Dialogue-based computer-assisted language learning systems for second language speaking development: A three-level meta-analysis

Abstract Speaking is often challenging for language learners to develop due to factors such as anxiety and limited practice opportunities. Dialogue-based computer-assisted language learning (CALL) systems have the potential to address these challenges. While there is evidence of their usefulness in second language (L2) learning, the effectiveness of these systems on speaking development remains unclear. The present meta-analysis attempts to provide a comprehensive overview of the effect of dialogue-based CALL in facilitating L2 speaking development. After an extensive literature search, we identified 16 studies encompassing 89 effect sizes. Through a three-level meta-analysis, we calculated the overall effect size and investigated the potential moderating effect of 13 variables spanning study context, study design and treatment, and measures. Results indicated a moderate overall effect size ( g = .61) of dialogue systems on L2 learners’ speaking development. Notably, three moderators were found to have significant effects: type of system , system meaning constraint , and system modality . No significant moderating effect was identified for education stage , L2 proficiency , learning location , corrective feedback , length of intervention , type of interaction , measure , and key assessment component . These findings suggest directions for future research, including the role of corrective feedback in dialogue-based CALL, the effectiveness of such systems across proficiency levels, and their potential in diverse learning contexts with the integration of generative artificial intelligence.