Abstract This study investigated how second language (L2) learners process the Korean numeral quantifier construction by using transferable and nontransferable information. For Chinese-speaking learners of Korean (Chinese group), agreement between an honorific numeral quantifier and a noun in Korean constitutes transferable information in the canonical structure and nontransferable L2-specific information in the scrambled structure. For Japanese-speaking learners of Korean (Japanese group), this information gives rise to crosslinguistic conflicts in both structures. The results from a self-paced reading task showed that the Japanese group did not exhibit sensitivity to grammatical errors in both structures, whereas the Chinese group detected the agreement violation in the canonical but not in the scrambled structure. When a context sentence was provided to license scrambling in the test sentence, however, another group of Chinese-speaking learners of Korean showed sensitivity to the violation. These findings suggest varying degrees of crosslinguistic influence in L2 sentence processing.