The study used VR equipment, standardized psychometric questionnaires, and multimodal physiological recording devices to establish an integrated assessment framework combining subjective and objective measures (Figure 1).

This study adopted a mixed methods design that combined subjective self-report measures with objective physiological signals to examine the effects of cultural heritage–based VR experiences on emotional health in older adults.

The VR environment was developed based on the Historic Centre of Macao, a United Nations Educational, Scientific and Cultural Organization World Heritage Site, selected due to its rich cultural significance and well-preserved architectural heritage. The inclusion criteria for the VR environment were (1) cultural and historical relevance, (2) visual richness and environmental complexity, and (3) the potential to evoke emotional and cognitive engagement. These criteria were informed by prior research suggesting that immersive and meaningful environments can enhance user engagement and modulate emotional and physiological responses in VR settings [36,37].

The experimental framework comprised 2 parallel VR conditions: a cultural heritage VR scene and a realistic replica VR scene. Both conditions were modeled on the same landmarks—the Ruins of St. Paul’s (A01) and St. Dominic’s Church (A02)—to ensure consistency in spatial content across conditions (Figure 2).

Participants were randomly assigned to either the cultural heritage VR group or the realistic replica VR group. Prior to the experiment, participants were blinded to their group allocation. The cultural heritage VR scenes incorporated creative enhancements, including visual aesthetic augmentation, soundscape design, light interactive elements, gamified tasks, and embedded cultural information. These features were intended to increase immersion, cultural engagement, and the elicitation of positive emotional responses among older participants. In contrast, the realistic replica VR scenes served as the control condition and were designed to provide a neutral, realistic replication reproduction of the same spaces. These scenes included no background music, no gamified interactions, and no cultural prompts. Visual presentation, auditory input, and interactivity were deliberately kept minimal to control for artistic and experiential factors. Representative snapshots of the virtual environments used in the study are presented in Figure 3.

The experimental setup comprised immersive VR hardware and physiological signal acquisition devices. VR experiences were delivered using a head-mounted display (Meta Quest 3/3S), which offers 4K-level resolution, a refresh rate of 90‐120 Hz, and 6 degrees of freedom tracking, ensuring stable and highly immersive visual performance. All VR content was developed by the research team led by PW at the School of Innovative Design, City University of Macau.

To ensure comparability across experimental conditions, the cultural heritage VR and realistic replica VR scenes were matched in spatial structure, viewpoint trajectories, and lighting parameters. The only differences between conditions concerned the presence or absence of artistic enhancements. Specifically, the cultural heritage VR condition incorporated environmental music, narrative voice prompts, cultural knowledge pop-ups, light interactive tasks, and gamified feedback, whereas these elements were deliberately excluded from the realistic replica condition.

Physiological data were collected using the EryLife S2 multimodal sensing device (Wecare Medical Co), which enables synchronized recording of EDA and HRV signals [38]. The sampling frequency for EDA was set at 4 Hz. Given that EDA signals primarily consist of slow-varying tonic and phasic components, relatively low sampling rates have been shown to be sufficient for capturing meaningful skin conductance responses [39]. Photoplethysmography signals were collected at 100 Hz for HRV analysis. This sampling rate is consistent with previous studies demonstrating that ≥100 Hz provides adequate temporal resolution for reliable interbeat interval detection and HRV estimation, in line with established HRV measurement principles [40]. EDA was measured using medical-grade dry electrodes attached to the inner side of the nondominant wrist, ensuring signal stability and wearing comfort for older participants. HRV data were obtained via a photoplethysmography sensor, providing complementary information on sympathetic and parasympathetic nervous system activity.

To systematically assess baseline emotional health prior to the VR intervention, psychological questionnaires were categorized into three functional domains:

This classification facilitated the distinction between short-term emotional changes (state variables) and relatively stable emotional characteristics (trait variables) in subsequent analyses and supported the examination of potential mechanisms underlying VR intervention effects across different psychological dimensions.

The 20-Item Positive and Negative Affect Schedule [41] was used to assess changes in positive affect (PA) and negative affect (NA) before and after the VR experience. The 16-Item Difficulties in Emotion Regulation Scale (DERS-16) [42] measured individual differences in emotion regulation difficulties, allowing exploration of how regulatory capacity may moderate VR-related emotional benefits. The 10-Item Connor-Davidson Resilience Scale (CD-RISC-10) [43] was administered to evaluate psychological resilience and its potential influence on emotional improvement. The 15-Item Geriatric Depression Scale (GDS-15) [44] served as a screening tool for depressive symptoms in older adults and was used to control for baseline emotional burden. The short version of the University of California, Los Angeles Loneliness Scale [45] assessed perceived social loneliness, enabling further examination of whether immersive VR experiences could alleviate loneliness-related emotional distress.

In addition, demographic variables (eg, living arrangements) and behavioral information (eg, prior exposure to VR) were collected via questionnaire.

All psychological scales used in this study are well-established instruments with demonstrated validity in prior research. To assess internal consistency within the current sample, reliability analyses were conducted by the research team using Cronbach α coefficients in SPSS (IBM Corp). The results indicated satisfactory internal consistency across all measures, with Cronbach α values exceeding the recommended threshold of 0.70.

Upon arrival at the laboratory, all participants received a standardized explanation of the study procedures and provided written informed consent. Participants then completed baseline psychological assessments, including PANAS, DERS-16, CD-RISC-10, GDS-15, and 8-Item University of California, Los Angeles Loneliness Scale (UCLA-8), to establish initial emotional states, emotional burden, and regulatory and resilience-related traits.

Following the questionnaire assessment, participants underwent a 2-minute resting-state EDA recording in a quiet environment with constant lighting conditions. This baseline recording was used to obtain individual skin conductance level (SCL) and SCR measures. After baseline data collection, participants were randomly assigned by computer algorithm to either the cultural heritage VR group or the realistic replica VR control group.

Participants then engaged in an immersive VR experience lasting approximately 8‐15 minutes using the Meta Quest 3/3S head-mounted display. During the VR session, EDA and heart rate activity were continuously recorded using the EryLife S2 Watch device. Immediately after the VR experience, participants completed the postintervention PANAS assessment and, when appropriate, participated in a brief semistructured interview to supplement subjective experience data (Figure 4).

EDA data were continuously collected during the VR experience using the EryLife S2 device and processed through a multistage standardized pipeline. First, the raw signals underwent automated artifact detection and low-pass filtering to remove motion-related noise and signal distortions caused by poor electrode contact. Second, baseline correction was performed using each participant’s resting-state SCL, ensuring comparability across individuals. Subsequently, key EDA features were extracted from the cleaned signals, including SCR frequency, peak amplitude, temporal characteristics, and area under the curve. Finally, z score normalization and percentage transformation were applied to optimize data distribution properties and to provide a robust basis for subsequent statistical analyses.

To address the RQs, a controlled experimental design was implemented. Specifically, RQ1 was examined by comparing emotional health outcomes between the cultural heritage VR condition and the realistic replication VR condition using self-report measures. RQ2 was investigated by analyzing physiological emotional responses based on EDA signals recorded during the VR experience. RQ3 was explored by examining the relationships between subjective emotional responses and physiological indicators, with the aim of identifying potential mechanisms underlying emotional health improvements associated with VR exposure.

To objectively compare autonomic nervous system responses under different VR conditions in older adults, between-group analyses were conducted using HRV and EDA indicators. For all measures, change scores (Δ=post–pre) were calculated at the individual level and used for statistical comparisons between groups. Continuous variables were expressed as mean (SD). Descriptive analyses were first performed to examine the distributional characteristics of the physiological data. Boxplots were used to visualize the median, IQR, and dispersion of each indicator. Subsequently, nonparametric Mann-Whitney U tests were applied to both HRV and EDA measures to assess between-group differences.

Descriptive analyses revealed clear differences in the distributional properties of EDA between the cultural heritage VR and realistic replica VR conditions. Examination of the boxplots (Figure 5) allowed further differentiation of the sources of these distributional differences.

Under the cultural heritage VR condition, the EDA box (representing the IQR, ie, the middle 50% of observations) was relatively compact, with the median located at a low level (approximately 25), indicating that baseline SCRs were clustered within a narrow range for most participants. However, the upper tail of the distribution was markedly extended and accompanied by several high-value outliers. This pattern indicates that a small subset of participants exhibited SCRs substantially higher than the group norm, thereby considerably increasing overall dispersion.

In the realistic replica VR condition, the EDA box was similarly concentrated in a low-value range, with a relatively low median, suggesting generally stable and homogeneous physiological responses across participants. Although a few high-value outliers were also observed, their number and influence on the overall distribution were limited, resulting in a lower degree of dispersion compared with the cultural heritage VR condition.

Overall, the primary differences between the 2 VR conditions lay in distributional dispersion and upper-tail structure rather than in a pronounced shift in central tendency. These findings suggest that cultural heritage VR experiences are more likely to elicit pronounced interindividual variability in autonomic responses, whereas realistic replica VR experiences are associated with more uniform and stable EDA response patterns.

The between-group comparisons of EDA indicators are presented in Table 2. No statistically significant differences were observed between the 2 VR conditions across all EDA measures, including mean EDA (P=.68; Cohen d=−0.16), SCL (P=.66; Cohen d=−0.16), SCR amplitude (P=.80; Cohen d=−0.09), and SCR peak rate (P=.87; Cohen d=0.06). All effect sizes were small (|d|<0.20), indicating limited practical differences in EDA responses between the 2 conditions.

Distributional analyses (Figure 6) further showed substantial overlap in medians and IQRs across groups, with no clear separation between conditions. Although several indicators exhibited occasional extreme values, these variations were primarily observed at the individual level and did not translate into consistent group-level effects.

From a physiological perspective, SCL reflects tonic sympathetic arousal, whereas SCR-related measures (amplitude and peak rate) capture phasic, event-related responses. These findings suggest that, under the present experimental conditions, EDA indicators did not demonstrate consistent group-level differentiation, potentially reflecting high interindividual variability and the complex nature of emotional arousal in immersive environments.

The between-group comparisons of HRV indicators are presented in Table 3. The results showed that a significant difference was observed for the mean HRV-derived value between the 2 VR conditions (P=.01; Cohen d=0.91), with higher change scores in the cultural heritage VR group (mean 21.48, SD 30.39) compared to the realistic replication VR group (mean −2.69, SD 21.03).

In contrast, conventional time-domain HRV indices, including the standard deviation of NN intervals (SDNN) and the root-mean-square of successive differences (RMSSD), did not show statistically significant differences between groups. Although SDNN exhibited a numerical difference (P=.06; Cohen d=−0.74), it did not reach statistical significance. Similarly, no significant difference was found for RMSSD (P=.48; Cohen d=−0.24).

Distributional analyses (Figure 7) indicated that the mean normal-to-normal (NN) interval (mean RR interval) tended to be higher under the cultural heritage VR condition, whereas substantial overlap was observed between groups for SDNN and RMSSD. Overall, the between-group difference was primarily reflected in the mean NN interval, while no consistent pattern was observed across standard time-domain HRV measures.

From a physiological perspective, RMSSD is generally considered to reflect parasympathetic activity, whereas SDNN represents overall HRV. However, as these conventional indices did not show significant differences, the physiological interpretation of the observed group difference should be approached with caution. The findings suggest that the effects of different VR conditions on autonomic regulation may be limited or influenced by substantial interindividual variability.

To ensure comparability between the cultural heritage VR group and the realistic replica VR group prior to intervention, chi-square tests were conducted on participants’ demographic and health-related background variables. A total of 12 indicators were examined, including sex, age group, educational attainment, marital status, living arrangement, occupational background, economic status, visual and auditory conditions, sleep quality, experience with smart devices, and prior exposure to VR or digital therapeutic applications (Table 4).

No statistically significant differences were observed between the 2 groups for any of the variables examined (χ² tests; all P>.05). These results indicate a high degree of baseline equivalence between groups in terms of sociodemographic structure, health status, and technology-related experience, thereby minimizing potential confounding effects attributable to sample composition and providing a reliable baseline for subsequent analyses of emotional and physiological outcomes.

Prior to the VR intervention, 2-tailed independent-samples t test were performed to compare baseline psychological trait measures between groups, including emotion regulation difficulties (DERS-16), psychological resilience (CD-RISC-10), and depressive symptoms (GDS-15). No significant between-group differences were detected for emotion regulation difficulties (t₄₇=−1.08; P=.29), psychological resilience (t_25.538=1.869; P=.07), or depressive symptoms (t_31.212=−0.688; P=.49; Table 5). These findings further confirm baseline homogeneity between the 2 groups at the psychological trait level, establishing an appropriate foundation for evaluating VR-induced emotional changes.

To examine the effects of VR exposure on emotional state, RM-ANOVA was conducted on PANAS PA and NA scores, with time (pre- vs postintervention) as the within-subject factor and group (cultural heritage VR vs realistic replica VR) as the between-subject factor.

Descriptive statistics are presented in Table 6. At baseline, PA and NA scores were comparable between the 2 groups. Following the intervention, PA increased and NA decreased in both groups, with more pronounced changes observed in the cultural heritage VR group.

For PA, RM-ANOVA revealed a significant main effect of time (F_1,47=211.20; P<.001; η²=0.818), indicating an overall increase following VR exposure. A significant main effect of group (F_1,47=5.67; P=.02; η²=0.108) and a significant time×group interaction (F_1,47=12.47; P=.001; η²=0.210) were also observed (Table 7).

For NA, a similar pattern was observed. RM-ANOVA showed a significant main effect of time (F_1,47=118.71; P<.001; η²=0.716), along with significant main effects of group (F_1,47=14.15; P<.001; η²=0.231) and a significant time× group interaction (F_1,47=16.97; P<.001; η²=0.265; Table 7).

Simple effects analyses (Tables 8 and 9) revealed a consistent pattern of significant time effects and differential group effects over time. At baseline, no significant between-group differences were observed in either PA or NA. Following the intervention, PA increased significantly in both the cultural heritage VR experience group and the realistic replication VR experience group; however, the increase was significantly greater in the cultural heritage VR group, resulting in higher postintervention PA compared with the control condition. Conversely, NA decreased significantly in both groups across time, with a significantly larger reduction observed in the cultural heritage VR group, which also exhibited significantly lower postintervention NA relative to the realistic replication VR group.

To further explore individual differences in emotional responses to VR, Pearson correlation analyses were conducted to examine associations among emotion regulation difficulties, psychological resilience, depressive symptoms, and emotional state before and after the intervention (Table 10).

Emotion regulation difficulties were significantly negatively correlated with psychological resilience (r=−0.530; P<.01) and positively correlated with depressive symptoms (r=0.526; P<.01). In addition, higher emotion regulation difficulties were associated with lower PA and higher NA both before and after the VR experience, indicating greater emotional burden among individuals with poorer regulation capacity.

Psychological resilience showed the opposite pattern, exhibiting significant positive correlations with PA and significant negative correlations with NA and depressive symptoms. These findings suggest that resilience may serve a buffering and facilitative role in VR-based emotional interventions.

Comparative analysis of subjective emotional changes and EDA responses revealed consistent patterns across modalities. Relative to the realistic replica VR condition, participants exposed to cultural heritage VR exhibited higher and more sustained levels of skin conductance during the experience, alongside larger increases in PA and more pronounced reductions in NA on self-report measures.

This convergence between subjective emotional outcomes and autonomic physiological responses indicates coherent emotion-related effects across measurement modalities under different VR content conditions, providing an empirical basis for subsequent multimodal interpretations of VR-induced emotional regulation effects.