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

Seitz, Catherine W.; Sali, Anthony W. · 2025 · Crossref

DOI: 10.3389/fpsyg.2025.1598594

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study investigates the cognitive mechanisms underlying learned distractor suppression, specifically how individuals represent statistical regularities to inhibit attentional capture by salient, task-irrelevant stimuli. While previous research established that observers suppress distractors appearing at high-probability locations, the nature of these probability representations—whether they are continuous, flexible, or rigid—remained unclear. The authors sought to determine if suppression is driven by continuous frequency summation, reinforcement learning (RL) prediction errors, or a categorical, all-or-nothing rule. The researchers conducted two experiments using an additional singleton search task. Participants searched for a unique shape target among distractors, with a salient color singleton appearing at a specific location with high probability (66.67%) in Experiment 1. Experiment 2 included a condition where the high-probability location periodically changed. The authors analyzed reaction time (RT) data using Hierarchical Bayesian Inference to fit computational models corresponding to accumulator, RL, and categorical hypotheses. These models accounted for global RT decay and trial history to isolate the specific learning mechanism governing suppression. The results demonstrated that participants successfully suppressed attentional capture at high-probability locations, evidenced by reduced capture effects and slower target selection at those locations compared to low-probability locations. Crucially, computational modeling revealed that the data were best explained by a combination of a global exponential decay in RT and a categorical learning mechanism. In this model, the location with the highest association for distractor appearance was suppressed in an all-or-nothing fashion, regardless of subtle variations in trial history. This categorical representation was more parsimonious than continuous accumulator or RL models, which failed to capture the stability of the suppression effect. The findings indicate that the magnitude of suppression is sensitive to overall probability differences between locations rather than moment-to-moment trial histories. These findings suggest that statistical learning for distractor suppression relies on a stable, category-like representation that shields attentional priority from unexpected outcomes. This mechanism allows for robust inhibition of predictable distractors, persisting even when probabilities shift or during extinction phases. The study implies that the visual system prioritizes stable, rule-like settings of attentional priority over flexible, trial-by-trial updates, offering a parsimonious explanation for the inflexibility often observed in learned distractor suppression. This insight clarifies how statistical regularities are encoded to maintain efficient visual search in complex environments.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-25; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• visual search