Across the reviewed literature, constructivism, experiential learning theory, and embodied cognition recur as key interpretive frameworks for understanding how immersive VR supports learning. These theories are commonly applied in VR learning research and offer complementary, testable expectations regarding learners’ attention, cognitive effort, and affective engagement—processes that can be examined using biometric measures (Radianti et al., 2020; Makransky and Petersen, 2021).

From a constructivist perspective, learning involves active exploration and selective attention to task-relevant information. In immersive VR neuroscience education, this suggests that learners engaged in knowledge construction should exhibit goal-directed gaze behaviors, such as sustained fixations on relevant anatomical structures and systematic visual exploration, which can be examined using eye-tracking metrics (Duchowski, 2018; Lai et al., 2013).

Experiential learning theory emphasizes iterative cycles of concrete experience, reflection, and conceptualization. Biometrically, these phases may be associated with distinct cognitive and affective signatures, including increased pupil dilation during demanding experiential activities or transient expressions of confusion followed by engagement during reflective insight, which can be explored through pupillometry and facial expression analysis (D’Mello et al., 2018; Souchet et al., 2022).

Embodied cognition highlights the role of bodily interaction in shaping cognition. In VR neuroscience education, embodied manipulation of three-dimensional neural models is expected to involve coordinated patterns of visual attention and action, measurable through eye tracking and complementary affective indicators (Makransky and Petersen, 2021; Jensen and Konradsen, 2018).

Together, these frameworks provide an integrated theoretical basis for linking biometric measures to learning-related processes in immersive VR environments and for informing mechanism-oriented instructional design.

VR-based neuroscience education encompasses a range of instructional applications that differ in learning objectives, interaction patterns, and cognitive–affective demands. In this review, VR applications are organized into three representative contexts: VR-based neuroanatomy learning, neurophysiology simulation, and clinical or diagnostic training. These contexts were selected because they reflect systematically distinct instructional goals and therefore provide analytically useful contrasts for interpreting biometric measures.

VR-based neuroanatomy learning primarily targets spatial understanding and structural knowledge acquisition (Aland et al., 2023). In this context, biometric measures can be used to examine how learners allocate visual attention across anatomical structures, the extent to which gaze is directed toward task-relevant features, and how visual exploration strategies evolve with increasing familiarity (Wainman et al., 2021). Unlike neuroanatomy learning, which focus on static structures, VR-based neurophysiology simulations emphasize dynamic process understanding and causal reasoning but rather static spatial relationships. Learners interact with simulations of neural signaling or circuit-level activity, often manipulating parameters to observe system behavior over time. Compared with neuroanatomy learning, this context places greater demands on temporal integration and cognitive effort, making pupillometry and affective measures particularly informative for examining fluctuations in cognitive load and emotional response during complex conceptual processing (Parong and Mayer, 2018). Beyond conceptual understanding, VR-based clinical and diagnostic training situates neuroscience knowledge within authentic decision-making contexts, often involving time pressure or uncertainty. This setting supports examination of visual search efficiency, attentional prioritization, and affective regulation, such as stress or confidence during task performance. Eye tracking can differentiate novice and expert gaze strategies, while affective indicators provide insight into emotional responses associated with high-stakes decisions (Shao et al., 2020).

By distinguishing VR applications according to learning goals and cognitive–affective demands, this framework enables context-sensitive interpretation of biometric data and strengthens their analytic value in immersive neuroscience education research.

Eye tracking has emerged as a powerful tool for investigating visual attention and cognitive processing in immersive learning environments, offering objective insight into how learners interact with VR-based instructional content (Duchowski, 2018). Fixation duration is commonly interpreted as an indicator of cognitive processing depth, with longer fixations reflecting more intensive information processing (Kiili et al., 2014). The “quiet eye” phenomenon—prolonged final fixations preceding action—has been associated with planning efficiency and expertise (Vickers et al., 2000). Saccadic paths and gaze heatmaps further reveal visual search strategies and collective attention patterns that inform interface design and learning efficiency (Duchowski, 2018).

In addition, Pupillometry offers a psychophysiological measure of cognitive effort that is difficult to consciously control (Souchet et al., 2022). Pupil dilation reliably reflects increased mental effort during complex tasks, making it a sensitive indicator of cognitive load across learning activities. When combined with fixation data, pupillometry provides a continuous, quantitative index of cognitive engagement that is less influenced by individual differences in visual scanning strategies. In immersive VR neuroscience education, these measures enable objective differentiation between on-task attention to relevant anatomical structures and off-task distraction, offering engagement metrics that extend beyond self-report methods (Just and Carpenter, 2013).

Combining traditional eye-tracking metrics with pupillometry provides a comprehensive framework for investigating learning mechanisms in immersive VR environments (Krejtz et al., 2018). This integrated approach can reveal how learners prioritize information and allocate cognitive resources, offering insight into how VR design influences attention, information encoding, and learning processes (Lee et al., 2023; Van Engen and McLaughlin, 2018).

However, eye-tracking measures do not constitute direct indicators of attention or learning. Fixation patterns, scan paths, and pupillary responses function as context-dependent signals whose interpretation depends on task structure, instructional design, and theoretical assumptions (Lai et al., 2013; Holmqvist and Andersson, 2017; Jarodzka et al., 2017). Gaze behavior reflects cognitive processing only when analytically situated within a specific learning context, instead of serving as a direct proxy for learning outcomes (Duchowski, 2018). Similarly, pupillometric responses index mental effort and processing demands but require careful experimental control to avoid conflation with learning gains (Just and Carpenter, 2013).

While cognitive processes are central to learning, affective factors—including emotions, motivation, and attitudes—play an equally critical role in educational outcomes (Wang et al., 2022). Facial expression recognition is a computer vision–based approach that objectively infers affective states by analyzing subtle facial muscle movements to estimate emotional valence and arousal (Li and Deng, 2020). This method captures both brief micro-expressions reflecting involuntary emotional reactions and sustained emotional trajectories that characterize learners’ ongoing experiences (Ekman and Friesen, 1978). Its educational value lies in revealing the dynamic interplay between emotion and cognition: positive states such as curiosity promote intrinsic motivation and deeper engagement, whereas prolonged frustration may hinder learning (Gruber et al., 2014). Linking affective patterns to learning outcomes is therefore essential for understanding learning as a holistic cognitive–emotional process (Adhanom et al., 2023).

In VR-integrated neuroscience education, the complexity of neuroanatomy and neural pathways can evoke strong emotional responses. Facial expression recognition enables researchers to identify affective transitions, such as shifts from confusion to curiosity during interaction with complex 3D brain models, or sustained frustration indicative of excessive cognitive load. This real-time affective data provides nuanced insight into learners’ experiences and complements cognitive measures such as eye tracking. However, applying facial expression analysis in VR presents technical challenges. Although tools such as iMotions and OpenFace have improved feasibility, head-mounted displays partially occlude facial features, requiring advanced algorithms for accurate inference (Tao et al., 2005). Accordingly, facial expression data should be interpreted as probabilistic and complementary. Despite these limitations, its integration is a critical step toward adaptive and responsive immersive learning environments.

The integration of eye tracking and facial expression recognition represents a leading approach within Multimodal Learning Analytics (MMLA), grounded in the view that learning arises from interacting cognitive and affective processes that cannot be fully captured by a single modality (Hernández-Mustieles et al., 2024). In VR-integrated neuroscience education, this multimodal approach holds promise for informing the development of adaptive learning systems, contingent on further empirical validation (Lai et al., 2013). In principle, such integration enables real-time adaptation of VR environments, for example, by dynamically adjusting the complexity of neuroanatomical models when sustained gaze fixation and affective signals indicate excessive cognitive load (Lécuyer et al., 2008).

However, substantial challenges remain. These include the technical difficulty of synchronizing heterogeneous data streams and the methodological complexity of fusing them into coherent analytical models (Chango et al., 2022; Qiu et al., 2023). Moreover, developing validated computational models that reliably infer learning states across diverse learners remains a critical barrier to scalable, biometric-driven immersive education (D’Mello et al., 2018).