Drivers of self-efficacy and learning in virtual reality training: The roles of presence, cognitive load, interest, and motivation

Virtual reality training (VRT) is increasingly adopted in education and industry, yet the mechanisms underpinning its effectiveness remain insufficiently understood. This study investigates the learning process involved in VRT by validating and extending the Cognitive Affective Model of Immersive Learning (CAMIL), a recent integrative theory of learning in immersive virtual reality, and its empirical refinement (CAMIL-E). Two aims were addressed: (1) to compare the effectiveness of VRT ( N = 131) and traditional training (TT; N = 147) on learning and self-efficacy; and (2) to use structural equation modelling (SEM; N = 131) to examine how presence, cognitive load, motivation, and interest shape learning and self-efficacy within VRT while accounting for baseline differences. VRT produced significantly greater knowledge gains than TT, while self-efficacy improved comparably across both training methods. SEM results showed that presence predicted situational interest (β = 0.18, p = 0.004), while the subsequent path from situational interest to learning was positive but non-significant (β = 0.14, p = 0.072), providing partial support for CAMIL-E. Post-training self-efficacy was primarily determined by baseline levels (β = 0.74, p < 0.001), rather than by presence or other cognitive and affective factors. Psychological readiness played a meaningful role, as baseline self-efficacy and intrinsic motivation predicted the degree of presence experienced during VRT. Extraneous cognitive load emerged as a consistent barrier, negatively predicting presence, situational interest, intrinsic motivation, self-efficacy, and learning. These findings extend CAMIL by highlighting the central roles of extraneous cognitive load and pre-training psychological readiness in shaping immersive learning outcomes.