It remains an open question whether second language (L2) learners can process linguistic properties at the syntax–discourse interface. This study examines this issue in the context of the L2 processing of Korean dative sentences under different information structure requirements. Given that discourse constraints associated with information structure tend to manifest more strongly in noncanonical than in canonical structures, we tested whether L2 learners of Korean show sensitivity to such constraints during online processing. In a story-continuation task, both native and nonnative speaker groups showed a strong preference for producing canonical dative patterns, indicating their comparable knowledge of the canonical status of Korean dative sentences. In a self-paced reading task, both groups spent longer reading times when the word order of dative sentences did not follow given–new information structure, but only for the noncanonical and not the canonical structure. These results suggest that L2 processing of dative structures at the syntax–discourse interface relies on the same parsing architecture that guides native-speaker processing.