Statistical context learning in visual search: distinct electrophysiological signatures of contextual guidance and context suppression.

Abstract

Facilitation of visual search by repeated distractor contexts is typically studied employing distractor configurations that are 100% predictive of the target location. Yet, real-world contexts vary in predictivity. We used electroencephalography (EEG) in human participants of either sex to explore how visual search facilitation arises from two distinct processing modes - contextual guidance and context suppression - that depend on the predictivity of distractor contexts, comparing repeated distractor arrangements that were either predictive or non-predictive of target location against a baseline of non-repeated arrangements. In Experiment 1, we manipulated context predictivity by shifting repeated contexts from predictive to non-predictive and vice versa, while in Experiment 2, we restricted repeated contexts to one side of the display to assess lateralized effects of the two processing modes. Both types of contexts behaviorally facilitated visual search, but facilitation was larger with predictive contexts. Making predictive contexts non-predictive reduced the facilitation while rendering non-predictive contexts predictive failed to produce gains. Half-display predictive contexts facilitated target detection on both sides, while non-predictive contexts facilitated same-side target detection only. EEG analyses revealed that predictive contexts triggered an early N1pc (guidance signal), followed by an enhanced N2pc (attentional selection) and an increased contralateral delay activity (CDA, indexing working memory processing of the target) in occipito-parietal regions, indicative of contextual guidance boosting the entire processing chain. By contrast, non-predictive contexts produced only an increased N2pc accompanied by reduced CDA, consistent with context suppression. These differential patterns demonstrate contextual guidance and context suppression to operate as electrophysiologically distinct processing modes.Significance Statement This study shows that the brain acquires two distinct processing strategies - contextual guidance and distractor suppression, respectively - depending on how reliably a learned visual context predicts the location of a searched-for target. Compared to non-repeated contexts, predictive contexts engendered enhanced responses in a sequence of lateralized event-related potentials, reflecting early signaling of the target location (N1pc), attentional engagement (N2pc), and processing of the target in working memory (CDA) - indicative of the brain operating in target-directed contextual-guidance mode. Non-predictive contexts, by contrast, only facilitated attentional engagement (increased N2pc), with scant (if any) representation of the context in working memory (reduced CDA) - indicative of the brain suppressing the distractor context to single out the target, so as to optimize search performance.