Virtual Reality vs. Augmented Reality: A Comparative Meta-Analysis of Immersive Learning Impact on STEM Education

Abstract

This meta-analysis investigated the comparative effectiveness of Augmented Reality (AR) and Virtual Reality (VR) in enhancing learning outcomes in STEM education. A total of 80 empirical studies published between 2015 and 2025 were analyzed using the PRISMA 2020 framework, encompassing 6,538 participants from 27 countries across primary, secondary, and tertiary levels. Effect sizes were computed as Hedges’ g under the DerSimonian–Laird random-effects model using R programming (Google Colab), Comprehensive Meta-Analysis (CMA v4), and JASP 0.18.3 for cross-validation. The findings revealed large and statistically significant pooled effects for both technologies (AR: g = 0.82 [0.64, 1.00]; VR: g = 1.07 [0.85, 1.29]). Moderate heterogeneity (AR I² = 61%; VR I² = 68%) justified the random-effects approach, while publication bias tests indicated symmetrical funnel plots and non-significant Egger’s regression results, confirming the stability of estimates. Subgroup analyses showed that effect sizes increased with higher educational levels and longer intervention durations, and varied across outcome domains—VR yielding the highest effects in psychomotor and spatial skills, while AR excelled in affective engagement. The results align with the Cognitive-Affective Model of Immersive Learning (CAMIL), affirming that immersive technologies facilitate dual cognitive and emotional pathways to learning. The study concludes that AR and VR are transformative pedagogical tools that significantly improve conceptual understanding, engagement, and skill mastery in STEM, positioning immersive learning as a cornerstone of Education 4.0 innovation.