Behavioral Engagement in VR-Based Sign Language Learning: Visual Attention as a Predictor of Performance and Temporal Dynamics

A new study from researchers Davide Traini, José Manuel Alcalde-Llergo, Mariana Buenestado-Fernández, and Domenico Ursino analyzes how behavioral engagement predicts learning outcomes in SONAR, a virtual reality application for sign language training. The team focused on three automatically tracked engagement indicators: Visual Attention (VA), Video Replay Frequency (VRF), and Post-Playback Viewing Time (PPVT). Using correlation analysis and binomial Generalized Linear Model (GLM) regression, they found that VA and PPVT were significant predictors of performance on a subsequent validation quiz, jointly explaining a substantial proportion of variance. Notably, VRF showed no meaningful association with learning success.

Going beyond simple outcomes, the researchers conducted a temporal analysis by aggregating moment-to-moment VA traces from all learners. This revealed distinct engagement dynamics: an initial acclimatization phase, oscillatory attention cycles during learning, and pronounced attentional peaks during the assessment quiz. Critically, these attention peaks were directly aligned with the most informationally dense segments of the training and validation videos. The findings underscore that sustained and strategically allocated visual attention is central to effective learning in immersive VR environments, and they highlight the value of passive, behavioral trace data for building predictive models of user engagement and performance.