Beyond immersion: distinguishing cognitive change from VR-supported performance

Introduction

Virtual reality (VR) is widely used across cognitive domains, including memory, attention, and executive functions (Cipresso et al., 2018; Riva et al., 2019), yet reported effects range from robust improvements to null findings. Differences in participant characteristics, intervention intensity, and task design may explain part of this inconsistency, but such explanations do not address how immersion itself is conceptualized and measured. This matters because the same immersive feature can provide task-relevant support in one intervention yet impose unnecessary demands in another. Clarifying when—and through what mechanisms—immersive features support cognition is therefore essential for designing effective interventions and explaining inconsistent outcomes.

Presence and agency are two experiential constructs central to immersive VR. Presence is the subjective sense of “being there” in a mediated environment (IJsselsteijn and Riva, 2003; Slater, 2003), whereas agency is the capacity for meaningful action with contingent feedback (Johnson-Glenberg, 2019). Together, they can strengthen task-relevant processing by embedding cognition within interactive, spatially structured contexts (Krokos et al., 2019; Makowski et al., 2017; Makransky and Petersen, 2021). Yet immersive features can also impose costs: navigation, spatial orientation, interface management, and discomfort may consume resources that would otherwise support the primary task (Kourtesis et al., 2019; Makransky and Petersen, 2021; Oh and Son, 2022). Cognitive benefit therefore cannot be inferred from engagement alone; it depends on whether task-relevant support outweighs task-irrelevant processing demands.

Such costs are not consistently measured in VR research, making it difficult to separate genuine cognitive change from performance supported by the immersive context (Kourtesis et al., 2019). Resolving this ambiguity requires linking specific immersive features to targeted cognitive processes and testing whether gains persist when environmental support is reduced or removed. This Opinion focuses on immersive VR as a cognitive intervention for nonclinical and subclinical populations; VR-based cognitive assessment, which uses VR for measurement rather than training, is outside our scope. The central question is therefore when immersion genuinely strengthens cognitive capacity rather than merely supporting task performance.

Within this scope, our argument is organized around three propositions. First, immersion is conditional: immersion supports performance only when its task-relevant contribution exceeds the extraneous load it adds (Parong and Mayer, 2018; Sweller, 1988). Second, immersion is directional: a feature’s effect depends on its alignment with the targeted process; spatial organization may support spatial encoding but divert resources when incidental to the mechanism being trained (Krokos et al., 2019; Makowski et al., 2017; Makransky and Petersen, 2021). Third, immersion can either scaffold or substitute for cognition, thereby shaping transfer: cues, prompts, and feedback support learning when they preserve task-relevant challenge but can produce nontransferable gains when they perform the learner’s cognitive work rather than strengthen internal control.

Together, these propositions constrain the gains immersive training can plausibly achieve. Near transfer (improvement on tasks resembling the trained task) is the most feasible target but remains susceptible to scaffolding effects. Ecological transfer (improvement in real-world functioning) is the most meaningful outcome because it tests whether gains matter beyond the training task. Far transfer (improvement on structurally dissimilar tasks) is less reliable and should not be assumed (Gobet and Sala, 2023). Finally, long-term maintenance is distinct from transfer and can be established only through delayed follow-up (Karbach and Verhaeghen, 2014).

These limits on transfer raise a broader evidentiary question: what must be demonstrated before improved performance in VR can be interpreted as cognitive change? We address this across five linked levels—technological properties, experiential constructs, cognitive mechanisms, moderators, and outcomes (Table 1). The proposed pathway is directional: design features shape experiences such as presence and agency as well as resource costs, which together influence whether the target process is exercised and whether any resulting gains transfer or persist. Moderators such as age, baseline ability, task–mechanism alignment, and cybersickness susceptibility may alter these links. This directional structure, however, is a framework to be tested, not an established causal chain. A more immersive system, stronger presence, or better trained-task performance alone therefore does not establish mechanism change or durable cognitive benefit.

TABLE 1. Levels of analysis and evidentiary requirements for evaluating immersive VR as a cognitive intervention tool.

The framework is directional, but a positive finding at one level does not by itself establish change at a downstream level. The third column specifies, for each level, the conclusion that cannot be drawn without additional evidence.

This evidentiary distinction clarifies what is needed when the Cognitive Affective Model of Immersive Learning (CAMIL; Makransky and Petersen, 2021) is applied to cognitive intervention. CAMIL explains how immersive affordances influence learning, but training performance can improve without durable learning or capacity change (Soderstrom and Bjork, 2015). For capacity training, this dissociation is critical: a feature that makes a task easier may raise performance while removing the cognitive work the intervention is meant to strengthen. We therefore extend CAMIL by making three intervention-specific requirements explicit: immersive features should align with the target mechanism, require users to exercise rather than offload that process, and produce benefits beyond VR-supported performance. Embodied cognition helps identify when sensorimotor engagement is mechanism-relevant (Wilson, 2002), whereas cognitive load theory helps identify when immersive demands are extraneous (Sweller, 1988). Together, these principles motivate a two-dimensional framework for evaluating whether mechanism-aligned engagement is achieved without excessive cognitive or physiological cost.

The presence–load decision matrix

The matrix operationalizes this framework along two continuous dimensions (Supplementary Image 1): immersive engagement and resource cost. Engagement reflects the joint contribution of presence and agency; presence without agency may produce absorption without task-relevant action, whereas agency without presence resembles a conventional interactive display. Resource cost includes extraneous cognitive load and physiological discomfort. Effort required to exercise the target process is not a cost but part of the intended training demand; navigation, interface management, or irrelevant environmental complexity becomes costly when it competes with that demand (Sweller, 1988). Cybersickness can further reduce available capacity and increase attrition (Oh and Son, 2022).

The four component constructs should be measured separately. Presence can be assessed with validated scales (Schubert et al., 2001; Witmer and Singer, 1998); agency can be indexed by task affordances for contingent, goal-relevant control and by self-report where appropriate; cognitive load by mental-effort ratings interpreted against task design (Paas, 1992); and discomfort by standardized cybersickness measures (Kennedy et al., 1993). Because effort ratings do not identify the source of load, target-process effort should be distinguished from navigation or interface burden using task-referenced or convergent measures. For the purposes of the matrix, these measures should remain continuous; “high” and “low” describe their position relative to a prespecified comparison, individual baseline, or validated scale anchor, not universal cutoffs. Engagement and cost may also covary, so both should be measured rather than inferred from device type.

The four quadrants (Supplementary Image 1) yield comparative, testable predictions, but their interpretation begins with mechanism alignment: immersive features must first plausibly engage the cognitive process the intervention is intended to strengthen. High engagement–low cost then provides the strongest case for benefit, with the best prospects for mechanism-relevant performance, transfer, and maintenance. High engagement–high cost may support immediate gains but is expected to yield weaker or less consistent transfer, with greater fatigue, discomfort, or attrition. Low engagement–low cost should add little beyond a matched nonimmersive condition, whereas low engagement–high cost is least favorable because it adds cognitive or physiological burden without task-relevant immersive support. The central prediction is therefore not that immersion helps, but that cognitive benefit depends jointly on mechanism alignment and the engagement–cost balance. Used prospectively, the matrix can guide both design and evaluation: the target mechanism is specified first, immersive features are selected to exercise that process, and their added engagement is weighed against the cognitive and physiological costs they introduce. These principles play out differently across cognitive domains. Memory and attentional–executive control illustrate how mechanism alignment, resource cost, and transfer depend on the process being trained.

Memory

Memory findings show why the comparison condition and the outcome measured matter. Krokos et al. (2019) found better recall with an immersive virtual memory palace than with a desktop version, whereas Boller et al. (2021) found, in older adults with subjective memory complaints, no added transfer benefit of immersive VR practice after method-of-loci training relative to an active VR control. Thus, the findings are not inconsistent: VR may outperform a less immersive format yet add little beyond an effective mnemonic strategy. What counts as benefit therefore depends on what VR is compared with and whether the outcome reflects trained-task performance, transfer, or maintenance.

Mechanistically, immersive VR is most likely to support memory when its spatial and sensorimotor features strengthen context-dependent encoding. A virtual memory palace may improve recall by linking items to distinct locations (Krokos et al., 2019), and presence may enhance encoding when those cues are salient and task-relevant (Makowski et al., 2017). The same spatial features can become costly, however, when navigation or environmental complexity is incidental to encoding and consumes working memory without supporting retention. Because such encoding is context-dependent, near transfer is most plausible; ecological transfer becomes plausible when training cues map onto real-world contexts, whereas far transfer and maintenance require direct evidence. The critical question is not whether VR improves recall in general, but whether its spatial and interactive features strengthen encoding without unnecessary processing costs.

Attentional and executive control

Goal-directed attentional control refers to sustaining attention to task-relevant stimuli while suppressing irrelevant information. It is closely linked to executive functions, including inhibition, working memory updating, and shifting (Miyake et al., 2000). Because these constructs share neural substrates, measurement approaches, and common variance (Friedman and Miyake, 2017; Niendam et al., 2012), we treat them as a unified attentional–executive control domain.

Attentional–executive findings are difficult to interpret uniformly because studies vary in comparison conditions, outcome measures, and environmental support. Immersive tasks can embed planning, inhibition, updating, and reorientation within naturalistic, multi-step scenarios. A virtual grocery task, for example, can require route planning, inhibition of distractions, goal updating, and reorientation as conditions change. Yet improved performance cannot be attributed to stronger internal control when engagement, expectancy, or contextual support offers an alternative explanation (Boot et al., 2013; Simons et al., 2016). Interpretation therefore hinges on whether immersive features require users to exercise attentional–executive control or instead make performance easier through environmental support.

Available evidence illustrates both the potential and the limits of immersion. Huang (2020) found greater gains in inhibition and task switching among adults aged 50 and older when the same exercise game (exergame) was delivered in an immersive rather than nonimmersive format, with presence mediating these effects. In a small exploratory sample of young adults, Grosprêtre et al. (2023) reported improvements in selective attention and inhibition after VR exergaming, whereas video exercise produced no such gains; other executive outcomes were unchanged. In contrast, Barreda-Ángeles et al. (2021) found that immersive 360° presentation increased presence and arousal but impaired focused attention and recall relative to screen presentation. Together, these findings suggest that immersion can support selected control processes when task demands engage them, but effects are process-specific and depend on the comparison condition; greater experiential intensity alone does not establish stronger attentional–executive control.

For attentional–executive training, the more informative test is whether improved regulation persists when environmental support is reduced or removed. Performance on untrained or ecologically relevant tasks is therefore more diagnostic than improvement within the trained VR scenario, and delayed follow-up is needed before durable change can be claimed. Matched nonimmersive controls, with engagement and expectancy measured, further help distinguish strengthened attentional–executive capacity from contextual support.

Toward standards in immersive cognitive research

Immersive VR holds genuine promise for cognitive intervention, but progress requires testing the directional links summarized in Table 1 rather than inferring benefit from presence or task performance alone. Future studies should manipulate specific immersive features using active comparisons matched for task content and expectancy where feasible; measure experiential engagement, resource costs, mechanism-specific cognition, and outcomes within the same study; preregister moderator hypotheses concerning age, baseline capacity, task alignment, and cybersickness susceptibility; and distinguish immediate performance from near, ecological, and far transfer as well as delayed maintenance, with transparent attrition reporting. These standards move the field from asking whether VR is broadly beneficial to testing when, for whom, and through which mechanisms immersion produces durable cognitive change. They also make negative or mixed findings more informative, because they help identify whether a weak effect reflects poor mechanism alignment, excessive cost, population-specific vulnerability, or inadequate transfer assessment rather than a generic failure of VR. On these terms, immersion is not a guarantee of benefit but a design parameter whose value must be demonstrated—mechanism by mechanism, population by population.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. The authors used ChatGPT to improve the language and readability of the manuscript. This included grammar editing, proofreading, and cross-referencing citations. The authors reviewed and edited the output as needed and take full responsibility for the content and integrity of the final publication.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Author disclaimer

The views and interpretations expressed in this article are solely those of the authors and do not necessarily reflect the official policies or positions of the Singapore government or any affiliated institutions or agencies.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/frvir.2026.1815501/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Jorge Martin Gutierez, University of La Laguna, Spain

Reviewed by: Putra A. K, State University of Malang, Indonesia

Shaoting Zeng, Beijing University of Technology, China