Feature-Based Statistical Learning Modulates Distractor Interference, Not Distractor Rarity

Interference from distractors can be reduced when they appear with frequently occurring features, suggesting that statistical learning attenuates distractor interference. Previous research on spatial statistical learning has shown that reduced interference may reflect both suppression of high-probability locations and enhanced capture by low-probability locations. Whether feature-based statistical learning follows the same pattern remains unclear. We ran two experiments combining behavioral and EEG measures. Experiment 1 replicated earlier findings that high-probability distractors interfered less with target search than low-probability distractors. Experiment 2 introduced an equal-probability baseline and recorded ERPs. Behaviorally, response times were faster for high-probability distractors than for both equal- and low-probability distractors, which did not differ from each other. Neurally, distractor-evoked N2pc amplitudes were smaller for high-probability distractors than for equal- and low-probability distractors. Whereas, the P_D component was reliably observed but did not differ across conditions. Target-evoked N2pc amplitudes were likewise unaffected by distractor probability. Taken together, these findings indicate that feature-based statistical learning primarily reduces interference of high-probability distractors, rather than enhancing rarity-driven capture of low-probability distractors, and that these effects are specific to distractor rather than target processing.