Recent empirical results have linked the N400 ERP with predictive language comprehension processes based on statistical learning (SL). However, links between SL abilities and N400 on the level of individual differences have so far been underexplored. The present study tested SL performance in 29 participants using speech segmentation and artificial grammar learning tasks, followed by EEG recordings of their N400 responses to sentences varying in cloze probability (high, intermediate, low). Mixed-effects models revealed that better online SL performance (SL-ON) was associated with larger N400 amplitudes across conditions. Additionally, working memory showed a significant main effect and interacted with SL-ON in modulating N400 amplitude, while cloze probability also had a robust, independent effect on it. These results demonstrate that individual differences in SL abilities contribute to N400 response variability, supporting the view that the semantic operations reflected by the N400 may involve some form of statistical learning as well. Our findings also raise the possibility that SL and CP tap into distinct levels of predictive mechanisms in language comprehension.
