Extensive research demonstrates the context-dependent nature of human memory. Information encoded alongside contextual cues (environmental or internal) benefits from the later presence of those cues during recall. The absence of these cues hinders recall. This context-dependence is especially impactful when learning multiple similar information sets in close proximity, as interference between sets can impair memory. Learning each set in a distinctive context mitigates interference by providing unique cues bound to each set. However, this creates context-dependent associations, potentially hindering recall when those cues are absent. Mental reinstatement—vividly imagining the original encoding environment—offers a solution, proving almost as effective as physically returning to the learning context. This study aims to harness context-dependence to enhance learning and retention using VR, which offers highly distinctive and controlled immersive learning environments. The focus is foreign vocabulary learning, a practically valuable and paradigmatic paired associate learning task. English-speaking participants learned Swahili and Chinyanja words, two phonetically similar languages, to maximize interference. Participants were assigned to either a single-context or dual-context group (learning both languages in one VR context vs. each language in its own unique VR context). The study hypothesized that the dual-context group would demonstrate reduced interference and better long-term retention, contingent on their subjective sense of presence (experiencing the VR as a real environment). A follow-up fMRI experiment investigated the neural mechanisms of context-specific reinstatement during retrieval.

Prior research extensively documented context-dependent memory, showing that learning and recall are enhanced when the context at encoding matches the context at retrieval. Godden and Baddeley's (1975) classic scuba diving experiment, and subsequent studies using less dramatic environmental changes (different rooms, noise levels), illustrated this effect, particularly prominent in recall tests. When learning multiple similar information sets, context-dependent learning can reduce interference and improve recall by associating unique cues with each set. However, reliance on contextual cues can hinder transfer to new contexts. Mental reinstatement, the act of vividly imagining the original context, mitigates this issue. Previous studies successfully used background images or videos to create distinct learning contexts, but these lack the immersion and ecological validity of real-world environments. Virtual Reality (VR) offers a powerful solution, creating immersive, distinctive, and controllable environments for studying context-dependent memory. A recent study successfully replicated context-dependent recall using two distinctive VR contexts. However, measuring participants' sense of presence (believability of the VR environment) is crucial, as it moderates the effect of VR-based contexts on memory outcomes.

The study employed a behavioral experiment and a follow-up fMRI experiment. The behavioral experiment involved 48 participants (24 in each context group: single-context, dual-context). Participants learned 80 foreign words (Swahili and Chinyanja) over two days, using two custom-built first-person desktop VR environments. Each context featured nine named areas. The learning process involved four rounds: one initial learning round followed by three test-learn cycles, incorporating expanding retrieval practice (increasing intervals between learning and testing). Learning occurred along predetermined paths within each VR context. To measure transfer and mental reinstatement, a non-VR test (T4) was administered on Day 2, using an explicitly cued mental reinstatement protocol (congruent or incongruent with the learning context). A surprise telephone test (T5) assessed one-week retention. A 10-item presence scale measured participants' sense of presence in the VR environments. The fMRI experiment involved 22 participants (all in the dual-context group), using the same learning and recall tasks. fMRI data were collected during T4 to assess neural correlates of context-specific reinstatement during retrieval using multi-voxel pattern analysis (MVPA) and representational similarity analysis (RSA). The fMRI experiment was designed to objectively measure context reinstatement during word recall, avoiding reliance on subjective reports or assumptions about the persistence of the mental reinstatement state into the word recall period. The use of verbal material allowed separation of the sensory modalities (visuospatial for contexts, verbal/auditory for memoranda), disentangling neural correlates of contextual support from memory retrieval itself.

Across groups, participants recalled 42% of words after two exposures (T2), exceeding expectations based on previous studies. No group differences emerged during initial learning (T1-T3). During T4 (non-VR test), congruent mental reinstatement significantly enhanced recall (52% vs 47%). Dual-context participants showed 38% fewer intrusions than single-context participants, indicating reduced interference. Importantly, the dual-context advantage in one-week retention (T5) was only observed among participants reporting high presence (92% vs 76%). The fMRI experiment revealed a strong relationship between representational fidelity (measured by RSA) and successful recall. Trials with high representational fidelity showed a 5% higher recall rate than low-fidelity trials, both at T4 and T5. The interaction between reinstatement prompt and representational fidelity was significant, primarily driven by T5 data: after incongruent mental reinstatement, high-fidelity reinstatement resulted in a 10.1% recall advantage compared to low-fidelity reinstatement. This advantage was absent in congruent reinstatement trials.

The study successfully harnessed context-dependence to enhance learning and retention of challenging material. Distinctive VR environments, combined with mental reinstatement, facilitated transfer of knowledge to non-VR settings. The high one-week retention rate (up to 92%) highlights the effectiveness of the approach. Three crucial conditions emerged: high sense of presence in the VR environment, unique contexts for each language, and assessment after a long delay. The fMRI results objectively quantified contextual support through mental reinstatement, demonstrating a direct link between neural context reinstatement and successful recall. The results expand on prior work showing that context-specific fMRI activity patterns can facilitate verbal recall. The finding that the recall advantage was most prominent in incongruent reinstatement trials at T5 suggests that overcoming incongruency through high-fidelity reinstatement may be a "desirable difficulty." The study's findings support the crucial role of spatial context in event representations, aligning with existing literature highlighting the primacy of spatial context in episodic memory. The active navigation of the VR environments during learning likely contributed to the observed contextual effects, contrasting with previous studies using passively presented VR backgrounds. The high initial learning rate (42%) may also be attributed to the engaging nature of the VR environments, potentially increasing learner motivation and engagement. This is consistent with the Cognitive Affective Theory of Immersive Learning (CAMIL).

This study demonstrates the successful use of context-dependence to enhance learning of challenging material, mitigating the negative effects of context-dependence via mental reinstatement. The approach resulted in high one-week retention, contingent on three conditions: high presence, unique contextual support, and a long retention interval. fMRI data provided an objective measure of mental context reinstatement, demonstrating its impact on recall. Future research should investigate the effect of more immersive VR systems, examine the role of specific brain regions in context reinstatement, explore the impact of context relevance, and quantify aspects such as learner engagement and motivation. The findings bear wide-ranging implications for education, training, and therapeutic interventions.

The study cued mental reinstatement, precluding an assessment of performance without explicit cues. The fMRI analysis focused on measuring context reactivation, limiting conclusions about specific brain structures' roles. Using a desktop VR system, rather than HMDs, potentially reduced participants’ sense of presence. The arbitrary relationship between contexts and learning material may have limited the memorability and transfer of information. Future research should address these limitations, using HMDs for improved immersion, exploring different reinstatement methods, and employing contexts relevant to the material.