On what it means to automatize a rule.

Area of Science:

• Cognitive Psychology
• Neuroscience
• Computational Neuroscience

Background:

• Rule-guided behaviors are initially controlled by prefrontal cortex networks.
• Practice can lead to the automatization of these behaviors.
• A neurocomputational model proposes a shift to a faster, direct network during automatization.

Purpose of the Study:

• To investigate the nature of automatization in rule-guided behavior after extensive practice.
• To test and refine a neurocomputational model of behavioral automatization.
• To determine if abstract rules or stimulus-response associations are learned.

Main Methods:

• Conducted two experiments with approximately 633,000 categorization trials.
• Designed experiments to specifically test predictions of a neurocomputational model.
• Analyzed behavioral data to infer the underlying learning mechanisms.

Main Results:

• Automatization did not involve learning an abstract, verbal-based strategy.
• Learning involved linking stimuli with shared visual features to a common motor response.
• Results support the neurocomputational model of automatization.

Conclusions:

• Behavioral automatization under rule guidance involves direct stimulus-response associations, not abstract rule abstraction.
• The neurocomputational model is supported, suggesting direct visual cortex to premotor cortex pathways are key.
• These pathways appear restricted to relevant visual dimensions, optimizing the automatic response.