Analyzing the relationship between argument structure construction (ASC) use and language learning has been an important area of investigation in second language (L2) studies from a usage-based constructionist approach. Previous studies have shown that advanced L2 learners’ language use demonstrates greater ASC diversity, less frequent ASC-verb combinations, and stronger ASC-verb associations. However, these investigations have been limited by methodological challenges in identifying ASCs and have predominantly focused on the written texts. To address these limitations, we employ a fine-tuned model to automatically extract ASCs from target and reference corpora, considering their semantic aspects. We then calculate ASC-based indices and, both alone and in combination with other lexicogrammatical indices, use them to predict L2 oral proficiency scores assigned by human judges. Our findings show that ASC-based indices alone explain 44% of the variance in scores. When combined with other indices, they provide complementary insights that enhance multivariate modeling of L2 oral proficiency.

Although lexical diversity is often used as a measure of productive proficiency (e.g., as an aspect of lexical complexity) in SLA studies involving oral tasks, relatively little research has been conducted to support the reliability and/or validity of these indices in spoken contexts. Furthermore, SLA researchers commonly use indices of lexical diversity such as Root TTR (Guiraud’s index) and D (vocd-D and HD-D) that have been preliminarily shown to lack reliability in spoken L2 contexts and/or have been consistently shown to lack reliability in written L2 contexts. In this study, we empirically evaluate lexical diversity indices with respect to two aspects of reliability (text-length independence and across-task stability) and one aspect of validity (relationship with proficiency scores). The results indicated that neither Root TTR nor D is reliable across different text lengths. However, support for the reliability and validity of optimized versions of MATTR and MTLD was found.