Ultimate attainment in L2 semantics: Where and why do learners fail to achieve native-like knowledge of verb meaning?

This study investigates the development of L2 lexical semantics in advanced Chinese learners of Japanese, focusing on the role of length of residence (LOR). Differences in semantic domains between learners' first language (L1) and second language (L2) often hinder accurate word meaning acquisition. Cross-linguistic differences between L1 and L2 are particularly evident in verb-noun collocations, which pose persistent challenges even for advanced learners. Sixty-six Chinese learners of Japanese with varied lengths of residency and 32 native Japanese speakers (NSs) participated in an acceptability judgment task involving the following three types of collocations: (1) L1-only collocations (e.g., *eakon-o akeru 'open the air conditioning'), which are valid in Chinese but not in Japanese; (2) L2-only collocations (e.g., eakon-o tsukeru 'turn on the air conditioning'), which are valid only in Japanese; and (3) L1-L2 collocations (e.g., doa-o akeru 'open the door'), which are valid in both languages. The results revealed that while NSs as a group rated L1-only collocations as unacceptable and L2-only collocations as acceptable, medium-LOR learners (3-7 years of residence) showed non-target-like patterns across the three collocation types. Long-LOR learners (more than 10 years) exhibited native-like judgments for L2-only collocations but continued to differ from NSs in their acceptance of L1-only collocations. Overall, these findings suggest that extended residence facilitates partial lexical-semantic integration in L2 achieved via positive evidence in the input, while persistent L1-mediated representations may constrain ultimate attainment in lexical-semantic development, which may persist due to a lack of negative evidence.