Speaker
Description
Recent studies suggest AI-generated images can enhance L2 vocabulary retention, but a key question remains: does the learner's act of generating an image — writing an English prompt that encodes a word's meaning — add value beyond merely viewing one? Prior findings are mixed. Drawing on dual coding theory, the generation effect, and levels-of-processing theory, this study proposes a within-subjects experiment to address that gap. Sixty-five first-year Vietnamese university students at B1 level, enrolled in an academic reading and writing course, will study 30 course-assigned academic words under three counterbalanced conditions rotated across learning sessions: learner-generated AI imagery, quality-matched provided imagery, and a word-only condition with sentence writing controlling for productive effort. Retention will be assessed through three-tier tests (productive recall, meaning recall, form recognition) immediately and after two and four weeks. Learner prompts will be coded for semantic elaboration to examine whether prompt quality predicts retention. Findings will offer evidence on when learner-driven image generation justifies classroom time, while providing empirical insight into B1 learners' emerging prompt literacy.
