[1]¿p#1 Interference from distractors can be reduced when they appear at locations or with features that occur frequently, suggesting that statistical learning enables the suppression of high-probability distractor locations and features. Recent studies have proposed that reduced interference in spatial statistic regularity arises not only from attentional suppression at high-probability positions but also from attentional capture at low-probability positions. However, it remains unclear whether feature-based statistical learning follows the same pattern—namely, attentional suppression of high-probability distractor features and attentional capture of low-probability distractor features. This study explored this possibility using both behavioral measures and event-related potentials (ERPs) by introducing an equal-probability baseline condition. The results showed that RTs to high-probability distractors were significantly shorter than low-probability distractors (Experiments1). Crucially, in Experiment 2, RTs to low-probability distractors showed no significant difference compared to equal-probability distractors, yet both were significantly longer than those for high-probability distractors. Consistent with these behavioral findings, ERP data from Experiment 2 further revealed reduced N2pc amplitudes for high-probability distractors relative to both low- and equal-probability distractors, with no amplitude difference between the latter two conditions. Furthermore, P3 amplitudes did not vary across conditions, suggesting that low-probability color singletons did not elicit a rarity effect. These findings indicate that the reduced interference associated with high-probability color distractors is primarily due to attentional suppression of the high-probability distractor feature itself, rather than attentional capture by low-probability distractor features. The implications of these findings for statistical learning and feature-based suppression mechanisms were discussed.