Modelling the complex dynamic effect of <scp>L2</scp> speaking anxiety in metaverse on students' academic enjoyment, pronunciation anxiety, speaking performance, and task load

Abstract This study examines the intricate relationship between second language (L2) speaking anxiety and its subsequent impacts on pronunciation anxiety, perceived task load, speaking performance, and academic enjoyment within the context of metaverse‐mediated English language learning, amidst the changing landscape of immersive technologies. The study examines a crucial issue: how virtual environments, despite promoting genuine interaction, may unintentionally exacerbate emotional barriers to L2 oral proficiency. A total of 117 high school students, aged 14–17 and possessing B1–B2 proficiency levels, participated in a structured debate task on Spatial.io, a three‐dimensional metaverse platform that simulates a virtual conference hall featuring personalized avatars and spatial audio. Using a quantitative cross‐sectional design, data were collected through validated instruments and analysed using partial least squares structural equation modelling (PLS‐SEM) in SmartPLS 4, with bootstrapping (5000 resamples) employed to address non‐normal distributions. The findings suggest that increased L2 speaking anxiety is a significant predictor of heightened pronunciation anxiety and perceived task load, while also having substantial negative impacts on speaking performance and enjoyment of academic activities. These pathways highlight the metaverse's socio‐technical affordances as a double‐edged sword: they boost engagement through immersive embodiment while increasing anxiety due to real‐time evaluative pressures, ultimately compromising fluency and positive affect. As a result, the findings support the implementation of innovative interventions, including AI‐integrated scaffolding and anxiety‐reducing protocols (such as pre‐task mindfulness modules or adaptive avatar feedback), to enhance metaverse platforms for equitable L2 oral development.