ChatGPT as a facilitator of vocabulary learning strategy use for lexical bundles across different difficulty levels

While many studies have investigated vocabulary learning strategies (VLSs) for single words, little research has examined effective VLSs for multi-word units such as lexical bundles (LBs), which are crucial for English for academic purposes (EAP). This study explored which VLSs are effective for Japanese university students to learn LBs at their current (i: B1) or slightly lower (i–1: A2) proficiency levels on the Common European Framework for Reference (CEFR). Participants received instruction in social (questioning), cognitive (gap-filling practice), and memory (functional grouping and L1 translations) strategies and studied 40 LBs (20 for A2 and B1). They also took A2- and B1- level LB tests. Their VLS uses were mediated and tracked by an AI tool ChatGPT, which engaged participants in interactive out-of-class learning. The results revealed that ChatGPT-assisted social strategies were beneficial across both A2 and B1 levels, while cognitive strategies were useful at A2 level but lost its sole effectiveness at more difficult (B1) LBs. As there was no significant score difference between both tests, deeper processing associated with VLS uses, rather than difficulty of LBs themselves, could be a determining factor for learning LBs. Theoretical and practical applications of the results are discussed.