Category-like representation of statistical regularities allows for stable distractor suppression

Area of Science:

• Cognitive Psychology
• Neuroscience
• Computational Modeling

Background:

• Statistical learning enables suppression of attentional capture by predictable salient distractors.
• This learning is often inflexible, persisting even without contextual cues.
• The precise learning mechanisms and probability representations are not fully understood.

Purpose of the Study:

• To replicate learned location-based distractor suppression.
• To investigate the underlying computational mechanisms of this suppression.
• To determine how distractor probability influences attentional capture and target selection.

Main Methods:

• Replication of learned distractor suppression in two experiments.
• Computational modeling to compare different learning mechanisms (frequency summation, reinforcement learning, categorical response).
• Analysis of attentional capture and target selection performance.

Main Results:

• Confirmed reduced attentional capture by high-probability distractors.
• Observed impaired target selection at high-probability distractor locations.
• Found that a combination of global response time decay and categorical learning best explained the data.

Conclusions:

• Learned distractor suppression is robust and linked to location probability.
• The magnitude of suppression is more influenced by overall probability differences than by trial-by-trial history.
• A categorical learning mechanism appears central to location-based distractor suppression.