This sentence-final probe recognition study reports that advanced Korean learners showed antecedent reactivation effects in reading clause-internal, object-scrambled sentences in Japanese. This finding runs counter to Clahsen and Felser's (2006) shallow structure hypothesis. Chinese learners, on the other hand, demonstrated no evidence of antecedent reactivation effects due to first language effects. Data came from advanced Korean and Chinese learners of Japanese and native speakers of Japanese (20 participants per group). Drawing on the Parallel Architecture framework (Jackendoff, 2002; 2007) and Simpler Syntax apparatus (Culicover and Jackendoff, 2005), the paper proposes a unified, fine-grained account of Korean learners' native-like processing and Chinese learners' first language-influenced processing of scrambling.