Do we mean the same? Semantic native-likeness in highly proficient second language users

Abstract

Research question:

Native-likeness (similarity between non-native and native users of the same language) depends on second language proficiency (L2p). However, evidence comes mainly from phonological or syntactic tasks, prompting an underexplored question: can higher L2p also entail a more native-like organization of semantic memory?

Methodology:

We asked high and low proficiency bilinguals to describe concepts in their second (English) and first (Spanish) language. We then compared their responses with semantic structure norms from native speakers of each language, and derived measures of semantic native-likeness through a validated semantic proximity measure.

Data and analysis:

We examined whether semantic native-likeness scores discriminated between high and low L2p levels at the group level (via inferential statistics) and at the individual level (via machine learning analyses).

Findings:

We found that semantic associations in English were more native-like in high than in low L2p bilinguals. Machine learning analyses showed that this pattern was highly consistent across individual subjects. Responses in Spanish did not differ between groups.

Originality:

This is the first study assessing semantic native-likeness through concept-feature associations in high and low L2p bilinguals, both at the group- and the individual-level.

Significance:

Our findings suggest that increased L2p entails a more native-like organization of semantic memory, informing models of bilingual cognition and its links with experience-related variables.