Walking settings affect aesthetic emotion during art appreciation in virtual reality

1 Introduction

Understanding how individuals engage with visual art has long been a central concern in empirical aesthetics. Previous research has typically examined art viewing in two distinct settings: physical museum environments and controlled laboratory settings. Laboratory studies offer rigorous experimental control (e.g., Leder et al., 2006) but often lack the contextual richness and immersive qualities of physical exhibition spaces. Conversely, museum-based studies provide high ecological validity by capturing natural viewing behaviors and psychological responses (e.g., Carbon, 2020; Pelowski et al., 2024), but they offer limited control over stimulus presentation and potential confounding variables. Bridging ecological validity and experimental rigor remains a fundamental methodological challenge in the study of aesthetic experience.

Virtual reality (VR) offers a powerful synthesis of these approaches by combining immersive realism with precise experimental manipulation (e.g., Cotter et al., 2022; Lin et al., 2020; Gotthardt et al., 2025). Recent advances in head-mounted displays (HMDs) and interactive systems have increased interest in VR-based art appreciation. Compared with two-dimensional displays, VR environments can enhance immersion and interactivity, potentially intensifying emotional and cognitive engagement (Lee and Youn, 2025; Lin and Kang, 2024; Fang, 2025). However, VR also introduces unique perceptual challenges, including constraints on operability, limited bodily freedom, and potential discrepancies in color, texture, and scale perception relative to physical artworks (Lin et al., 2020; Matsumoto and Okada, 2025). Importantly, VR enables systematic manipulation of spatial variables, such as artwork size and viewing distance, that are difficult to isolate in physical galleries. Thus, VR does not merely replicate physical art viewing; it also creates new experimental opportunities to decouple and test specific components of the aesthetic encounter.

Locomotion is a particularly distinctive dimension of the museum experience. In physical museums, viewers are not passive observers; they actively regulate their distance from artworks, approaching or receding according to size, detail, and attentional focus (Carbon, 2020; Yang and Kawabata, 2024). This embodied engagement is a foundational component of aesthetic experience and may modulate both affective and evaluative responses. Yet many current virtual museum systems rely on controller-based “teleportation,” which lacks the sensorimotor contingencies of natural walking. It remains unclear whether immersive VR can reproduce, or even enhance, the emotional and experiential benefits of physical locomotion. Furthermore, because VR allows direct manipulation of spatial scale, it raises important questions about whether display size systematically affects aesthetic evaluation (the “size effect”) and whether physical walking modulates these effects (the “walk effect”).

To address these questions, the present study used the interactive VR art appreciation system “Midokoro Walk®,” developed by Dai Nippon Printing Co., Ltd. (Isoda, 2022). This system was designed to digitally reproduce seventeenth-century ceiling paintings from the Mazarin Gallery at the Bibliothèque nationale de France, allowing users to wear an HMD while physically walking along a constrained 4 m handrail. By integrating calibrated spatial scaling and non-visual sensory cues, the system aims to evoke a sense of architectural vastness within a limited physical footprint. Such calibration is essential because perceived scale in VR can vary substantially across individuals due to factors such as interpupillary distance and perceptual biases (Kim and Interrante, 2017; Mine et al., 2020). In our experiment, artwork size was systematically manipulated while participants either walked or remained stationary during viewing.

Beyond subjective measures of emotion and viewing quality, we included an exploratory assessment of eye movements. Eye tracking provides objective indices of attentional allocation and visual exploration, which are intrinsically linked to aesthetic processing. Although previous research has compared gaze behavior between physical and VR exhibitions (Gulhan et al., 2021; 2023), few studies have examined gaze patterns as a function of artwork size and locomotion in immersive environments. Because walking dynamically alters both viewing distance and perspective, it may influence the attentional mechanisms underlying art appreciation. Incorporating these measures, therefore, allows a more comprehensive understanding of how embodied interaction shapes aesthetic experience. Building on findings that museum-based viewing leads to more positive evaluations than classroom-based viewing (Ishiguro et al., 2021), this study investigates whether VR can replicate these emotional benefits through controlled, embodied interaction.

Accordingly, we addressed three primary research questions: (1) Does immersive VR art viewing enhance emotional states and perceived viewing quality? (2) Does manipulating artwork size influence affective and evaluative responses (size effect)? (3) Does physical walking within VR modulate aesthetic, emotional, and attentional outcomes (walk effect)? We hypothesized that (H1) viewing artworks in immersive VR would increase positive emotions and decrease negative emotions from pre-to post-viewing, and that (H2) improvements in perceived viewing quality would mediate these changes. We further predicted that (H3) larger display sizes would enhance aesthetic evaluations and emotional responses relative to smaller sizes, and that (H4) physical walking during VR viewing would enhance immersion and positive emotional responses while positively influencing aesthetic evaluations relative to stationary conditions. To address these questions, we deliberately implemented a virtual museum environment focused on observing external artwork. Specifically, we aimed to evaluate the ecological validity and embodied mechanisms of replicating physical museum viewing behaviors. By systematically combining subjective ratings, behavioral tracking, and eye-tracking metrics, this study clarifies the multifaceted impacts of spatial scale and embodied mobility in virtual art appreciation.

2 Materials and methods

2.1 Participants

We expected a moderate effect size for the main effect of walking condition or size condition and conducted an a priori power analysis using G*Power (Faul et al., 2007) to determine the sample size required to estimate η _*p* ² = 0.06. The analysis indicated that 23 participants would be sufficient to achieve 80% power at α = 0.05. Drawing on our experience with online experiments, we increased the sample size to 25 participants to ensure an adequate number of participants with an error rate below 20%. Seventy-one university students (49 females; M = 21.68 years, SD = 1.55) participated in the experiment and received 2,500 Japanese Yen as compensation. All participants had normal or corrected-to-normal vision. Participants were randomly assigned to one of three viewing conditions: no-walking (n = 25), forward-backward walking (n = 25), or lateral (right-left) walking (n = 21). The study was approved by the Research Ethics Committee of Keio University, and all participants provided written informed consent before participation.

2.2 VR settings and materials

The study used the Midokoro-Walk® system (Dai Nippon Printing Co., Ltd.), which enables users to view a painting while physically moving within an environment that seamlessly integrates real-world and virtual spaces. The standalone software controlling the VR settings ran on a Meta Quest Pro headset connected to the participant’s headset (1832 × 1920 px per eye, 72 Hz target refresh rate), which included an eye-tracking function with an average accuracy of 1.652° and a precision of 0.699° (Wei et al., 2023). This device supports six-degrees-of-freedom (6DoF) tracking with high-resolution optics and a high refresh rate, effectively minimizing display latency and motion jitter. For eye tracking, only gaze movements directed toward the paintings were recorded, and fixation points were stored as two-dimensional coordinates for each stimulus. Participants wore the headset and navigated a virtual space while holding a handrail and using a controller in their other hand. The experiment included three movement conditions: forward-backward walking, lateral walking, and teleportation. For the walking conditions, we installed a handrail (approximately 4 m long and 0.8 m high) in the laboratory. Participants assigned to the walking conditions used the physical handrail alongside the corresponding virtual handrail in the VR environment. Although the virtual gallery space measured 30 m × 30 m, participants could physically move only within a 4 m × 4 m area located 1 m from the displayed painting. The system provides 1:1 spatial tracking aligned with the participant’s physical locomotion within a designated 4 m × 4 m area, ensuring consistent congruence between visual motion and physical bodily sensations. Therefore, Participants’ positions in the virtual gallery were precisely matched to their physical positions within this observation area. During locomotion in the gallery space, participants’ spatial positions in the VR environment were recorded as two-dimensional coordinates from a top-down perspective for each trial.

In the VR space, each painting was displayed on the wall at one of three sizes. One of 12 Western paintings served as the visual stimulus in each trial. Three display sizes were prepared for each painting within the VR space: 2 m, 6 m, and 12 m in width (referred to as sizes S, M, and L, respectively). The vertical height of each painting was adjusted to preserve its original aspect ratio. The size condition was manipulated within participants, such that each participant viewed four paintings in each of the three size conditions, for a total of 12 paintings, with each painting viewed once. Combinations of paintings and sizes were counterbalanced across participants.

Three movement conditions were prepared for viewing paintings in VR: (1) no-walking, (2) forward-backward walking, and (3) lateral (right-left) walking. In the no-walking condition, nine viewing spots were arranged in a 3 × 3 grid in front of each painting within a 4 m × 4 m gallery space. Participants used the controller to “teleport” between these spots and viewed the paintings without physical movement. In the forward-backward walking condition, three virtual handrails (each 4 m long) were placed perpendicular to the painting (Figure 1). Participants physically walked along the real handrails installed in the laboratory. The virtual and real handrails were precisely aligned, and participants could teleport laterally between the three rails in VR using the controller. In the lateral (right-left) walking condition, three 4 m handrails were placed parallel to the painting. As in the forward-backward condition, participants held the handrail while moving side to side to view the paintings. In addition, participants could teleport between the three handrails at different front-back positions using the controller, allowing them to adjust their viewing distance from the painting within the VR space.

2.3 Procedure

In each condition, participants completed a sequence consisting of (1) the pre-questionnaire phase, (2) the art appreciation in VR phase, and (3) the post-questionnaire phase.

• The pre-questionnaire phase: To assess changes in positive and negative moods before and after the art appreciation phase, we used the generic version of the Positive and Negative Wellbeing Umbrella in the UCL Museum Wellbeing Measures Toolkit (Thomson and Chatterjee, 2014), which consists of a positive (six items) and negative (six items) mood scale in which participants rated six mood items for each mood using five-point Likert scales (Positive: Active, Alert, Enthusiastic, Excited, Happy, and Inspired; Negative: Distressed, Irritable, Nervous, Scared, Unhappy, and Upset). The mood index was developed as a less time-consuming measure based on the Positive Affect Negative Affect Scale (Watson et al., 1988), a well-established and reliable measure of psychological mood that is sensitive enough to detect significant changes between pre- and post-measurements (Thomson et al., 2020).
• The art appreciation in VR phase: During each painting-viewing session, participants were instructed to freely verbalize any thoughts or feelings they had about the artwork or their own viewing experience. For example, they could describe what they saw in the painting or share any emotions or thoughts that came to mind. No prompts or questions were provided. Participants could speak freely for up to five min for each painting and could signal the experimenter when they had finished. They were also informed that remaining silent was acceptable if they had nothing to say. After viewing each painting, participants provided verbal responses to nine items designed to assess their subjective viewing experience and emotional reactions. The items were as follows: (a) “How long do you think you spent viewing the painting? Please answer in min and s” (subjective viewing time). (b) “To what extent did you focus on fine details in the painting?” (c) “To what extent did you attend to the painting as a whole?” (d) “To what extent did you want to view the painting from various positions?” (e) “To what extent did you wish to look carefully without moving?” (f) “To what extent was the viewing experience overwhelming?” (g) “To what extent did you feel immersed in the painting?” (h) “To what extent did you feel fear during the viewing experience?” (i) “To what extent did you enjoy the viewing experience?” The item assessing subjective viewing time was always asked first. The remaining eight items were presented in randomized order for each participant to control for potential order effects. Responses were recorded using a seven-point Likert scale ranging from 1 (“not at all”) to 7 (“very strongly”). This viewing and evaluation procedure was repeated for all 12 paintings.
• The post-questionnaire phase: To examine changes from before to after the viewing experience, participants completed the generic version of the Positive and Negative Wellbeing Umbrella in the UCL Museum Wellbeing Measures Toolkit again. They also completed the 12-item Generic Wellbeing Questionnaire in the UCL Museum Wellbeing Measures Toolkit (Thomson and Chatterjee, 2014), using a five-point Likert scale ranging from 1 (“None of the time”) to 5 (“All of the time”). The questionnaire, referred to here as the “museum experience,” included items such as “I felt happy”, “I felt engaged, “I felt safe and secure”, etc. In addition, participants’ general interest in art was assessed using the Vienna Art Interest and Knowledge Questionnaire (VAIAK; Specker et al., 2020), which includes 11 items rated on a seven-point scale; only the Art Interest items were analyzed.

2.4 Data analysis

2.4.1 Participants and data exclusion

Behavioral and gaze data from 17 of the 71 participants were excluded from the final analysis because of wireless network instability and hardware-related issues that caused data corruption. Consequently, the final analytic sample consisted of 54 participants (Mean age = 21.51, SD = 1.56; 35 females). Among these participants, 18 were assigned to the no-walking condition, 18 to the forward–backward walking condition, and 18 to the lateral (right–left) walking condition.

2.4.2 Gaze behavior processing

Gaze behavior was recorded using the Meta Quest Pro’s built-in eye-tracking system, which provided (x, y) coordinates over time. To ensure algorithmic consistency with the iMap4 protocol (Lao et al., 2017), fixations were identified using the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm, with a velocity-threshold filter applied. Fixations occurring outside the stimulus boundaries were excluded from the analysis. Duration-weighted fixation maps were generated and then subjected to individual Z-score normalization for each participant to account for individual differences in viewing duration and fixation counts. Bilinear interpolation was used during rendering to ensure smooth spatial transitions in fixation intensity. For statistical validation, a linear mixed-effects model (LMM) was fitted to the pixel-level intensities for representative stimuli, using Viewing Condition as a fixed effect and Participant as a random effect. However, as the painting size was counterbalanced across participants, which resulted in insufficient data points for specific size-stimulus cells, the evaluation of size effects on gaze was limited to a qualitative comparison of aggregated heatmaps.

2.4.3 Locomotion and spatial behavior

Locomotion behavior was quantified using five metrics: (1) total movement distance (overall distance in VR, including joystick-based movement); (2) jump count (number of discrete joystick operations); (3) walking distance (distance traveled via physical walking only); (4) movement variance (spatial variability of locomotion); and (5) actual viewing time. Each index was analyzed using LMMs with Painting Size (S, M, L) as a within-subjects factor and Viewing Condition as a between-subjects factor. To investigate the spatial distribution of participants within the virtual gallery, dwelling times were calculated by dividing the 4 m × 4 m movement area into a 3 × 3 grid. These dwelling time values were analyzed using LMMs with the same fixed and random effects structure described above.

2.4.4 Psychological and emotional measures

Changes in positive and negative moods were analyzed using two-way mixed-design ANOVAs with Period (pre-vs. post-viewing) and Viewing Condition as factors. Subjective painting evaluations were analyzed using LMMs with Painting Size and Viewing Condition as fixed effects, and Participant, Painting Stimulus, and Presentation Order as random effects. Stepwise multiple regression analyses were performed to identify which of the eight evaluation items predicted emotional outcomes and museum experience scores (12-item Generic Wellbeing). The significant predictors were then used in mediation analyses, in which art interest (VAIAK scores) was included as a background covariate. The analyses used maximum likelihood estimation with bias-corrected bootstrap confidence intervals (5,000 resamples). For all LMMs, degrees of freedom were estimated using the Kenward–Roger method, and Bonferroni corrections were applied to all multiple comparisons with an alpha level of 0.05.

3 Results

3.1 Changes in mood before and after art appreciation and evaluation of the appreciation experience

To examine the emotional impact of the VR art experience, changes in positive and negative mood were analyzed using a 2 (Period: before vs. after) × 3 (Viewing Condition: no-walking, forward–backward walking, lateral walking) mixed-design ANOVA (Figure 2). For positive mood, the results revealed a significant main effect of period [F (1, 68) = 51.98, p < 0.001, η _*p* ^*2* = 0.433], indicating a substantial increase in positive affect after the VR experience. However, there was no significant main effect of viewing condition [F (2, 68) = 0.335, p = 0.716, η _*p* ^*2* = 0.010] and no significant interaction between period and condition [F (2, 68) = 2.059, p =0 .135, η _*p* ^*2* = 0.057]. Similarly, for negative mood, a significant main effect of period was observed [F (1, 68) = 10.599, p < 0.01, η _*p* ^*2* = 0.135], indicating a decrease in negative affect after the experience. No significant effects were found for viewing condition [F (2, 68) = 0.860, p = 0.428, η _*p* ^*2* = 0.025] or the interaction [F (2, 68) = 0.068, p = 0.934, η _*p* ^*2* = 0.002]. Together, these findings indicate that the VR art experience significantly enhanced positive emotions and reduced negative emotions, regardless of the locomotion condition.

To evaluate viewing-experience quality, mean scores from the 12-item Generic Wellbeing Questionnaire were compared across conditions (no-walking: Mean = 43.80 (SD = 4.88); forward–backward walking: Mean = 46.36 (SD = 6.31); lateral walking: Mean = 43.62 (SD = 5.89). A one-way ANOVA revealed no significant difference in overall experience quality across viewing conditions [F (2, 68) = 1.74, p = 0.18, η _*p* ^*2* = 0.05]. Given the consistent emotional improvement across conditions, we conducted mediation analyses to explore whether the subjective quality of the experience accounted for the observed mood changes, incorporating art interest (VAIAK scores) as a covariate to control for individual differences.

For positive mood (Figure 3A), direct effects from pre-to post-appreciation were significant across all three viewing conditions; however, no significant indirect effects through experience quality were observed. Experience evaluations significantly predicted post-viewing positive mood in the lateral walking condition, but not in the no-walking or forward–backward conditions. Furthermore, art interest significantly influenced art experience scores as a background covariate, suggesting that personal interest in art may modulate the quality of the experience when physical locomotion is restricted. For negative mood (Figure 3B), direct effects between pre- and post-measurements were significant in the no-walking and forward–backward conditions, but not in the lateral walking condition. While no indirect effects were found, experience quality scores influenced post-viewing negative mood specifically in the no-walking condition.

We further examined the correlations between the assessed indices within each viewing condition (Table 1). Across all locomotion conditions, significant positive correlations were observed between pre- and post-appreciation positive mood scores. For negative mood, pre- and post-appreciation scores were significantly correlated in the no-walking and forward–backward walking conditions, whereas this stability was not observed in the lateral walking condition. Notably, post-appreciation positive mood was significantly correlated with the museum experience score in both the forward–backward and lateral walking conditions, but not in the no-walking condition. By contrast, in the no-walking condition, art interest was significantly and positively correlated with both post-appreciation positive mood and the quality of the museum experience; these correlations did not reach significance in either of the two walking conditions.

TABLE 1. Correlations among pre- and post-viewing mood scores, museum experience quality, and art interest across the three viewing conditions.

*p < 0.05, p < 0.01, *p < 0.001.

3.2 Individual painting evaluations: effects of artwork size and locomotion

To examine the effects of viewing condition and painting size on the subjective ratings of individual paintings, linear mixed-effects models (LMMs) were fitted for each questionnaire item. Viewing condition and painting size were entered as fixed effects, while participant, painting stimulus, and presentation order were included as random effects. Significant main effects and interactions for several items are summarized in Table 2. No significant effects were identified for subjective viewing time (Item 1) or fear (Item 8) (all p > 0.20).

TABLE 2. Effects of viewing condition, painting size, interaction, and post hoc tests.

Viewing condition = no-walking, forward-backward walking, lateral walking. Painting size = S, M, L. p-values were adjusted using Bonferroni correction. Degrees of freedom were estimated using the Kenward-Roger method. Bold indicates significant effects (p < .05).

3.2.1 Size effects on visual exploration

A significant main effect of painting size was found for attention to fine details (Item 2) [F (2, 756.85) = 8.56, p < 0.01]. Large paintings received significantly lower ratings than both medium and small paintings [t (758) = −2.72, p = 0.02, d = −0.23; t (751) = −4.12, p < 0.01, d = −0.35]. Similarly, for attention to the painting as a whole (Item 3), a significant main effect of size emerged [F (2, 763.90) = 4.50, p = 0.02], with large paintings being rated lower than small paintings [t (763) = −2.84, p = 0.01, d = −0.24]. Regarding the desire to look carefully without moving (Item 5), a significant main effect of size was also observed [F (2, 753.76) = 12.75, p < 0.01], with ratings decreasing progressively as painting size increased (S > M > L) [S vs. M: t (762) = −2.50, p = 0.04, d = −0.21; S vs. L: t (756) = −2.54, p = 0.03, d = −0.22]. These results indicate that larger scales in VR may paradoxically reduce the subjective sense of detailed scrutiny and holistic attention.

3.2.2 Locomotion and embodied interaction

For the desire to view from various positions (Item 4), both a significant main effect of size [F (2, 763.19) = 25.56, p < 0.01] and a significant interaction between size and condition [F (4, 761.04) = 3.07, p = 0.02] were observed. Overall, small paintings elicited a significantly lower desire for movement than medium and large paintings [t (761) = −6.02, p < 0.01, d = −0.52; t (765) = −6.32, p < 0.01, d = −0.54]. Follow-up analyses of the interaction revealed that for medium and large paintings, participants in the forward–backward walking condition expressed a stronger desire to change viewing positions compared with those in the no-walking condition [t (107) = 2.66, p = 0.03, d = 0.57; t (107) = 2.48, p = 0.04, d = 0.53].

For feeling overwhelmed (Item 6), significant main effects were found for painting size [F (2, 763.72) = 39.94, p < 0.01] and viewing condition [F (2, 68.00) = 5.37, p = 0.01]. Large paintings elicited the strongest feelings of being overwhelmed (S < M < L) [S vs. M: t (764) = −2.70, p = 0.02, d = −0.23; M vs. L: t (764) = −6.05, p < 0.01, d = −0.52]. Participants in the forward–backward walking condition reported significantly stronger feelings of being overwhelmed than participants in the no-walking condition [t (68) = 3.27, p < 0.01, d = 0.70].

Furthermore, significant main effects of viewing condition were observed for immersion (Item 7) [F (2, 68.00) = 3.54, p = 0.03] and enjoyment (Item 9) [F (2, 68.00) = 3.33, p = 0.04]. In both cases, the forward–backward walking condition produced significantly higher ratings compared with the no-walking condition [Immersion: t (62.33) = 2.52, p = 0.04, d = 0.48; Enjoyment: t (68) = 2.57, p = 0.04, d = 0.50]. Overall, these findings suggest that both immersive locomotion and large-scale presentation enhance critical aspects of the viewing experience. Specifically, forward–backward walking reliably increased feelings of being overwhelmed, immersion, and enjoyment, suggesting that embodied movement significantly amplifies the affective quality of aesthetic experiences in VR.

3.3 Behavioral measures

3.3.1 Locomotion metrics: Distance, jumps, variability, and viewing time

To examine whether viewing conditions and painting size influenced behavioral performance in the VR environment, linear mixed-effects models (LMMs) were fitted for four key behavioral indices: (1) total movement distance (including joystick-based teleportation); (2) jump count (the frequency of discrete joystick operations); (3) movement variability (the spatial dispersion of locomotion); and (4) actual viewing time. For all models, viewing condition and painting size were entered as fixed effects, whereas participant ID, painting stimulus, and presentation order were included as random effects.

• Total movement distance: A significant main effect of viewing condition was observed [F (2, 51.11) = 4.30, p = 0.02, η _*p* ² = 0.14]. Participants in the no-walking condition moved significantly greater total distances than those in the forward–backward walking condition [t (51) = 2.82, p = 0.02, d = 1.10]. A significant main effect of painting size also emerged [F (2, 580.55) = 17.29, p < 0.01], with small paintings eliciting shorter movement distances than both medium [t (558) = −3.95, p < 0.01, d = −0.38] and large paintings [t (576) = −5.07, p < 0.01, d = −0.55; Figure 4].
• Jump count: Significant main effects were found for both viewing condition [F (2, 51.16) = 8.69, p < 0.01] and painting size [F (2, 591.18) = 15.68, p < 0.01]. Post-hoc comparisons revealed that jump count was significantly higher in the no-walking condition than in the forward–backward walking condition [t (51) = 4.15, p = 0.02, d = 1.51]. Viewing small paintings also resulted in fewer joystick operations than viewing medium [t (559) = −3.44, p < 0.01, d = −0.33] and large paintings [t (576) = −5.51, p < 0.01, d = −0.53].
• Movement variability: For spatial dispersion of the viewer’s position, a robust main effect of painting size was observed [F (2, 591.09) = 170.11, p < 0.01]. Spatial dispersion increased significantly with artwork scale: large paintings produced greater variability than medium paintings [t(572) = 5.29, p < 0.01, d = 0.51], which in turn produced greater variability than small paintings [t (576) = 12.56, p < 0.01, d = 1.21]. Furthermore, a significant interaction between viewing condition and size was observed [F (4, 591.09) = 8.27, p < 0.01]. Simple effects tests indicated that for large paintings, participants moved more variably in both the forward–backward [t (57.10) = 2.51, p = 0.04, d = 8.12] and lateral walking conditions [t (57.00) = −2.95, p = 0.01, d = −9.51] compared with the no-walking condition. In the forward–backward walking condition, variability was significantly greater for large than for small paintings [t (572) = 3.30, p < 0.01, d = 3.08], and in the lateral walking condition [t (573) = 3.25, p < 0.01, d = 3.03], variability was significantly greater for medium than for small paintings.
• Actual viewing time: The analysis of actual viewing time showed a marginal main effect of painting size [F (2, 578.73) = 2.97, p = 0.05; Figure 5], and a marginal interaction between viewing condition and size [F (4, 574.49) = 2.04, p = 0.09]. Pairwise comparisons showed that medium paintings elicited longer viewing times than small paintings [t (578) = 2.42, p < 0.05, d = 0.24]. Within the lateral walking condition, viewing time for medium paintings was significantly longer than for small paintings [t (578) = 2.42, p < 0.05, d = 0.41] and marginally longer than for large paintings [t (578) = 2.10, p = 0.10, d = 0.35].

3.3.2 Spatial distribution of viewing locations across viewing conditions

To examine spatial viewing behavior in the VR environment, participant positions were sampled at 10 Hz. Spatial distributions were visualized separately for each condition: violin plots were used for the forward–backward and lateral walking conditions, and a 3 × 3 grid was used to visualize viewing positions within the 4 m × 4 m movement area for the no-walking condition.

Forward–backward walking condition: The virtual gallery space was divided into three equal vertical segments relative to the painting: Left, Center, and Right. The distance from the painting at each time point was extracted for each segment and painting size. To statistically evaluate differences in viewing position, an LMM was applied with Area (Left, Center, Right) and Painting Size (Small, Medium, Large) as fixed effects. Subject, Painting Stimulus, and Trial Order nested within Subject were treated as random effects. Because continuous walking trajectories produced repeated observations nested within trials, a nested random-effects structure (1 | Subject: Trial) was included to accurately account for within-subject trial-level variability and reduce inflated Type-I error rates. The analysis revealed a significant main effect of Painting Size [F (2, 201.44) = 324.14, p < 0.01] and a significant Area × Painting Size interaction [F (2, 310.39) = 2.47, p < 0.01; Figure 6A]. Post-hoc comparisons indicated that viewing distance was significantly shorter for small paintings compared to medium paintings [t (210) = −15.23, p < 0.01, d = −2.10], and shorter for medium compared to large paintings [t (174) = −10.55, p < 0.01, d = −1.60]. Furthermore, for medium-sized paintings, viewing distances in the Center area were significantly greater than in both the Left [t (299) = 2.57, p = 0.03, d = 0.30] and Right areas [t (295) = 2.41, p < 0.05, d = 0.28].

(2) Lateral walking condition: In the lateral walking condition, the 4 m × 4 m movement area was divided horizontally into three equal segments: Near, Middle, and Far. Violin plots visualized the distribution of left–right viewing positions for each size (Figure 6B). An LMM with Area (Near, Middle, Far) and Painting Size as fixed effects revealed a significant main effect of Painting Size [F (2, 187.80) = 4.19, p = 0.02] and a significant Area × Painting Size interaction [F (2, 383.22) = 2.47, p = 0.04]. Simple main effects testing showed that the mean X-coordinate was significantly smaller (more centered) for small paintings than for large paintings [t (198) = −2.74, p = 0.02, d = −0.39]. Follow-up comparisons indicated that this size effect was present only in the Near area, where small paintings produced significantly smaller X-coordinates than large paintings t (440) = −4.02, p < 0.01, d = −0.38).

(3) No-walking condition: In the no-walking condition, the VR viewing area was divided into a 3 × 3 matrix comprising three horizontal zones (Left, Center, Right) and three vertical zones (Near, Middle, Far), resulting in nine discrete viewing regions. Dwelling times within each region were aggregated across participants and visualized as heatmaps for each painting-size condition (Figure 7). An LMM analysis identified significant main effects for both Left–Right Area [F (2, 1,154.62) = 117.02, p < 0.01] and Near–Far Area [F (2, 1,150.54) = 3.97, p = 0.02]. Significant two-way interactions were observed for Left–Right Area × Size [F (4, 1,149.54) = 6.55, p < 0.01], Near–Far Area × Size [F (4, 1,150.52) = 32.61, p < 0.01], and Left–Right Area × Near–Far Area [F (4, 1,149.23) = 4.28, p < 0.01]. Importantly, the three-way interaction between all factors was also significant [F (8, 1,149.40) = 25.69, p < 0.01]. Post-hoc comparisons for the main effects revealed that dwelling time was significantly longer in the Center than in both the Left [t (1,135) = 11.06, p < 0.01, d = 0.66] and Right areas [t (1,128) = 12.77, p < 0.01, d = 0.76]. Furthermore, Far areas exhibited significantly longer dwelling times than Middle areas [t (1,145) = 2.81, p = 0.01, d = 0.17]. Detailed analyses of the interaction by painting size revealed distinct spatial preferences. In the large-painting condition, participants spent significantly more time in the Far area compared to both the Near area [t (1,078) = 9.73, p < 0.01, d = 0.59] and the Middle area [t (1,145) = 10.16, p < 0.01, d = 0.64]. In the Medium-painting condition, dwelling time was significantly shorter in the Near area compared to the Middle area [t (1,019) = −3.93, p < 0.01, d = −0.25] and the Far area [t (1,057) = −3.99, p < 0.01, d = −0.25]. By contrast, in the small-painting condition, dwelling time was significantly longer in the Near area than in the Middle [t (1,114) = 5.87, p < 0.01, d = 0.35] and Far areas [t (1,148) = 4.47, p < 0.01, d = 0.26].

3.3.3 Integrated analysis of spatial engagement across conditions

To compare spatial engagement patterns across all viewing conditions, we employed a unified 3 × 3 matrix covering the 4 m × 4 m movement area. This approach enabled direct statistical comparison of mean dwelling times (Figure 7) using two separate LMMs, one focused on the horizontal (Left–Right) and the other on the vertical (Near–Far) spatial dimensions.

Horizontal Distribution (Left–Right): An LMM with Viewing Condition, Painting Size, and Left–Right Area (Left, Center, Right) as fixed effects revealed a significant main effect of Area [F (2, 2,918.35) = 191.90, p < 0.01] and a significant Left–Right Area × Size interaction [F (4, 2,882.36) = 21.99, p < 0.01]. Dwelling time in the Center region was significantly longer than in both the Right [t (2,914) = 15.69, p < 0.01, d = 0.87] and Left regions [t (2,915) = 14.94, p < 0.01, d = 0.84]. Participants spent significantly more time in the peripheral regions (Left and Right) for large paintings than for small paintings [Right: d = 0.36; Left: d = 0.43]. Conversely, in the Center region, dwelling time decreased as painting size increased, being significantly shorter for large paintings compared to both medium [d = −0.27] and small paintings [d = −0.62; Figure 7].

Vertical Distribution (Near–Far): The analysis of the vertical dimension revealed significant main effects for Area [F (2, 2,903.29) = 6.28, p < 0.01] and Size [F (2, 2,894.34) = 6.87, p < 0.01]. We identified a significant three-way interaction among Near–Far Area, Painting Size, and Viewing Condition [F (8, 2,879.91) = 3.58, p < 0.01], indicating that locomotion conditions uniquely modulated how participants adjusted their distance relative to artwork scale. Analysis of the Near–Far × Size interaction revealed systematic adjustments: for large paintings, dwelling time was longest in the Far area [d = 1.13 vs. Near], while for small paintings, it was longest in the Near area [d = 0.99 vs. Center]. The Near–Far × Viewing Condition interaction showed that in the no-walking condition, participants exhibited a significant bias toward the Far area [d = 0.28 vs. Center], whereas in the lateral walking condition, dwelling time was significantly shorter in the Center area compared to both the Near [d = −0.29] and Far areas [d = −0.30], reflecting more extensive side-to-side exploration (Figure 7).

3.4 Gaze analysis

To investigate the attentional dynamics underlying VR art appreciation, gaze data were processed using a pipeline developed for spatial and algorithmic consistency with the iMap4 protocol (Lao et al., 2017). Fixations were identified using the Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm, combined with a velocity-threshold filter to exclude saccades and noise. Only fixations within the paintings’ boundaries were included in the analysis. To account for individual differences in viewing behavior and total fixation counts, duration-weighted fixation maps were generated for each participant and subjected to Z-score normalization. Bilinear interpolation was employed during rendering to ensure smooth spatial transitions in fixation intensity, mirroring the Gaussian smoothing and mapping protocols used in high-fidelity eye-tracking research.

The resulting heatmaps for each of the 12 stimuli revealed highly consistent visual exploration patterns across the three viewing conditions (no-walking, forward–backward walking, and lateral walking) and the three artwork sizes (Small, Medium, and Large). As shown in the representative heatmaps (Figure 8), participants’ gaze consistently converged on semantic focal points, such as human figures, faces, and central compositional elements, regardless of whether physical locomotion was required or presentation scale. This spatial stability suggests that while the affective experience of art is significantly modulated by embodied interaction and scale, the fundamental attentional allocation to the artwork’s features remains robustly anchored to the visual content in the VR environment.

To quantitatively verify whether locomotion conditions influenced the spatial distribution of attention, a linear mixed-effects model (LMM) was fitted to the pixel-level fixation intensities for each stimulus. The model included Viewing Condition as a fixed effect and Participant as a random effect. Consistent with the qualitative observations, the analysis revealed no significant main effect of Viewing Condition for the representative stimuli examined. For instance, in Stimulus 1, there were no significant differences in pixel intensity between the no-walking and forward–backward walking conditions [t (51) = −0.84, p = 0.40], or between the lateral and forward–backward walking conditions [t (51) = −0.20, p = 0.84]. These results suggest that the density and spatial allocation of visual attention directed toward the paintings did not vary significantly as a function of the viewer’s physical movement or the navigation method employed in the virtual gallery. This dissociation between stable gaze behavior and varying subjective evaluations highlights the unique role of embodied locomotion in modulating the higher-order psychological dimensions of aesthetic experience.

4 Discussion

4.1 Changes in mood and the mediating role of the viewing experience

The present study examined how viewing paintings in a virtual reality environment with an HMD affected viewers’ emotional states and subjective experiences. In accordance with prior research indicating that exposure to art can promote wellbeing (e.g., Ishiguro et al., 2021), significant increases in positive mood and decreases in negative mood were observed following VR art viewing. Comparable increases in positive mood and decreases in negative mood have been observed in creative green prescription programs that integrate artistic endeavors with nature experiences (Thomson et al., 2020). This pattern aligns with the framework of neuroaesthetics, which posits that aesthetic experiences involve complex reward-related neural processing that actively mitigates stress (Williams et al., 2018; Magsamen, 2019). These findings suggest that VR art appreciation positively affects participants’ psychosocial wellbeing. While the present study focused specifically on art appreciation, the strength of VR lies in its ability to creatively arrange diverse content within virtual spaces, enabling the integration of not only art but also natural environments. Mediation analysis further revealed that improvements in museum experience ratings, as evaluated using the Generic Wellbeing Questionnaire in the UCL Museum Wellbeing Measure Toolkit (Thomson and Chatterjee, 2014), significantly contributed to post-viewing increases in positive mood, highlighting the importance of subjective experiential quality for emotional benefits. This mediating role of subjective experience is consistent with the information-processing model of aesthetics (e.g., Leder et al., 2004), which suggests that emotional outcomes depend heavily on successful cognitive mastery and continuous evaluation of the artwork rather than mere visual exposure.

4.2 Impact of artwork scale and embodied locomotion on evaluation

Evaluative responses to the artworks were influenced by both the display dimensions and walking conditions. Larger displays elicited stronger feelings of being overwhelmed and a greater desire to view artworks from multiple positions, while smaller displays prompted more focused, stationary viewing and increased attention to fine details. These findings suggest a robust scale effect, in which the visual size of an artwork shapes perceptual strategies and experiential interpretation. Seidel and Prinz (2018) showed that alterations to a painting’s dimensions or placement can influence its perceived aesthetic value. Specifically, their study found that enlarging a painting and positioning it higher on a wall were perceived as more aesthetically pleasing. These findings suggest a reciprocal relationship between perceived greatness and the spatial position and scale of an artwork. This phenomenon can also be interpreted through the psychological construct of “awe,” which is primarily triggered by perceptual vastness and a subsequent need for cognitive accommodation (Keltner and Haidt, 2003). Furthermore, walking was associated with heightened feelings of immersion, amplified awe responses, and increased enjoyment. This pattern suggests a “walk effect,” a phenomenon related to interacting with the artwork through physical movement. This finding aligns with the theory of embodied simulation in aesthetics (Freedberg and Gallese, 2007; Cabbai et al., 2024), suggesting that physical engagement and motor resonance enhance the empathic and emotional bond with the artwork. These effects suggest that VR can introduce affordances for movement-based engagement that are difficult to replicate in traditional screen-based presentations. This has implications for both museum exhibition design and art education. Navigation in the virtual reality environment did not produce a uniform change in emotional response; rather, its effects were associated with the dimensions of the presented artworks.

4.3 Behavioral indicators of spatial viewing strategies

Analysis of walking behavior revealed how participants actively adjusted their spatial position relative to the artwork within the viewing space during VR art appreciation. The artwork’s size significantly influenced the viewing distance. Specifically, participants moved closer to smaller paintings but moved away from larger works. They prioritized observing details at close range while focusing on the overall composition from a distance. This behavioral pattern closely mirrors field observations in physical museums, where visitors naturally adjust their physical distance to accommodate the scale of the canvas and optimize their perceptual understanding of the artwork (Carbon, 2017). The forward-backward walking condition also contributed to more efficient spatial adjustment. Compared with the no-walking condition, forward-backward walking resulted in shorter movement distances, fewer joystick-based jumps, and shorter viewing durations, indicating that participants could rapidly locate an optimal distance without unnecessary exploration. This efficiency highlights the importance of body-based self-motion cues; continuous vestibular and proprioceptive feedback during physical locomotion facilitates accurate spatial updating, mitigating the spatial cognitive costs and disorientation commonly associated with teleportation-based VR navigation (Cherep et al., 2020). By contrast, the no-walking condition may have constrained participants’ viewing positions, often keeping them farther back, and required more time to complete each trial.

Viewing location preferences further clarified these strategies. Participants predominantly stayed near the frontal center, particularly for small artworks, while time spent in rear areas increased as paintings became larger. Although the pattern was subtle, participants tended to view paintings from a slightly right-shifted position, a pattern potentially related to individual lateral dominance (e.g., handedness, ocular dominance), though this remains speculative. Importantly, movement variability (an index of positional exploration) systematically decreased from large to small paintings and was consistently smallest in the forward-backward walking condition. This suggests that efficient distance control supports a more stable and deliberate viewing posture. In contrast, the no-walking condition exhibited the highest variability, likely reflecting restricted ability to fine-tune viewing position, thereby prompting compensatory adjustments. Taken together, these findings indicate that viewers actively adjust both distance and lateral position in response to artwork size and available movement freedom to obtain perceptually optimal visual information. Moreover, VR forward-backward walking appears to reproduce the most natural and efficient museum-like viewing behavior. Thus, the present study shows that VR forward-backward walking promotes realistic and fluent viewing behavior, and behavioral metrics in VR sensitively reflect viewers’ cognitive viewing strategies.

4.4 Dissociation between visual attention and affective experience

A key finding of the present study is the clear dissociation between affective experience and visual attention, as revealed by gaze analysis. Although psychological evaluations and locomotion patterns demonstrated substantial variation across conditions, the spatial distribution of fixations exhibited remarkable consistency, aligning with the artworks’ semantic features. The stability of the gaze indicates that spatial visual exploration is predominantly driven by bottom-up visual saliency, such as faces or high-contrast areas, which are intrinsic to the artwork’s composition (Itti and Koch, 2001). Conversely, the subjective emotional responses were strongly modulated by the top-down context of embodiment and scale. This dissociation aligns with the dual-process models of aesthetic appreciation, which posit that lower-order perceptual analysis and higher-order affective evaluations are partially independent functional stages (Leder et al., 2004; Pelowski et al., 2017). Consequently, embodied interaction functions not as a primary filter for visual attention but rather as an affective amplifier, thereby modulating how the viewer experiences the artwork without altering where they look within the artwork.

4.5 Implications, limitations, and future directions

The findings of this study demonstrate that viewing VR art fosters positive emotional responses and offers unique experiential affordances. The substantial scale and walk effects observed in the evaluation of artworks underscore the capacity of VR to manipulate spatial experience (particularly display size and embodied mobility) in ways that exceed the physical constraints of traditional museums. This aligns with the foundational principle that physical walking in virtual environments significantly enhances presence and immersion compared with virtual travel metaphors (Usoh et al., 1999; Boletsis and Cedergren, 2019). By leveraging these unique affordances, VR art experiences can be positioned not only as entertainment but also as scalable interventions within the emerging field of digital therapeutics (DTx) for individuals with limited physical mobility (Riva et al., 2016).

Despite these contributions, some limitations must be addressed to guide future inquiry. First, the study sample consisted primarily of young adults with non-expert backgrounds in art. Therefore, the generalizability of the observed emotional and behavioral responses, including the walk- and scale-dependent patterns, must be verified in broader populations, such as older adults and expert art viewers, whose aesthetic processing and motor profiles may differ substantially. Second, the stimulus set focused exclusively on representational paintings with explicit semantic features, particularly human faces and figures. The remarkable consistency of gaze distribution across conditions may be strongly tied to these salient semantic focal points. Therefore, alternative interpretations should acknowledge that these gaze patterns may not generalize to abstract art, where visual exploration is less anchored in predefined semantic structures (Brieber et al., 2014). Third, the theoretical claims regarding the “walk effect” and “scale effect” must be carefully considered within the specific parameters of our experimental setup. The parameters include the range of artwork dimensions chosen, the physical boundaries of the 4 m × 4 m tracking area, and the short-duration viewing tasks. Fourth, in terms of user comfort, none of the participants reported severe cybersickness or requested to discontinue the experiment. The 1:1 physical locomotion design maintained complete concordance between visual feedback and body-based self-motion cues, including vestibular and proprioceptive inputs. This design avoided the visual-vestibular sensory conflicts commonly observed in virtual locomotion setups such as controller-based navigation or teleportation, thereby ensuring a high level of comfort.

Although the physical forward-backward walking condition significantly enhanced immersion, the limited dimensions of the laboratory environment still imposed boundaries that contrasted with the vast, open galleries of major cultural institutions. Future research should examine how these effects develop across expansive virtual environments. Additionally, the present study focused on evaluating the reproduction of real-world museum viewing behavior (i.e., observing artworks from an external perspective). However, VR technology has the unique potential to transcend physical boundaries by allowing viewers to step directly into the artistic space itself. Recent research has demonstrated that converting two-dimensional paintings into three-dimensional, immersive environments can significantly enhance narrative transportation, emotional engagement, and hedonic motivation (Lee and Youn, 2025). Future research should explore how embodied locomotion interacts with such transformed artistic spaces by directly comparing the psychological and behavioral impacts of traditional external observation with those of entering and exploring the interior of an artwork. Comparing the emotional and cognitive effects of VR viewing with on-site museum experiences could clarify the extent to which digital replicas can simulate the authentic engagement found in physical institutions (Marín-Morales et al., 2019). The integration of multimodal sensory enhancements, such as spatial audio and haptic feedback, or the exploration of social VR platforms where multiple viewers interact simultaneously, could further elucidate the interpersonal dimensions of virtual art appreciation. As VR archives continue to evolve, it will be important to understand how immersive digital environments can enhance, rather than replace, on-site museum experiences in the domains of education and cultural accessibility.

5 Conclusion

This study demonstrates that embodied interaction in virtual reality significantly enhances the art appreciation experience. While the fundamental allocation of visual attention remains anchored to the artwork’s semantic features, higher-order psychological responses are strongly modulated by display scale and physical locomotion. Specifically, forward-backward walking promotes the most efficient spatial exploration and heightens feelings of immersion and enjoyment. These findings reveal a critical dissociation between perceptual processing and affective evaluation. Overall, manipulating spatial scale and embodied mobility in VR offers powerful new affordances for designing immersive virtual museums, advancing cultural accessibility, and developing art-based DTx.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. Gemini 3 Pro (Google, Mountain View, CA, USA) was used solely to refine wording and clarity in select portions of the manuscript text. Gemini did not generate data, perform analyses, or contribute to scientific interpretation or conclusions. All scientific content and interpretations remain the responsibility of the authors.

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.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/frvir.2026.1914418/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Ramy Hammady, University of Southampton, United Kingdom

Reviewed by: Yogi Udjaja, National Taiwan Normal University, Taiwan

Jelena Atanacković Jeličić, University of Novi Sad, Serbia