green colors, each with different motion direction distributions.
• Experimental conditions varied the structure of these distributions (e.g., bimodal vs. uniform).
• Behavioral responses were analyzed to infer the influence of learned priors, particularly during stimulus absence.

Main Results:

• When motion distributions differed significantly (bimodal vs. uniform), subjects formed a single, generalized prior.
• When motion distributions were similarly structured (e.g., both bimodal), subjects formed two distinct, color-specific priors.
• These distinct priors significantly impacted behavior, demonstrating learned specificity.

Conclusions:

• The brain can form distinct, stimulus-specific priors based on learned statistical regularities.
• The structure of the statistical learning environment influences whether generalized or specific priors are acquired.
• Results suggest potential suboptimality in statistical learning under certain conditions, explainable by a Bayesian framework.