The Development of Native-Like Multiword Processing in a Second Language: A Proficiency Effect

A recent psycholinguistic framework, the Chunk-and-Pass model, posits that language acquisition involves learning how to appropriately chunk language input into larger representational units. Previously, Kosaka (2024) reported that low-proficiency second language (L2) learners may lack native-like chunking abilities at the multiword level; however, that study did not examine more advanced L2 learners. The present study extends this previous work by recruiting advanced L2 learners who share the same language background as in the original study—namely, Japanese. A Bayesian analysis revealed that while low-proficiency L2 learners did not demonstrate robust native-like chunking processes, advanced L2 learners exhibited processing tendencies similar to those of L1 speakers. Furthermore, both advanced L2 learners and L1 speakers were sensitive to the meanings of multiword units as cohesive wholes, whereas low-proficiency learners showed reduced sensitivity. These findings provide empirical support for the theoretical assumptions of the Chunk-and-Pass model.