Beyond Location Probability: Neural Correlates of Feature-Spatial Mapping Learning in Visual Search

Visual statistical learning enables humans to extract regular information from the environment, thereby optimizing the allocation of attentional resources. Previous studies have demonstrated that the visual system can learn the spatial probability of target occurrence and feature-specific spatial regularities for distractors. It remains unclear how target feature-location contingencies guide search when high-probability locations have equal overall target probability, or how such learning is reflected in ERP activity. The present study used a visual search paradigm that manipulated target-location probability and feature-spatial matching relationships, while simultaneously recording EEG data. Behavioral results showed that reaction times were significantly faster when targets appeared at feature-matched high-probability locations than when they appeared at feature-mismatched high-probability locations or low-probability locations. However, analyses by target feature showed that the feature-matching advantage was observed for color targets but not for shape targets. Electrophysiological results showed no significant differences in N2pc amplitude among the three conditions; when averaged across target features, LPC amplitudes were significantly smaller in both high-probability location conditions than in the low-probability location condition, but did not differ significantly between feature-matched and feature-mismatched high-probability locations when averaged across target features. These findings provide evidence that feature-spatial statistical regularities can influence visual search, while the feature-matching advantage was specific to color targets in the present study.