Abstract
This study investigates how individual differences in Working Memory Capacity (WMC) modulate the real-time syntactic processing and comprehension of complex sentences by intermediate-to-advanced second language (L2) learners of Japanese. Utilizing an eye-tracking-while-reading paradigm combined with an automated Japanese Reading Span Test, sixty-four native English-speaking learners of Japanese read sentences containing either subject-extracted relative clauses (SRCs) or object-extracted relative clauses (ORCs) under canonical and scrambled word orders. Eye-movement measures, including first-pass reading times, regression path durations, and total fixation times, were analyzed across critical disambiguating regions using linear mixed-effects models. The results revealed a significant processing disadvantage for ORCs over SRCs, particularly at the embedded verb and head noun regions, manifested in elevated regression path durations and higher regression rates. Crucially, higher WMC significantly attenuated processing disruption at the clause-final predicate, enabling more efficient structural reanalysis and leading to significantly higher offline comprehension accuracy. Conversely, lower-WMC learners exhibited persistent processing breakdown during complex integration points. These findings provide empirical support for capacity-constrained models of L2 sentence parsing, demonstrating that cognitive resource availability critically governs the real-time resolution of head-final syntactic dependencies in a non-native language.