Studies relied on different nature scenes for relaxation. In Zhang et al. (2023), for example, a panorama photo of Osmanthus trees (a common Chinese garden plant) was used for the intervention and a white wall without any smells as a group control. In Qi et al. (2022), four different environments were used, each presented as a three-dimensional 360-degree panoramic photo. The nature scenes included a lawn, a rose garden, an Osmanthus tree garden, and a pine forest. As in the other study, a white wall with no smell or audio was used for the control group. The work by Lopes et al. (2022) used a 360-degree computer-generated video of a forest environment; the study did not have a control group. The environment used in Abbott and Diaz-Artiles (2022) was a forested area including a campsite, a river, and a wood cabin; for the control condition, subjects were exposed to a stressful stimulus. In Schebella et al. (2020), a 360-degree video of low biodiversity, moderate biodiversity, and high biodiversity nature scenes was used, along with a virtual urban environment as a control condition. In Hedblom et al. (2019), 360-degree photos of a park and forest were used, while for the control condition, an urban environment was used. Lastly, in Amores et al. (2018), a beach scene was presented in 360-degree video format. For the control condition, it was mentioned that no stimulus was presented and no specific instructions were given to the participants, other than to not close their eyes.

While the studies above used HMDs as the display modality, other studies relied on projectors and display screens/TVs to provide the virtual experience. In Takayama et al. (2022), a video image of a forest environment was projected in a room. In Shin et al. (2022), a 270-degree video of a busy cafe scene was presented under three conditions, two of which included a nature scene. In Zhao et al. (2022), an LED display screen was used to transmit a video of a forest in autumn colors; the control condition had no stimulus. In Putrino et al. (2020), a high-definition projector was used to show a soothing natural landscape, comprised of either ocean scenes or forest scenes. In Song et al. (2019), a 4K-compatible display television was used and the visual stimulus was a photograph view of a forest landscape. The work in Sona et al. (2019), utilized an artificial window comprised of three high-resolution 4K LED screens with speakers to present a park scene and a lounge scenario. In Serrano et al. (2016), an HD video projector was used where participants could watch relaxing scenes, including a beach and a landscape covered in snow. Lastly, the studies in Putrino et al. (2020); Sona et al. (2019) relied on a hybrid setting, where real plants were also used. More details about the virtual environment presentation modality can be seen in Section 3.5.

Audio stimuli were used in most of the studies, with only two studies focusing on the visual-olfactory senses Zhang et al. (2023); Song et al. (2019). The audio was related to nature and often included bird and water sounds. Among the selected articles, only three studies used complementary sounds, in addition to nature ones: Lopes et al. (2022) included a guided self-reflection message presented in a young woman’s voice; Putrino et al. (2020) and Serrano et al. (2016) also included relaxing music. For the studies that used the control condition with urban environments Hedblom et al. (2019); Schebella et al. (2020), it was common for traffic sounds to match the environment. In Sona et al. (2019), only instrumental music was played. Finally, the work in Shin et al. (2022) compared an open window (with virtual nature) condition versus a closed window (with/without nature) in a coffee shop. In the open window condition, sounds of nature could be heard. In all three conditions, background babble noise, typical of a coffee shop, was also presented at a moderate volume level.

All scents used in the studies were related to nature for the intervention conditions with the natural environments. Some relied on natural scents, such as the work in Zhang et al. (2023), which used real osmanthus flowers to produce the olfactory stimulus, and the work in Qi et al. (2022) which used real samples of lawn, rose flowers, osmanthus flowers, and pine. In Song et al. (2019) they used Hinoki cypress leaf oil (Kiseitec Co., Wakayama, Japan) extracted by steam distillation from leaves and twigs of Hinoki cypress trees. On the other hand, eight studies used different types of diffusers to emit smells. In Takayama et al. (2022), the essential oil of Abies sachalinensis (Sakhalin fir) was used. The authors in Shin et al. (2022) used two ultrasonic essential oil diffusers with the aroma of coffee and wet dirt (nature condition). In Zhao et al. (2022), the Aroma Shooter device was used to provide a forest scent. In Abbott and Diaz-Artiles (2022), scents were produced with the Olorama Scent Generator (Olorama Technology Ltd, Spain) and smells of sausage, coffee, wet ground, pine, woods, roses, lavender, and honey were used. In Putrino et al. (2020), yuzu, hinoki, roman chamomile, and lavender essential oils were used. In Sona et al. (2019), scents of rosewood, geranium, ylang-ylang, olibanum (frankincense), and hyssop were used in the natural outdoor condition and composition of rosewood and cardamom in the indoor condition. In Serrano et al. (2016), a lavender scent was distributed with a ceramic diffuser. In Lopes et al. (2022) a SENSIKS (SENSIKS, Netherlands) multisensory pod was used with scents of dirt, forest, and grass.

Lastly, three studies used their methods of aroma emission. In Schebella et al. (2020), scents were applied to cotton pads and attached to the HMD, a range of 16 nature scents (e.g., several types of grass, soil, leaves, and flowers) and four urban scents (e.g., dust and turpentine) were initially tested on a convenience sample of 11 people. In the end, they used only one scent in the urban and low biodiversity scenes, and three different natural scents in the high biodiversity scene, however, they did not specify the exact name of the aroma. All scents were selected from the Demeter Fragrance Library ¹ . In Hedblom et al. (2019), the authors used their own computer-controlled olfactometer (Lundström et al., 2010), where city odors included diesel, tar, and gunpowder and for park odors they used grass and different forest odors, including two evergreen species and mushroom. In Amores et al. (2018), an essence necklace was used to emit a lavender essential oil scent.

The methodology for selecting scents in multisensory VR studies varies, often reflecting the unique objectives and contexts of each research project. For instance, Zhang et al. (2023) selected the scent of Osmanthus, a flower renowned as one of the ten most famous flowers in China, where the study took place, thus highlighting the cultural relevance for scent choice. Takayama et al. (2022); Song et al. (2019) focused on scents from trees that complemented the visual aspects of the VR environment, thus enhancing the sensory congruence for users. Other studies, however, such as that conducted by Putrino et al. (2020); Amores et al. (2018); Serrano et al. (2016), opted for scents such as lavender and Roman chamomile essential oils, chosen for their documented effects in reducing anxiety and stress in the scientific literature.

Qi et al. (2022) took it one step further and used a portable odor sensor (COSMOS XP-329 III R; Rank Value, range 0–2000) to measure and select odor concentrations that closely mimicked the landscapes used in the virtual environments. Schebella et al. (2020) engaged participants in the scent selection process, where 11 subjects identified which scents they found to more closely match the natural or urban settings in the experiment. Seven scents were accurately recognized and subsequently utilized in corresponding VR environments, ensuring the olfactory experience matched the visual setting. Moreover, Sona et al. (2019) not only referenced literature for scent selection but also conducted preliminary tests with 12 subjects to determine the intensity of ambient scents and identify perception thresholds. Other studies Shin et al. (2022); Lopes et al. (2022); Zhao et al. (2022); Abbott and Diaz-Artiles (2022); Hedblom et al. (2019), chose scents based on the landscapes in their experiments.

In addition to the audio-visual-olfactory stimuli detailed above, three studies also investigated tactile/somatosensory stimuli. In Lopes et al. (2022), heating elements and four fans surrounding the sides, back, and front of the users, as well as acoustic vibrations on the user’s seat were used. In Zhao et al. (2022), fans were mounted on the top of the table pointing toward the user’s head and table surface, as well as infrared lights underneath the table were used for temperature control. Lastly, in Serrano et al. (2016), artificial grass was installed on the ground, allowing the participant to touch and feel it during the immersive VR experience.

Among the selected studies, seven presented the virtual reality experience using display/projectors in an indoor setting (Serrano et al., 2016; Hedblom et al., 2019; Sona et al., 2019; Putrino et al., 2020; Shin et al., 2022; Takayama et al., 2022; Zhao et al., 2022). In Shin et al. (2022), three conditions were tested: no window (brick wall), closed window (nature view), and open window (nature view) with all conditions including the scene of a coffee shop. Participants were randomized into one of three groups. Participants were immersed in a 270-degree VR simulation within the laboratory. The participant’s viewpoint featured a lifelike projection of the student union coffee shop, spanning the front and left walls. Additionally, the right wall showcased three distinct conditions. Throughout all conditions, a coffee fragrance and ambient conversation noises were used at a moderate volume level. In the open window condition, the immersive experience was further enhanced by the sounds of birds chirping/singing in the background, accompanied by the aroma of wet dirt. In Takayama et al. (2022), participants were exposed to a 20-min digital shinrin-yoku environment that reproduced visual, auditory, and olfactory elements, each session accommodated a maximum of five subjects, and an adequate distance between individuals was maintained. Video images of a forest were projected onto three walls and the ceiling of the experimental room.

The work in Zhao et al. (2022) focused on multimodal augmentation via lighting, audio, video, airflow, heating, and scent elements to improve the productivity and wellbeing of workers in an open-plan office. In this work, a comparison was made between three conditions (neutral vs. unscented forest vs. scented forest), and a stress induction phase was also performed. For the neutral condition, the LED panels displayed no visuals, whereas for the other two conditions (with scent and without scent) a scene forest was displayed. Another study created a recharge room with multisensory (visual, auditory, and olfactory) nature-inspired experiences where healthcare workers could relax and reduce stress Putrino et al. (2020). The environment was carefully designed to evoke a serene natural setting. Silk imitation plants were strategically placed to simulate lush greenery, while scenes of natural landscapes were projected onto a blank wall. The ambiance was further enhanced by soft, color-matched lighting, which complemented the projected landscapes. High-definition audio recordings of nature sounds were combined with relaxing music. To complete the multisensory immersion, essential oils, and calming scents were diffused using an essential oil diffuser, heightening the impression of being fully surrounded by a natural environment.

The study in Song et al. (2019) employed four different conditions, each lasting 90 s, namely,: a forest landscape image featuring Hinoki cypress trees without any accompanying odor (visual-only), a gray image infused with Hinoki cypress leaf essential oil (olfactory-only), a forest landscape image of Hinoki cypress trees combined with the scent of Hinoki cypress leaf essential oil (combined stimulus), and a gray image without any odor (control condition). The study conducted by Sona et al. (2019) examined five distinct conditions, including a control condition with no window, sound, or scent. Additionally, two audio-visual conditions were assessed: a nature condition with a window displaying nature scenes, accompanied by bird sounds and neutralizing air/no scent, and a lounge condition featuring a window displaying a lounge scene, instrumental music, and neutralizing air/no scent. Furthermore, the study explored two audio-visual-olfactory conditions where scents were introduced as an additional element.

In Serrano et al. (2016), the experiment was performed in a VR-room, in which a virtual reality environment was created where participants were immersed inside a two-story house called House of Relaxation; the goal was to induce relaxation and a sense of presence and to analyze whether adding touch and smell altered the experience outcomes. Inside the environment, users could perform different actions, such as modifying the house’s decoration, watching a video clip through a projector, changing the intensity of the lights and the color of the walls, opening or closing doors and windows, moving between the two floors of the house, moving or adding furniture, changing the ground to grass or sand, or change the exterior landscape to reflect another natural context (e.g., beach or snow). The lavender scent was used to stimulate the olfactory sense, and for the tactile stimulus, they used artificial grass. They compared four conditions: audio-visual, audio-visual-olfactory, audio-visual-tactile, and audio-visual-olfactory-tactile.

The other seven studies relied on VR HMDs to present the visual stimuli (Amores et al., 2018; Hedblom et al., 2019; Schebella et al., 2020; Abbott and Diaz-Artiles, 2022; Lopes et al., 2022; Qi et al., 2022; Zhang et al., 2023). The most recent study compared three conditions, i.e., olfactory-only, visual-only, and visual-olfactory) to a control condition Zhang et al. (2023). The experiment took place in a room with white walls and ceilings, ensuring a consistent indoor environment to prevent the subjects’ moods from being influenced by external factors, such as weather and light. In the control group, an odorless condition was maintained, with participants exposed to an image of a white wall, devoid of any stimuli associated with plant smells or garden landscapes. In the olfactory condition, participants experienced a natural concentration of fragrance emitted from Osmanthus fragrans. For the visual stimulus, a panoramic photo of Osmanthus trees was displayed. The visual-olfactory condition involved the combination of both visual and olfactory stimuli.

In Qi et al. (2022), a total of 308 volunteers participated in the study and were divided into 14 independent groups. These groups were designed to explore three different stimulus scenarios: audio-visual, audio-olfactory, and audio-visual-olfactory. Four plant landscapes (lawn, rose garden, osmanthus tree garden, and pine forest), with each landscape accompanied by its respective scent when olfactory stimuli were integrated, plus a single birdsong group (1 group) that visualized a white wall and received no odor stimulus, and a control group (1 group) of participants who viewed only a white wall (no birdsong and no odor). Each condition comprised 22 participants, ensuring an equal distribution across the study.

In Lopes et al. (2022), two conditions were compared: audio-visual and audio-visual-olfactory-tactile. Participants were seated inside a SENSIKS multisensory booth, which provided a fully immersive experience. In the audio-visual condition, participants were immersed in a forest environment for approximately 1 minute, accompanied by the soothing sounds of nature. Additionally, a soft voice of a young woman whispered in the background, prompting participants to engage in self-reflection on existence and consciousness. In the audio-visual-olfactory-tactile condition, scents of nature, such as dirt, forest, and grass were included, while fans and heating elements located on the sides, back, and front of the participant were used to simulate wind and sunlight matching the visual content.

Another study that showed a positive impact of introducing olfactory stimuli to reduce negative affect and state anxiety levels was the work in Abbott and Diaz-Artiles (2022). There, they analyzed an audio-visual and an audio-visual-olfactory condition with a task to induce stress between conditions. They built two VR environments using Unreal Engine 4. The first VR environment was designed as a nature-inspired forested area, complete with a campsite, a river, and a wood cabin. To enhance the immersion, realistic sounds of nature such as a rushing river and chirping birds were incorporated. Additionally, when olfactory stimuli were required, scents of nature were introduced. The second VR environment was specifically tailored for the stress induction portion of the experiment, known as the Trier Social Stress Test (TSST). This environment was modeled after a small auditorium, featuring approximately 30 static audience members. These virtual audience members remained passive throughout the session, providing no feedback to the subjects, whether positive or negative. Lastly, in Amores et al. (2018), a baseline condition (no stimulus) was compared to an audio-visual-olfactory condition. The participant was comfortably seated while wearing the Essence necklace, which emitted a fragrance of lavender. Participants were presented with a 360-degree video of a beach setting and the audio accompanying the video.

Two studies compared between a natural and an urban environment (Hedblom et al., 2019; Schebella et al., 2020). In Schebella et al. (2020), three natural environments, with different levels of biodiversity (low, moderate, and high) were compared to an urban environment control condition. A significant effect of multisensory biodiversity was observed on stress recovery related to the urban environment. Moreover, in Hedblom et al. (2019) an urban area was compared to an urban park and an urban forest. Each condition included audio-visual-olfactory stimuli appropriate for the scenes. The two virtual nature conditions resulted in greater stress reduction relative to the urban environment. The perceived pleasantness was associated with the olfactory and auditory stimuli, but not with the visual stimulus. The study suggested that olfactory stimuli may be better at facilitating stress reduction than visual stimuli and highlights the relevance of multisensory approaches for stress reduction.

In total, 13 studies performed subjective measurements to analyze the mood changes associated with the multisensory VR exposure. Four studies administered the Perceived Restorativeness Scale (PRS) to measure how restorative an environment was (Sona et al., 2019; Shin et al., 2022; Takayama et al., 2022; Zhao et al., 2022). In Sona et al. (2019), questions about mood, fatigue, and arousal (before and after exposure) were also measured. Two studies used the Restoration Outcomes Scale (ROS) to analyze the subjective restorativeness, where Shin et al. (2022) requested the questionnaire only after VR exposure, while in Takayama et al. (2022) the measurement was requested before and after exposure. Two other studies used a semantic differential (SD) method. In Qi et al. (2022), a survey concerning the overall quality evaluation of the environment was used, while in Song et al. (2019) the participant’s subjective spatial impressions were collected via a questionnaire with four opposite adjectives, each evaluated on 13 scales.

Two studies measured mood with the Positive and Negative Affect Schedule (PANAS) (Abbott and Diaz-Artiles, 2022; Takayama et al., 2022). Two studies used the Visual Analogue Scale (VAS) before and after exposure. While in Schebella et al. (2020) VAS was used to measure the perceived stress, anxiety, insecurity, calmness, and happiness levels of the participants, in Serrano et al. (2016) it was used to evaluate levels of sadness, joy, anxiety, and relaxation moods. One study used the Profile of Mood States (Takayama et al., 2022). The State-Trait Anxiety Inventory (STAI) was used in two studies. In Qi et al. (2022), the full 20-item version was used, whereas in Abbott and Diaz-Artiles (2022) a modified version including only six-items was used. The work in Serrano et al. (2016) mentioned using the state scale of the STAI but did not report the results; instead, it reported results using the Self-Assessment Manikin (SAM) (before and after exposure) and the Presence Self-Assessment Manikin (after exposure).

Single-item measures were also used across different studies, including the Relaxation Rating Scale (RRS) which was collected in Amores et al. (2018) before and after the exposure, and in Lopes et al. (2022) just after exposure. Overall, five studies acquired some type of subjective measure only after the virtual experience (Hedblom et al., 2019; Song et al., 2019; Lopes et al., 2022; Shin et al., 2022; Zhao et al., 2022). Moreover, five studies used tasks to induce stress prior to exposure. In Zhao et al. (2022), participants were given a 3-min reading comprehension task similar to those given on a graduate standardized test. In Abbott and Diaz-Artiles (2022) and Schebella et al. (2020), the modified trier social stress test (TSST) was used. In Hedblom et al. (2019), mild electric shocks were used as a stress trigger. Finally, in Sona et al. (2019), a sequence of tasks was used to induce stress, including a single N-back task, a Stroop task, and an Attention Network Task.

The addition of olfactory stimuli in virtual nature experiences has resulted in inconsistent outcomes concerning effects on emotional restoration and positive/negative emotion. Table 5 shows a summary concerning the restorative effects mentioned in the studies. While some studies have reported an improvement in mood and stress reduction, others have reported no significant differences. For example, in Shin et al. (2022), the conditions where the participants looked out the window at nature showed greater restorative qualities than the absence of nature, but comparisons between the closed window condition (only view of nature) and the open window condition (sight, smell, and sounds of nature) showed no significant differences. Several participants did, however, report the open window condition to be more captivating. Results of one-way ANOVAs showed significant restorativeness/restoration across conditions for the ROS subscales ‘Being Away’ and ‘Fascination’ (p < 0.05).

In Qi et al. (2022), the audio-visual-olfactory stimuli led to an increased physiological restoration and overall perceived quality, whereas the audio-olfactory stimuli, despite having no significant effect on psychological restoration, showed an increase in perceived overall feelings of attraction to the landscape and a sense of overall harmony. For the audio-visual and audio-visual-olfactory conditions, no significant differences in psychological effects were found. Only for the auditory stimulus was the STAI-S value significantly lower than the control group (p = 0.04). In another study Zhao et al. (2022), in turn, after the two virtual forest conditions (with and without nature smell), ratings of perceived restoration and focus were significantly p < 0.05 higher than the neutral condition (i.e., post-hoc comparisons, using a Wilcoxon signed-rank test, audio-visual-olfactory, Z = 17.50, p = 0.000 and audio-visual Z = 13.0, p = 0.000). However, stress development was slower and recovery was greater in the forest without scent condition compared to the other conditions. In addition, Friedman’s test showed a significant (p < 0.05) difference in restoration perception among the three conditions (X ²(36) = 39.78, p = 0.000). Also, in Takayama et al. (2022) negative mood states (measured using POMS), significantly decreased (p < 0.01) after the experience (i.e., tension-anxiety, depression, anger-hostility, fatigue, and confusion). In addition, psychological restorative effects were confirmed, with a significant decrease (p < 0.01) in “negative affect” (measured using PANAS) and a significant increase (p < 0.01) in the sense of restoration (measured using ROS) after the experience.

Regarding negative emotions, in Abbott and Diaz-Artiles (2022), participants experienced a notable reduction in STAI-6 scores following the audio-visual-olfactory condition compared to the audio-visual condition (p = 0.03). Furthermore, when contrasted with their stressed state, there was a significant decrease in both PANAS Negative Affect (p = 0.003) and STAI-6 scores (p = 0.001) after the audio-visual-olfactory. Additionally, STAI-6 scores showed a significant decline from the baseline in the audio-visual-olfactory condition (p = 0.013). Results suggested that the addition of olfactory stimuli to the VR environment aided in reducing negative affect and state anxiety levels.

For positive emotions, significant improvements in relaxation were reported in Lopes et al. (2022); Amores et al. (2018). In Amores et al. (2018), the results showed that the multisensory VR experience was able to promote increased relaxation relative to the baseline condition, although not a significant difference. The findings corroborate previous studies that showed forest-related stimuli significantly increasing the participants’ feelings of comfort and relaxation relative to a control condition (p < 0.05) Song et al. (2019). Similar findings were shown in Lopes et al. (2022) where greater relaxation was achieved as more senses were stimulated (p < 0.05). Moreover, the work in Sona et al. (2019) the correlation between the two multisensory environment experiences showed that nature views were perceived as more pleasant (r = 0.74, p < 0.01) and restorative (r = 39, p < 0.01) than the control group (no stimulus). Adding nature smells facilitated the recovery of personal resources (mood, fatigue, arousal) via greater scent pleasantness and fascination (r = 0.18, p < 0.10). Moreover, correlation analyses between experiences showed that the view was perceived as more pleasant (r = 0.53, p < 0.01) and the environment as more restorative (r = 0.22, p < 0.05) in the nature simulations than in the lounge simulations. In Serrano et al. (2016), the results showed that the VR experience was effective in inducing relaxation F(1,132) = 90.31, p < 0.001, where positive moods and the sense of presence increased. Moreover, a significant decrease in arousal (F(1,132) = 92.04, p < 0.001) was observed after all four VR experiences. However, no statistical differences were found between the four groups on emotions and sense of presence. In Hedblom et al. (2019), the perceived pleasantness, average over environments and sensory stimuli rated on a 1–100 scale, was significantly higher in the park environment (Mean = 69.21, SD = ±11.1) compared to both the forest (Mean = 62.7, SD = ±12.74; p = 0.01) and the urban area (p < 0.001). Perceived pleasantness was also higher for the forest relative to the urban area (Mean = 37.19, SD = ±14.22; p < 0.001). There were significant differences in the ratings of perceived pleasantness among the three environments, as indicated by the sensory stimuli (t = 90.01; p < 0.001; pairwise comparisons).

Regarding perceived stress, Schebella et al. (2020) showed that a multisensory experience (visual, auditory, and olfactory) was associated with better recovery in all measures of wellbeing relative to a visual-only experience. Median anxiety recovery scores were significantly greater in the audio-visual-olfactory high biodiversity environment (33.00) than in the visual-only (17.50) using a Mann–Whitney test (U = 90.50, z = 1.971, p = 0.047). When compared to the virtual urban environment, stress recovery was most effective in a low-biodiversity environment (p < 0.05). In Putrino et al. (2020), a greater reduction in perceived stress after the multisensory immersive experience was reported (p < 0.001), where the level of stress had an average reduction of 59.6% in the self-reported ratings. Similarly, in Hedblom et al. (2019), the park and forest conditions resulted in significant stress reduction, whereas the urban area condition did not.

The study described in Zhang et al. (2023) investigated the effects of olfactory and olfactory-visual stimulation on blood pressure and pulse. During the stimulation, it was observed that systolic blood pressure (SBP) values remained relatively unchanged. However, diastolic blood pressure (DBP) values exhibited a significant increase (ΔDBP = 4.37 ± 1.69 mmHg, p < 0.05). Additionally, the pulse pressure difference (PP) values demonstrated a notable decrease during both stimulations (ΔPP = −4.56 ± 1.24 mmHg, p < 0.05). Furthermore, during olfactory stimulation, pulse values also showed a significant decrease (ΔP = −2.34 ± 1.16 bmp, p < 0.05). In contrast, visual stimulation did not lead to significant changes in SBP, DBP, or PP. These findings suggest that the physiological effects were greater when the olfactory system was stimulated. Conversely, in Takayama et al. (2022) no significant differences in DBP, SBP, or pulse rate were seen before and after experiencing the digital forest bathing environment.

Only one study looked at skin temperature (ST) changes (Qi et al., 2022). While the results did not show significance, the restorative tendency for auditory stimuli was greater than in the control group.

Similarly, only one study measured salivary amylase (Takayama et al., 2022). No significant effect was observed when comparing the data before and after the multisensory digital forest experience.

The study described in Song et al. (2019) demonstrated that the combined visual and olfactory stimuli yielded significantly reduced oxygenated hemoglobin (oxy-Hb) concentrations in both the left (control, −0.08 ± 0.10 μM; visual-olfactory, −0.48 ± 0.09 μM; p < 0.05) and right (control, 0.02 ± 0.10 μM; visual-olfactory, −0.46 ± 0.08 μM; p < 0.05) prefrontal cortices, compared to the control condition. Additionally, when exposed to the olfactory stimulus alone, a significant decrease in oxy-Hb concentration was observed in the right prefrontal cortex compared to the control condition (olfactory, −0.32 ± 0.12 μM). This decrease in oxy-Hb concentration aligns with previous literature suggesting its association with physiological calming (Hoshi et al., 2011). Overall, these findings suggest greater relaxation effects resulting from the combined visual and olfactory stimuli.

Regarding the cardiovascular system, there were five studies that show results related to relaxation and stress reduction in multisensory experience (Song et al., 2019; Schebella et al., 2020; Lopes et al., 2022; Takayama et al., 2022; Zhao et al., 2022). In Lopes et al. (2022), despite not showing a statistically significant difference, the mean of the high frequency (HF) component of the electrocardiogram (ECG) signal (across all participants) for the multisensory condition was higher compared to the audiovisual-only condition. These findings are corroborated by other studies that showed increases in HF during deep relaxation situations, such as meditation, or a decrease in HF during stressful situations. In addition, the standard deviation of the absolute first difference of the ECG feature showed a significant negative correlation (r = −0.46, p < 0.05) with the subjective relaxation rating for the multisensory condition, thus suggesting that the users felt more relaxed in the multisensory experience. Furthermore, the results also showed a decrease in the maximum blood volume pulse (BVP) signal between the audio-visual and multisensory (audio-visual-olfactory-haptic) conditions, suggesting greater relaxation potential in the latter condition (Parent et al., 2020).

In Takayama et al. (2022), the heart rate significantly decreased (p < 0.01) and HF was significantly higher (p = 0.014) during the exposure to multisensory forest bathing compared with that during the resting state. On the other hand, the study in Zhao et al. (2022) used heart rate variability (HRV) to compare three conditions during stress-inducing and recovery tasks. They found significant differences (p < 0.05) in HRV among these conditions. Specifically, the audio-visual condition showed a notably higher HRV than the neutral condition across all metrics. Additionally, HRV in the audio-visual condition was generally higher than in the audio-visual-olfactory condition. The results suggest that the nature experience without scent was the most effective in promoting relaxation.

Regarding the comparison between urban and natural environments, in Schebella et al. (2020), results showed that there were no significant differences in recovery from stress induction in terms of heart rate and HRV. However, regarding the stimulus in the experience, the multisensory experience showed that the recovery from stress induction relative to a visual-only condition was higher in terms of reduced heart rate and HRV. Moreover, the greatest recovery from stress induction based on median HR recovery scores (X ²(2) = 9.234, p = 0.007) was found in the low biodiversity condition, relative to the moderate biodiversity natural environment. Lastly, in Song et al. (2019), findings derived from HRV (i.e., ln(LF/HF)) showed that only the visual stimulus resulted in significantly decreased sympathetic nervous activity compared to the control condition (control, −0.26 ± 0.17; visual, −0.67 ± 0.17; p < 0.05), which indicates a decrease in stress. When comparing the four conditions (control, visual, olfactory, and visual-olfactory) there was no significant difference in heart rate.

Four studies reported changes in electrodermal activity (EDA) when including multisensory experiences. In Zhang et al. (2023), in the visual condition, the amplitudes of skin conductance (SC) (ΔSC = 0.19 ± 0.01 μΩ, p < 0.05) increased significantly relative to the control group. In addition, SC values significantly increased under the olfactory–visual stimulus method (ΔSC = 0.45 ± 0.34 μΩ, p < 0.05), and this increase was significantly higher than that of the olfactory-only condition. These findings may indicate that the effect of smelling and seeing garden plants provides more excitement than just smelling or seeing. On the other hand, the work in Qi et al. (2022) showed that in terms of the physiological restoration, the skin conductance level (SCL) value revealed a negative effect in the audio-olfactory stimulation, only the audio stimulus resulted in higher levels of restoration than the control group, illustrated by the SCL (p = 0.04).

According to the findings in Lopes et al. (2022), the participants in the multisensory condition exhibited a lower average of the low-frequency component in the EDA signal compared to the audio-visual condition. An elevation in the low-frequency component of EDA has been linked to stress (Posada-Quintero et al., 2016). Moreover, a Pearson correlation analysis revealed significant correlations (p < 0.05) between EDA features and relaxation ratings. Notably, the mean of the negative first difference of the EDA signal showed a correlation of r = 0.42, and the standard deviation of the EDA signal had a correlation of r = −0.43. Lastly, in the study by Hedblom et al. (2019), a greater decrease in EDA was observed in natural (park) environments compared to urban environments, with a significant difference (β = 2.02, (t(2471) = 2.45; p < 0.02)). A similar trend, though not statistically significant, was also noted between park and forest environments (β = 0.56, (t(2471) = 0.67, p = 0.25)). The study used a regression model to explore the relationship between perceived environmental pleasantness and SCL values during stress and recovery periods. It was found that only odor pleasantness significantly predicted SCL during both periods (stress period: β = −0.164; p = 0.04; recovery period: β = −0.188; p = 0.03), with auditory pleasantness showing marginal significance in the recovery period (β = −0.45; p = 0.05).

Most studies measuring electroencephalographic activity focused on alpha waves, which represent a state of calm and feelings of quietness and relaxation. Two studies, in addition to alpha waves, also showed results related to beta waves, which are related to effortful thinking and active attention. In Zhang et al. (2023), there was a significant increase in alpha and beta wave power during visual and combined olfactory-visual stimulation compared to before the stimulation. Specifically, alpha waves increased from 13.90 ± 7.19 μV to 20.10 ± 9.45 μV (p < 0.001) in the visual condition and from 16.03 ± 5.03 μV to 18.31 ± 6.77 μV (p < 0.05) in the olfactory-visual condition. Beta waves also rose from 8.27 ± 5.09 μV to 13.78 ± 6.79 μV (p < 0.001) in the visual condition and from 10.12 ± 4.35 μV to 11.52 ± 4.87 μV (p < 0.05) in the olfactory-visual condition. No significant changes were noted with olfactory stimulation alone. However, when compared to a control group, all three conditions showed a significant increase in both alpha and beta wave powers.

In Qi et al. (2022), audio-visual-olfactory stimuli led to greater restoration, but only for beta wave power (compared with the only-audio condition). As for the alpha wave power, correlations were seen with a lifting effect in the audio-visual condition and a negative effect in the audio-olfactory condition (p = 0.00) with a relatively equal mean difference when compared to the audio-only condition. Combining audio (birdsong) with olfactory stimuli from environments such as rose gardens and pine forests led to a significant decrease in α power (p ≤ 0.01). An osmanthus garden also notably lowered α power (p = 0.00), indicating that certain plant landscapes may induce negative physical arousal. Regarding β band power, adding visual and visual-olfactory stimuli to audio-only (birdsong) scenarios generally improved restoration (p = 0.00). However, in lawn settings, the addition of visual-olfactory stimuli had a variable impact on β values (p = 0.00). The work in Amores et al. (2018) also investigated changes in theta wave, which can be observed during meditation and drowsiness. A relaxation metric based on the Renyi entropy of different EEG frequency bands was proposed and showed that the relaxation score increased significantly (p < 0.05) between the audio-visual-olfactory experience and the control condition (no stimulus).

Several studies have faced limitations due to demographic constraints (e.g. (Sona et al., 2019; Qi et al., 2022; Shin et al., 2022; Zhang et al., 2023)). These studies often do not account for the full spectrum of diversity in age, gender, ethnicity, or socioeconomic status. This lack of diversity can hinder the generalizability of the findings, potentially leading to conclusions that may not apply broadly across different demographic groups. For example, the majority of the participants in the studies were comprised of university students, which could not fully represent the target population for multisensory immersive experience interventions. Moreover, very few studies had balanced gender representation in the participant pool. Some previous studies have reported increased chances of motion sickness (known as cybersickness) for females relative to males (e.g. (Kelly et al., 2023)). As such, future studies should strive to include a more diverse sample, including from a patient population, to enhance the generalizability of the findings.

Moreover, the reliability of the reported findings is often compromised by limited sample sizes (e.g. (Amores et al., 2018; Song et al., 2019; Schebella et al., 2020; Abbott and Diaz-Artiles, 2022; Lopes et al., 2022; Takayama et al., 2022; Zhao et al., 2022)). Small sample sizes can result in poor statistical power, making it challenging to detect true effects or differences, and may lead to results that are not replicable in larger, more diverse samples. Researchers should aim to increase sample sizes in future studies. Larger sample sizes would improve the statistical power and the robustness of the findings, contributing to more reliable and valid research outcomes.

Another significant limitation, as noted by Song et al. (2019), is the lack of studies with populations with heightened daily stress (e.g., first responders). Overlooking high-stress groups can lead to a gap in understanding their specific needs and responses. Future studies should concentrate on these populations, as targeted research can provide important insights into stress management and resilience, aiding in the development of effective interventions and therapies for those heavily impacted by stress.

Lastly, cultural background plays a pivotal role in shaping an individual’s preference for specific natural odors, deeply influencing the design and customization of the olfactory element in a virtual experience. For example, the Japanese practice of Shinrin-Yoku highlights the cultural value placed on the immersive experience of forest scents, such as those from pines, cedars, and spruces. Furthermore, the type of vegetation and biodiversity characteristic of a region can significantly influence the natural odors that individuals find comforting. For instance, individuals from regions rich in coniferous forests may have a preference for the scents of pine or fir, while those from tropical areas might be drawn to the fragrances of flora like banyan or kauri trees. Acknowledging these cultural and regional differences is crucial in creating more inclusive and personalized virtual reality experiences. By offering a wide selection of natural scents, from the refreshing aroma of a dense forest to the salty breeze of a beach, virtual environments can cater to a global audience. This approach not only enhances the immersive quality of virtual experiences but also fosters a deeper connection between users and the virtual natural environments they choose to explore, based on their personal experiences with nature. Further research should explore the impact of olfactory preferences on overall relaxation potential.

Studies, such as those by (Song et al., 2019; Zhang et al., 2023), focused predominantly on visual and olfactory stimuli, while works, such as (Amores et al., 2018; Putrino et al., 2020; Takayama et al., 2022), emphasized multisensory experiences. The absence of a comprehensive integration strategy and comparison of the three sensory inputs can lead to a fragmented understanding of their synergistic potential and impact. Future research should strive to integrate and compare audio, visual, and olfactory stimuli in a cohesive manner. Studies could explore how these stimuli interact and influence mental health outcomes when used together, as opposed to in isolation or in paired combinations. This integrated approach can provide deeper insights into the multisensory experience, particularly in how olfactory stimuli complement and enhance visual and auditory experiences.

A critical limitation noted in studies, such as that by Zhang et al. (2023), is the exclusive focus on physiological indicators, neglecting psychological aspects. Conversely, research by Shin et al. (2022); Abbott and Diaz-Artiles (2022); Putrino et al. (2020); Sona et al. (2019); Serrano et al. (2016) have primarily relied on subjective analyses. This one-dimensional approach can result in an incomplete understanding of the study subject, as it fails to capture the holistic impact of the interventions or phenomena being studied. Future studies should aim to incorporate both objective (physiological) and subjective (psychological) measures. This dual approach would provide a more comprehensive understanding of the effects being studied. For instance, while physiological data can offer concrete evidence of bodily responses, psychological data can provide insights into personal experiences and perceptions. The integration of both types of data can lead to a more nuanced and complete understanding of the research findings.

In fact, several studies relied on a single objective measurement (e.g. (Amores et al., 2018; Hedblom et al., 2019; Zhao et al., 2022)), thus may have missed some important interactions between the central and nervous systems, which have been linked to different emotional states and behaviors. In turn, the work by Qi et al. (2022) relied on a single subjective measurement, which can result in a narrow understanding of participant experiences. Additionally, the studies by Lopes et al. (2022); Putrino et al. (2020); Amores et al. (2018) relied on single-item self-reports. Multi-item scales and the incorporation of a wider array of subjective tools should be used in future studies to capture a more comprehensive view of participant experiences and perspectives.

Studies, such as (Song et al., 2019; Zhang et al., 2023), were the only ones to analyze the stimuli and their effects at the individual level. As smell is very subjective (pleasant for some, unpleasant for others), performing analyses at the group level may average out some interesting individualized outcomes. Future research should highlight both individual and group analyses to provide deeper insights into how each stimulus uniquely contributes to the overall outcome, essential in sensory and psychological research.

Most studies offered a limited range of nature scenes, often focusing on a single type of environment (e.g., a forest). This limited range fails to consider the diverse preferences and responses individuals might have to different natural settings. Only a few works (e.g. (Serrano et al., 2016; Hedblom et al., 2019; Putrino et al., 2020; Qi et al., 2022)) offered different types of nature scenes. Future research should expand the range of environmental options presented to participants. Offering a diverse array of nature scenes, each with distinct characteristics (e.g., beach, mountain, urban green space), can cater to varied individual preferences and potentially yield richer data on environment-person interactions.

Personalization can play a crucial role in how individuals interact with and respond to different settings, impacting their psychological and physiological wellbeing. For example, Reid et al. (2015) demonstrated that odors can evoke emotional and nostalgic responses, enhancing the immersive quality of digital experiences. Incorporating options for personalizing the environmental experience can be a significant step forward in future research. Allowing participants to choose or even modify their environment according to their preferences can provide insights into how personalization affects their experience and response.

Moreover, as highlighted by Lopes et al. (2022), virtual experiments are often short in duration, in the order of a few minutes. Such short durations may limit the effectiveness of heart rate variability, which can often rely on tens of minutes. To capture more detailed biological features, extending experiment times to at least 10 min is recommended for more reliable and comprehensive physiological assessments.

Studies such as Amores et al. (2018) and Serrano et al. (2016) highlight a crucial limitation: the mismatch of sensory modalities. For instance, using lavender essential oil in a beach scenario can create sensory dissonance. This inconsistency can disrupt the coherence of the user experience, potentially diminishing the intended effect of the environment, whether it be for relaxation, stress reduction, or immersion. Future research should focus on creating sensory congruence in multisensory environments. This means ensuring that all sensory inputs are harmoniously aligned to reinforce each other.

Of the studies analyzed, only nine utilized objective measurements to quantitatively assess the effects of sensory stimuli. Objective measurement via wearables, for example, may allow for close to real-time monitoring of user affective states, as well as the development of biomarkers that could be used to quantitatively monitor intervention outcomes in the long term. Wearables have been shown to be a useful ally for stress management interventions (e.g. (Smith E. N. et al., 2020; Gomes et al., 2023)), thus should be considered in future studies. Identifying the most sensitive and informative wearable devices or measurements will be crucial for the understanding of the role of olfactory stimuli in digital nature bathing.

Moreover, only four of the nine studies relied on monitoring neural data while the user was immersed in the virtual environment (Amores et al., 2018; Song et al., 2019; Qi et al., 2022; Zhang et al., 2023). Unlike other wearable devices that are typically placed on wrists, chests, or arms, the collection of neural data, while the user is wearing an HMD, may be challenging. There are recent innovations in sensor-embedded HMDs that may help overcome this limitation (e.g. (Cassani et al., 2020; Bernal et al., 2022; Moinnereau et al., 2022)). Physiological measures derived from EEG can provide new insights into the immersive experience and its mental health impact. For instance, the works by Abbasi et al. (2019); Wu et al. (2023) showed the use of EEG to study emotional responses to olfactory stimuli. Future studies should consider the use of such devices to allow for real-time monitoring of mental/cognitive states, as well as the development of new neuromarkers of VR nature exposure outcomes.

While this review did not specifically focus on multisensory experiences relying on only haptic/tactile/somatosensory stimuli, three of the fourteen studies did include audio-visual-olfactory-tactile stimuli. Recent studies are showing that as more senses are stimulated, a greater sense of presence, immersion, and engagement can be achieved (De Jesus Jr et al., 2022; Gougeh et al., 2022), as well as improved neural plasticity (Amini Gougeh and Falk, 2023). Vibroacoustic therapy has shown to be useful for stress management (Boyd-Brewer, 2003). Additionally, the effects of thermal perception have been shown to have restorative benefits Lyu et al. (2022); Song et al. (2024). Future studies should explore the inclusion of tactile/somatosensory stimuli to further enhance the relaxation potential of the intervention.

Lastly, we attempted to quantify the quality of the reviewed studies via different quality checklists. While the majority was rated as moderate to high quality, two were deemed as low quality, which could have introduced biases and affected the overall reliability of the findings reported herein. Future works should aim to follow quality guidelines in (Tufanaru et al., 2020) to ensure outcomes are reliable.