Cognitive overload in multimedia language learning: a review of attention limits and instructional design strategies

Abstract

Background Cognitive overload and speaking anxiety are major barriers to second language acquisition, particularly in English as a Foreign Language (EFL) and English as a Second Language (ESL) contexts. Multimedia learning, artificial intelligence (AI), mobile learning (m-learning), and cognitive-affective interventions have been proposed to enhance language performance and reduce anxiety; however, their overall effectiveness remains unclear. Objective To evaluate the effects of multimedia-assisted language learning, cognitive-affective interventions, learner-related factors, and speaking anxiety across educational settings and to quantify their effectiveness through meta-analysis. Methods This meta-analysis followed systematic evidence-synthesis procedures. A total of 38 studies published between 2013 and 2026 were included. Risk of bias was assessed using the ROBINS-I tool, and evidence quality was evaluated using GRADE criteria. Random-effects inverse-variance models were applied to calculate standardized mean differences (SMDs) with 95% confidence intervals (CIs). Heterogeneity was assessed using the I2 statistic. Publication bias was examined through funnel plots and Egger’s regression test, while robustness was evaluated using leave-one-out sensitivity analyses. Results Significant positive effects were observed across all four domains (p < 0.05). Multimedia and technology-enhanced language learning demonstrated a pooled SMD of 0.44 (95% CI: 0.31–0.58; I2 = 45.6%). Anxiety-reduction and cognitive-affective interventions showed an SMD of 0.51 (95% CI: 0.40–0.62; I2 = 13.5%). Studies examining anxiety, academic achievement, and learner-related factors yielded an SMD of 0.51 (95% CI: 0.41–0.62; I2 = 27.5%). The strongest effect was observed for interventions targeting speaking anxiety in educational settings (SMD = 0.54, 95% CI: 0.47–0.62; I2 = 16.7%). ROBINS-I assessments indicated predominantly moderate risk of bias, whereas GRADE ratings ranged from high certainty for multimedia-based interventions to low-to-moderate certainty for context-specific approaches. Egger’s test detected no significant publication bias (p > 0.05). Sensitivity analyses confirmed the stability of pooled estimates, with effect sizes ranging from 0.52 to 0.54. Conclusion Multimedia-assisted learning, AI-enhanced educational environments, and cognitive-affective interventions significantly improve language learning outcomes and reduce speaking anxiety. Speaking anxiety remains a key determinant of language achievement, while psychological factors such as self-efficacy and confidence play crucial roles in successful language acquisition.