Statistical Learning of Feature-Spatial Mappings and Its Modulation by Cue-Induced Expectation

Visual statistical learning enables individuals to extract regularities from the environment, thereby optimizing attentional allocation. Previous research has mainly focused on unidimensional location or feature probability learning, leaving unclear whether individuals can acquire multidimensional mappings between target features and high-probability spatial locations, and how cue-induced expectations modulate the neural expression of such mappings. We employed a location probability paradigm combined with multivariate pattern analysis (MVPA) and recorded EEG. Reaction times showed a significant gradient from the feature-matched high-probability location to the low-probability location, indicating successful acquisition of feature-spatial mappings under both cue conditions. The N2pc did not reliably differentiate the two high-probability location conditions, providing no detectable evidence of feature-dependent modulation of early lateralized selection. Predictive cues elicited brief pretarget alpha lateralization toward the cue-associated HPTL. Search-array-locked MVPA discriminated feature-matched from feature-mismatched HPTL trials under both cue conditions, with stronger decoding under predictive cueing. These findings show that learned feature-spatial regularities were expressed behaviorally and in distributed neural activity, and that predictive cues modulated their neural expression before and after search-array onset.