Are Virtual Forests Just for Relaxation, or Can They Enhance the Benefits of Therapy?

Abstract

Forest bathing (Shinrin-Yoku in Japanese) is used as an intervention for improving mental health, with VR being used to create virtual forests for relaxation. Background/Objectives: In this research, we added therapeutic intent to a virtual forest with the goal of reducing social anxiety, with and without therapeutic instruction. Methods: Fifty-eight first-year psychology students were randomly assigned to one of three conditions: virtual forest only, therapeutic exercises only, and both combined. Results: All three conditions enhanced restorative effects equally. However, only the therapeutic exercise-only condition showed a tendency to reduce social anxiety. Participants in the combined condition reported more positive experiences and showed better comprehension of therapy content in the virtual forest. Conclusions: While the non-VR approach may offer immediate relaxation and possible anxiety reduction, combining the virtual forest with therapeutic exercises may yield better outcomes for sustained engagement and understanding over multiple therapeutic sessions.

1. Introduction

In a world where there is considerable psychological distress, coupled with limited mental health treatment resources, various technologies promise to potentially make mental health treatments more effective and readily accessible. One class of technologies seeks to create relaxed and mindful states that reduce stress and overall symptomology. A second class of technologies focuses on structured interventions that seek to address specific issues (such as motivational interviewing to reduce addictive behavior, and exposure therapy to treat various fears and anxieties). There has been little previous research on how to combine relaxing technologies with therapeutic instruction, and none that we could find on therapeutic instruction within a virtual forest. In this paper, we address this gap by looking at the impact of using therapeutic instructions in a virtual forest, focusing on social anxiety as the target mental health issue.

1.1. Social Anxiety

Social anxiety disorder (SAD) is a silent disorder usually masked as shyness or introversion, delaying diagnosis and treatment [1]. Individuals with social anxiety avoid speaking in public and expressing opinions due to a persistent fear of judgment and negative evaluations [2,3]. It is closely associated with early school dropout and subsequent depression risk [4].

SAD is influenced by a combination of genetic, temperamental, personality, and environmental factors, such as childhood experiences and peer interactions, which contribute to its development and persistence [5]. Early adverse experiences, such as childhood abuse, have been shown to dysregulate the HPA axis, leading to heightened cortisol reactivity and an increased susceptibility to social anxiety [6]. Additionally, SAD is often associated with deficits in emotion regulation, where individuals struggle to manage emotional responses to anxiety-inducing situations, further exacerbating the disorder [7,8]. Emotionally provocative stimuli, particularly negative facial expressions (e.g., anger and contempt), were found to elicit heightened neural responses in brain regions linked to emotion processing in socially anxious individuals, including increased activity in the insular cortex (specifically to angry faces), and the anterior cingulate cortex (faces showing disgust) [7,9,10]. These heightened responses may reflect an over-sensitivity to social cues of disapproval, a characteristic feature of SAD.

SAD disproportionately affects young people. It has an early onset (between 13 and 15 years old), often emerging in childhood or adolescence [11,12,13]. Approximately 9–14% of the young adult population in North America have SAD [14,15,16]. University students are also impacted, with approximately 10% exhibiting elevated social anxiety [17]. Notably, a study on 6825 participants aged 16–29 in various regions across Asia, North America, and South America found that over a third met the threshold for SAD [18]. Thus, social anxiety is pervasive, and its reduction would have a beneficial effect on large numbers of people.

1.2. Traditional Intervention to Reduce Social Anxiety

Traditional interventions for treating SAD are variants of cognitive behavioral therapy (CBT) and include applied relaxation, social skills training, and cognitive restructuring [19]. Applied relaxation focuses on progressive muscle relaxation, which is easy for individuals to practice during daily activity. Social skills training helps individuals with SAD to learn and practice social skills using techniques such as behavioral rehearsal, corrective feedback, and positive reinforcement. In cognitive restructuring, therapists work collaboratively with patients to identify and then contradict the often inaccurate, distressing, negative thoughts that tend to occur automatically as an effect of SAD.

Mindfulness-based interventions (MBIs) are alternative treatments for SAD [20,21,22,23]. MBIs encompass a range of psychotherapy techniques centered on developing non-judgmental present-moment awareness and engaging in meditative practices. Practicing mindfulness cultivates attention focusing, self-compassion, and attention shifting while reducing safety behaviors and cognitive distortions [21,24].

1.3. Access to SAD Treatment

In a study involving people who reported anxiety or depression (N = 1956), researchers found that 23.6% of the participants had been diagnosed with SAD by healthcare professionals, but only 40% of them were receiving pharmacotherapy or psychotherapy [25]. Another study reported that over 80% of individuals with SAD receive no treatment, while those who did seek treatment only sought help after 15 to 20 years of experiencing symptoms [13].

Barriers to seeking treatment include social interaction fears, shame, stigma, limited information, and financial cost [11,26,27]. Additionally, lengthy appointment wait times and sparse service availability impede access [28,29]. Barriers to treatment are negatively correlated with age, indicating that younger individuals perceive more hindrances [27]. In educational settings, children and adolescents with SAD are prone to be the targets of bullying and teasing, while educators may know little about how to recognize SAD and manage the condition [29,30].

Recent advances in digital therapy and virtual treatment for SAD have increased access to treatment [31]. However, among 1154 online mental health applications, just 38 applications were relevant to SAD, with only 5 focused on symptom management through techniques like relaxation and cognitive bias modification [32]. Well-designed applications can provide various forms of training (in areas such as social skills, cognitive restructuring, mindfulness practices, and exposure therapy), allowing people with mild forms of disorders (not requiring diagnosis and clinical treatment) to receive assistance. However, rigorous studies with randomized-control trials and larger samples are needed to demonstrate the validity of applications [31].

1.4. Real and Virtual Nature

With advancements in virtual reality (VR), researchers have explored the potential of virtual nature experiences to provide similar therapeutic benefits to real nature experiences, particularly for individuals in urban environments with limited access to natural settings. The biophilia hypothesis suggests an inherent biological connection between humans and nature, with natural environments enhancing emotional well-being [33]. Gaekwad et al. found that exposure to nature improves affective states more than urban environments, increasing positive emotions and decreasing negative ones.

Virtual nature environments can replicate these effects, offering similar benefits for those unable to access real nature. Studies have shown that exposure to virtual nature can reduce stress, anxiety, and negative emotions while fostering positive psychological states such as increased vitality and relaxation [34,35,36]. A systematic review of 21 studies on virtual nature’s psychological effects found consistent emotional recovery, with virtual nature more effective at reducing negative emotions than boosting positive ones [37]. Another study assessed the impact of daily exposure to virtual nature on college students. It found decreased anxious arousal (panic) and anxious apprehension (worry) after 3–4 weeks of exposure [38].

Additionally, physiological benefits—including decreased heart rate, lower cortisol levels, and reduced blood pressure—have been observed, supporting the potential of VR nature therapy as a viable intervention for mental wellness [39,40,41]. Despite growing interest in VR-based nature therapy, a systematic review highlighted the need for more rigorous research, as many existing studies rely on small sample sizes and lack high-quality experimental designs [42], a problem that has continued up to the time of this writing.

Researchers have examined various virtual nature environments such as forests, oceans, and other natural landscapes. A study assessing 16 different nature environments found that the perceived relaxation scores from sunset beaches, ocean waves, and green forests were the highest [43]. However, some participants commented that they felt unsafe due to the openness of the sea.

Shinrin-Yoku, or forest bathing, originated in Japan in the 1980s as a practice of immersing oneself mindfully in nature using multiple senses to promote relaxation and stress reduction [44,45]. Forest bathing involves engaging with the forest atmosphere without structured guidance and has been shown to provide psychological and physiological benefits [44]. Over time, this practice evolved into forest therapy with structured sessions led by trained professionals, incorporating techniques such as guided meditation, walking, and cognitive behavioral therapy (CBT) to enhance well-being [46,47]. Research supports both structured and unstructured approaches to forest therapy, with researchers striving for well-researched mechanisms that facilitate mental and physical health improvements [44,45,46,47].

Researchers have designed immersive virtual forests that allow users to navigate freely and explore interactive elements within the virtual environment [48,49,50]. For instance, Mohamad Yahaya et al. [49] integrated breathing exercises and stretching activities at checkpoints within the virtual forest, as well as two tasks (collecting red balls and taking photographs). They found that higher-quality audio-visual elements contributed to a stronger sense of presence and spatial awareness, thereby improving the user experience. While their study suggested that tasks may be successfully implemented in a virtual forest without reducing their beneficial effects, the tasks were simple, and only eight participants were used in their study.

Adding interactive features to a virtual nature environment may enhance mental well-being. Wang et al. investigated the intervention effects of urban vs. park environments, with interactive activities such as flying a kite, enjoying the idyllic scenery, fishing, and watering plants [51]. The experiment was conducted once a week over four sessions with 195 participants having mild to moderate anxiety and depression. Both groups (urban and park) experienced a reduction in anxiety and depression.

However, the integration of psychotherapy techniques with interactive features into virtual nature applications remains an emerging research area. One study examined an e-mental health VR application that combined Mindfulness-Based Cognitive Therapy (MBCT) with nature-based activities, such as focusing on a sprouting flower [52]. However, due to a small sample size of 30 participants, the study did not yield significant results, and the study did not consider the impact of using therapy in a virtual forest. Cognitive overload from various sensory stimuli (e.g., visual, auditory, etc.), including potential cybersickness, may have hindered participants’ ability to focus on the mindfulness practice, leading to non-significant results. Cognitive load, related to limited working memory capacity, can reduce learning effectiveness when excessive information overwhelms the working memory system [53]. Factors like task complexity, interaction levels, avatar representation, and cybersickness can all influence cognitive load and learning performance in VR [54,55]. Given the mixed results thus far, and the paucity of research targeting the use of therapeutic instruction within virtual forests, further research is needed to see if therapeutic instruction might enhance, or interfere with, the beneficial properties of virtual forests.

In response to this research gap, we developed a VR application aimed at addressing mild social anxiety by integrating evidence-based therapeutic techniques within a virtual forest setting. Users navigated a virtual forest that was intended to be calming while receiving guided therapeutic instructions from a virtual therapist. These instructions incorporated elements of CBT, meditation, and mindfulness practices, equipping users with coping strategies to manage social anxiety. Our application was designed for individuals who might have been reluctant to engage in traditional in-person counseling.

1.5. Research Questions and Hypotheses

The following research questions were addressed in our study:

• Does the use of a virtual forest setting (with or without therapeutic instruction) reduce social anxiety (and improve self-rated restoration effects) to a greater extent than therapeutic instruction provided without a virtual forest?
• Is therapeutic instruction (either by itself or in a virtual forest setting) more effective at reducing social anxiety (and improving self-rated restoration effects) than virtual forest bathing (without therapeutic instruction)?
• Does the addition of therapeutic instructions make a virtual forest less attractive (measured in terms of user experience, immersiveness, and sentiment) than it would otherwise be?

2. Materials and Methods

2.1. Participants

A total of 58 students (18–25 years old; women = 36, men = 22) were recruited from a first-year psychology course at the University of Toronto and participated in this study. The students received course credit for their participation. Participants were randomly assigned to three different conditions without any pre-screening for baseline social anxiety levels or VR familiarity. Due to ethical requirements, participants were not diagnosed with SAD, and all participants had minor or mild social anxiety rather than clinical levels of anxiety. There were three conditions: Condition A (VR forest only), Condition B (combined VR forest and therapy exercises), and Condition Non-VR (non-VR, i.e., therapy exercises only). Participants from Conditions B and Non-VR needed to participate in two sessions to complete all the checkpoints in their condition. Participants in Condition A only needed to participate in one session as they could complete the walk through the forest in 30 min.

Participants in Condition B and the Non-VR condition had the option of completing only the first session or of also completing a second session (with a minimum of five days between sessions to increase the amount of time that any effect of reducing social anxiety might have) for additional course credit (one credit per session). A total of 9 of the 40 students in the two relevant conditions continued their participation in a second session (18–19 years old; women = 5, men = 4; B = 5, Non-VR = 4, see

Table 1. Number of participants in each condition and session.

Session	A	B	Non-VR
1	18	22	18
2		5	4

). The attrition rate for both Condition B and Condition Non-VR is around 0.77. A chi-square test of independence found no significant difference in attrition between Condition B and the Non-VR condition, suggesting attrition was similar across both conditions and that there was no significant attrition bias. Relatively few participants completed the second session because most participants had either completed the experimental course credits that they required after the first session or wanted to try out other experiments provided by different research labs.

2.2. Materials and Apparatus

The materials and apparatus used in this experiment consisted of (1) self-reported questionnaires assessing social anxiety, restoration effects, usability, and immersiveness; (2) Meta Oculus Quest 2 VR headsets and virtual forest application; and (3) therapy instructions.

2.2.1. Psychological Measurements

The Social Interaction Anxiety Scale (SIAS) is a self-report 20-item questionnaire rated on a 5-point Likert scale that evaluates the stress experienced when interacting with others [56]. Researchers have reported good test–retest reliability with an intraclass correlation coefficient of 0.92 over a 4-week interval, and internal consistency with Cronbach’s α = 0.94 when applied to a wide range of people, including graduate students, community volunteers, and patients with SAD [56,57].

We also used a 6-item Restoration Outcome Scale (ROS) to assess a more sensitive and temporary state, i.e., restoration experienced after visiting favorite places, which consisted of relaxation, calmness, attention restoration, and clearing thoughts [58,59,60]. ROS is commonly used to assess forest-related experience in many studies, with Cronbach’s α ranging between 0.87 and 0.92 [59,61,62].

2.2.2. VR Usability Measurements

We used the Virtual Reality Neuroscience Questionnaire (VRNQ) to assess the quality of VR software in terms of user experience, in-game assistance, and VR-induced symptoms and effects [63]. The Cronbach’s α values for each domain of VRNQ were 0.89 for user experience, 0.90 for in-game assistance, and 0,89 for VRISE [63]. We removed 3 of the 20 questions in this questionnaire because they were not relevant to our study (the 3 questions asked about non-existent features in our VR forest: two-handed interaction, how things were picked up, and how easy it was to use items), leaving 17 items that assessed user experience in VR.

After completing the questionnaires, participants were asked about their experience in using the virtual forest (“Please describe your overall experience in the virtual forest. What aspects of the virtual environment, interactions, or content, if any, stood out to you or influenced your experience positively or negatively?”).

2.2.3. VR Headset and Software

Participants engaged in the virtual forest using the Meta Oculus Quest 2 VR headset. The virtual forest application was designed and implemented using Unity (2020.3.26f1 version) by a team of developers from the Interactive Media Lab, University of Toronto. Assets, including forest environment (NatureManufacture’s Forest Environment), animals (WDALLGRAPHICS and PROTOFACTOR, INC), and architectural buildings (StarGames studio) were bought from the Unity Asset Store.

There were eight checkpoints in the forest, and participants navigated from the start to the destination via a path through the forest (a total of 30 min of walking from the start to the final destination). The path is indicated by a wavy red line on the map shown in the figure, and the checkpoints are represented in the map view as yellow dots (as shown in Figure 1a).

The virtual forest was filled with tall coniferous, and deciduous, trees. The ground was covered with long grass, flowers, and rocks, and occasionally small animals would appear such as deer, squirrels, and foxes, and those animals would move around in the forest. There was also a river with flowing water and the scene portrayed a bright day with partly cloudy skies, adding to the immersive atmosphere of the setting. Tree density varied within the forest, but in general, the sky was quite visible, and the overhead forest canopy was limited (Figure 1b).

In addition to the map that could be accessed at any time with a button press, there were also wooden road signs in the forest pointing the way to the next checkpoint. Each checkpoint had a different architectural building that participants could pause at, or explore, before moving on to the next checkpoint. In Condition B (virtual forest combined with therapy instructions), participants had to stay at the checkpoint area and listen to the audio instructions on techniques to reduce social anxiety (given by the virtual therapist). Links to the video demonstrations of the virtual forest can be found in the Supplementary Materials section.

2.2.4. Experimental Conditions

The following conditions were used in the experiment.

• Condition A—VR forest only: In this VR condition, participants had to navigate through a virtual forest using controllers and joysticks without receiving any therapeutic exercises.
• Condition B—combined VR forest and therapy exercises: Similar to Condition A, participants in this VR condition still navigated through the same virtual forest, but they had to pause their virtual walk and listen to therapeutic exercise instructions that were designed to reduce social anxiety at each checkpoint.
• Condition Non-VR—therapy exercises only: In this Non-VR condition, participants received therapeutic exercise instructions displayed on a laptop, without exploring or walking in the virtual forest. The exercise instructions, identical in content and design to Condition B, were shown on a lab-developed website, accompanied by the same forest sounds used in Conditions A and B. Before the session started, participants were asked to imagine that they were in a forest.

2.2.5. Therapy Instructions

As mentioned in Section 2.2.4, participants from Conditions B and Non-VR had to listen to the virtual therapist’s instructions at each checkpoint. The therapeutic instruction included psychoeducation, meditation exercises, and quizzes that assessed participants’ learning [64]. Materials, from the social anxiety treatment manuals developed by Andrews [65], mindfulness-based stress reduction by Kabat-Zinn [66], and mindfulness-based cognitive therapy developed by Segal et al. [67], were incorporated and modified into our therapeutic instructions on reducing social anxiety and enhancing overall mental wellness. The complete psychotherapy instruction script can be found in Appendix A.

An overview of the 8 checkpoints’ instruction content is provided below:

• Checkpoint 1: In accordance with a typical cognitive behavior therapy first session’s content, this checkpoint focused on psychoeducation [65]. The psychoeducation content was developed based on Andrews’ treatment manual (Pages 3–18). The psychoeducation part was followed by a quick abdominal breathing exercise (from Andrews’ manual) teaching people how to slow down the rate of respiration to cope with shallow and rapid breathing when anxiety level is high.
• Checkpoint 2: This checkpoint introduced the concept of mindfulness, highlighting the concept of focusing on the present moment and becoming aware of thoughts and feelings based on Segal et al.’s mindfulness-based cognitive therapy manual [67]. To help cultivate awareness, participants were asked to engage in the four-senses mindfulness exercise (the five-sense exercise minus taste) to bring awareness to each sense [68].
• Checkpoint 3: We used both Andrews and Segal et al.’s manuals to develop this section of cognitive restructuring [65]. Challenging thoughts and cognitive restructuring are fundamental components of CBT. Participants also received two hypothetical situations, asking the participant to become aware of the thoughts that those situations evoke, to evaluate whether those thoughts are reasonable assessments, and to reappraise the situation if the thoughts were not helpful.
• Checkpoint 4: The fourth checkpoint exercise was the mountain mediation developed by Kabat-Zinn [69].
• Checkpoint 5: Using Andrews’ manual (Pages 33–34), the material in this checkpoint discussed avoidance and its negative impact on social anxiety [65]. While avoidance might seem helpful, it often exacerbates anxiety and prevents personal growth.
• Checkpoint 6: The sixth checkpoint exercise was sound meditation, adapted from Segal et al. to engage participants with forest sound played in the VR headset [67].
• Checkpoint 7: The seventh checkpoint exercise was about self-compassion, which incorporates Hofmann et al.’s Loving-kindness Meditation [70]. It discusses the inevitability of setbacks in conquering social anxiety and emphasizes the importance of treating oneself kindly when encountering those difficult times.
• Checkpoint 8: The eighth checkpoint was similar to a closure session, which guides participants to reflect on what they have learned and how they can apply it in the real world, followed by a simple breathing meditation, Three Minute Breathing Space [66,67].

2.3. Procedure

The study was posted on the first-year psychology course’s participant pool system at the University of Toronto in the winter 2024 semester. Eligible participants for enrolment had to be 18 years old or above and not clinically diagnosed with social anxiety disorder. The experiments were carried out in person at the Interactive Media Lab, University of Toronto.

Participants used the provided laptops to complete the consent form, fill in demographic information, and answer a set of questionnaires assessing social anxiety levels and restoration effects. Participants were randomly assigned to one of the three conditions. After watching a three-minute tutorial video on how to complete the VR/Non-VR task, participants were provided with either the Meta Oculus Quest 2 VR headset or the laptop displaying the Non-VR website to start the task. There were two rounds of VR tasks, with a 5 min break between them. Each round lasted about 15 min. Participants were also asked to rate their motion sickness and discomfort levels during the break. After completing the second round, participants were asked to complete another set of questionnaires on user experience, social anxiety levels, and restoration effects as the final step of the experiment. As many participants might not be able to complete all 8 checkpoints in the first session, participants could decide whether they wanted to participate in the second session to complete the whole program. The total time to finish one session was approximately 50 min.

In the second session, the SIAS was administered at the end of the session so that social anxiety at that point could be compared with the baseline social anxiety prior to the first experimental session.

2.4. Data Analysis

Data analyses were carried out using R (Version 4.2.1), R Studio (Version 2023.06.0+421), and Python (Version 3.12) to evaluate the effects of different experimental conditions on participants’ social anxiety, quiz performance, user experience, and restoration outcomes.

We began by conducting a principal factor analysis on the Social Interaction Anxiety Scale (SIAS) with Varimax rotation. We chose the number of factors to select based on a scree plot. Item reliability analysis was then used to assess the internal consistency of the selected factor, calculating Cronbach’s alpha for both pre- and post-session data. We then iteratively removed items that did not contribute significantly to the overall reliability, retaining a set of items that had high internal consistency. The average of the SIAS factor items was calculated to develop a scale rating consistent with the summated rating scale method described by Spector [71]. The differences in the resulting social anxiety scale ratings, between pre- and post-session measures, were then analyzed for each condition. This analysis was visualized using Cumming plots, highlighting changes in social anxiety across the three conditions.

In addition to analyzing social anxiety, we conducted a Fisher’s exact test to evaluate differences in quiz accuracies across two conditions (B and Non-VR) in three checkpoints to further assess participant’s understanding of the therapy instructions. User experience was assessed using one-way ANOVA and post hoc Tukey tests, comparing overall quality and enjoyment ratings (from the Virtual Reality Neuroscience Questionnaire (VRNQ)) across conditions. To examine restoration outcomes, a mixed ANOVA was conducted to assess the effects of conditions and phases on restoration ratings (calculated as the average of the six items from the Restoration Outcome Scale).

Furthermore, we analyzed participants’ open-ended comments using NLTK’s Sentiment Intensity Analyzer to compute sentiment scores, which ranged from −1 (most negative) to +1 (most positive). These sentiment scores were then evaluated using one-way ANOVA and independent t-tests to explore differences in sentiment across conditions.

3. Results

3.1. Social Anxiety Measurement

3.1.1. Factor and Reliability Analysis

We conducted a factor analysis to identify the underlying constructs for the 20-item Social Interaction Anxiety Scale (SIAS). The KMO measure was 0.82, indicating sampling adequacy. Bartlett’s test of sphericity was significant, χ²(190) = 569.63, p < 0.001, suggesting the factorability of the correlation matrix. The scree plot suggested only one factor (Figure 2). Varimax rotation was applied to achieve a clearer structure.

Table 2. Factor loadings of the SIAS 20 items.

SIAS Items	Factor Loadings
1. I get nervous if I have to speak with someone in authority (teacher, boss, etc.).	0.637
2. I have difficulty making eye contact with others.	0.509
3. I become tense if I have to talk about myself or my feelings.	0.472
4. I find it difficult to mix comfortably with the people I work with.	0.803
5. I find it difficult to make friends my own age.	0.359
6. I tense up if I meet an acquaintance in the street.	0.606
7. When mixing socially, I am uncomfortable.	0.768
8. I feel tense if I am alone with just one other person.	0.380
9. I have difficulty meeting people at parties, etc.	0.371
10. I have difficulty talking with other people.	0.707
11. I find it difficult to think of things to talk about.	0.433
12. I worry about expressing myself in case I appear awkward.	0.749
13. I find it difficult to disagree with another’s point of view.	0.529
14. I have difficulty talking to people I am attracted to.	0.184
15. I find myself worrying that I won’t know what to say in social situations.	0.722
16. I am nervous mixing with people I don’t know well.	0.692
17. I feel I’ll say something embarrassing when talking.	0.814
18. When mixing in a group, I find myself worrying I will be ignored.	0.601
19. I am tense mixing in a group.	0.771
20. I am unsure whether to greet someone I know only slightly.	0.539

Note. Items above are from the Social Interaction Anxiety Scale (SIAS) [56].

shows the factor loadings of the 20 items, in which Item 14, “I have difficulty talking to people I am attracted to”, had the lowest loading with 0.184, suggesting a poor contribution to the factor. All the other loadings were above 0.3.

Item reliability analysis was conducted to identify a parsimonious representation of the social anxiety factor while maintaining strong internal consistency. We began with all 20 SIAS items using post-session 1 data. Items were sequentially removed based on their impact on internal consistency, retaining only those that did not lead to a large drop in Cronbach’s alpha. This led to the inclusion of five items (7, 12, 15, 17, and 19). This five-item scale demonstrated strong internal consistency, with Cronbach’s alpha values of 0.87 for the pre-session data and 0.93 for the post-session data (

Table 3. Five-item scale retained after item reliability analysis.

Phase and Cronbach’s Alpha	SIAS Items
S1 Pre = 0.87 S1 Post = 0.93	7. When mixing socially, I am uncomfortable. 12. I worry about expressing myself in case I appear awkward. 15. I find myself worrying that I won’t know what to say in social situations. 17. I feel I’ll say something embarrassing when talking. 19. I am tense mixing in a group.

).

Peters et al. developed a short form of Social Interaction Anxiety with six items [72]. We also applied Peters’ six items and calculated the Cronbach’s alpha for our pre- and post-session data. The internal reliability for the six-item Peters scale was 0.76 for the pre-session data and 0.81 for the post-session data, both of which were well below the Cronbach’s alpha measures of internal consistency that we obtained for our five-item scale. (0.87 and 0.93), in the pre- and post-sessions, respectively. Thus, we chose to use our five-item version of the SIAS as the short version of the SIAS rather than the Peters scale.

3.1.2. Social Anxiety Change Across Three Conditions

We derived a social anxiety scale by calculating the average of the five items from Section 3.1.1. We then subtracted the post-session social anxiety from the pre-session social anxiety, with the resulting differences in social anxiety across the three conditions in Session 1 shown in the Cumming plot in Figure 3. Condition Non-VR (therapy only) showed a greater (but statistically non-significant) drop in social anxiety than conditions A (VR forest Only) and B (combined VR forest and therapy), suggesting therapy-only instructions reduced social anxiety more, possibly because participants in that condition were not distracted by the VR environment and focused more on the therapy content. The virtual forest may make people feel relaxed, but therapeutic instruction may be needed to reduce social anxiety.

The nine participants (five in the combined condition and four in the Non-VR condition) who stayed in both sessions (only participants in Conditions B and Non-VR returned for the second session) were too few for statistical analysis. However, the corresponding Cumming plot in Figure 4 again shows a tendency for Non-VR (i.e., therapy only) participants to report a greater drop in social anxiety (vs. the combined condition).

3.2. Checkpoint Quiz Performance

We applied Fisher’s exact test to analyze participants’ performance between Condition B and the Non-VR condition in the first session’s three checkpoints (see

Table 4. Contingency Table of Participant Accuracy in the Three Checkpoints.

Checkpoint	Condition	Correct	Incorrect
1	B	41	1
	Non-VR	35	1
2	B	19	2
	Non-VR	17	1
3	B	40	2
	Non-VR	12	18

Note. These data were collected from a total of 21 participants in Condition B and 18 participants in Condition Non-VR; Checkpoints 1 and 3 have two questions. Data from one participant in Condition B across all three sessions and three participants in Condition Non-VR in Checkpoint 3 were lost due to technical issues.

). The results did not differ between the two conditions in the first two checkpoints (Checkpoint 1: p = 1; Checkpoint 2: p = 1). However, the result indicated a significant difference in Checkpoint 3’s accuracy between the two conditions, p < 0.001, odds ratio = 0.04, 95% CI [0.01, 0.18], and Cramér’s V = 0.58. More participants in Condition B answered correctly compared to those in Condition Non-VR, suggesting Condition B helped participants sustain a more accurate understanding of the therapy content over time.

3.3. User Experience Questionnaire

The participants in the VR conditions assigned higher ratings on user experience, suggesting that VR conditions were preferred to the Non-VR condition. For the Session 1 data, one-way ANOVA showed a significant effect of condition on the user experience overall quality rating, F(2, 55) = 5.40, p < 0.01, η² = 0.16. Post hoc Tukey tests revealed that Condition Non-VR (M = 4.00, SD = 1.19) had significantly lower overall quality ratings compared to Conditions A (M = 5.17, SD = 1.25, p < 0.05, Cohen’s d = 0.96, 95% CI [0.24, 1.67]) and B (M = 5.00, SD = 1.07, p < 0.05, Cohen’s d = 0.89, 95% CI [0.22, 1.56]), as shown in Figure 5a. No differences between the three conditions were found for sound, graphics, immersion, and enjoyment ratings in Session 1.

While only a few participants continued to Session 2, one-way ANOVA revealed a significant effect of condition on enjoyment for those that did, F(1, 7) = 13.37, p < 0.01, η² = 0.66. Post hoc Tukey tests revealed that Condition Non-VR (M = 4.50, SD = 1.00, N = 4) had significantly lower enjoyment ratings compared to Condition B (M = 6.40, SD = 0.55, N = 5), p < 0.01, Cohen’s d = 2.45, 95% CI [0.36, 4.55], as shown in Figure 5b. No differences between the two conditions were found for sound, graphics, immersion, and overall quality ratings in Session 2. Thus, even though Condition Non-VR showed the greatest reduction in social anxiety in Session 1, we hypothesize that the greater user experience of the two VR conditions in Session 1, and the greater enjoyment of the two VR conditions in Session 2, suggest that VR may help to sustain engagement over multiple sessions.

3.4. Open-Ended Comments

We were interested in whether the presence of VR had an impact on sentiment. Would people prefer to be in the VR forest? We ran an independent samples t-test to examine the effect of VR vs. Non-VR groups on the sentiment score of participants’ comments in the first session. There was a borderline significant difference in ratings between the Non-VR (M = 0.38, SD = 0.46) and the VR (M = 0.58, SD = 0.34) conditions, t(56) = −1.87, p = 0.067, 95% CI [−0.42, 0.01], Cohen’s d = 0.53, 95% CI [−1.09, 0.04]. This suggested a greater positivity in the two VR conditions versus the Non-VR (therapy only) Condition. Participants may have perceived the experience as more enjoyable, engaging, or rewarding compared to Condition Non-VR. As shown in Figure 6 and

Table 5. Top and bottom sentiment score comments.

Condition	Comments	Sentiment Score
Comments with highest three sentiment scores across three conditions
A	I found the sounds and relaxing graphics to be of great stress relief. i also found that it was easy to become immersed into and feel a great sense of calm. The structures were nice as they were all different and invoked a sense of interest and wonder. I enjoyed the experience thoroughly.	0.98
B	I like how there are cute animals included in there, but it’ll be better if they are all pushable (pushing animals is satisfying idk why). Also, when I was exploring the river I was expecting something under the surface (maybe fish) but turns out there’s none, which makes me feel kind of disappointed… But overall the experience is good, the forest is pretty and the sound effect is really realistic. I’ll come back if I have time. Thank you for the experiment!	0.95
Non-VR	I enjoyed the modules, very helpful, thank you.	0.85
Comments with lowest three sentiment scores across three conditions
A	I loved the environment that was created with the headset and I found myself immersed in the VR world but the feeling of nausea I had took me out of the experience and was the only thing that made me uncomfortable and what I didn’t like.	−0.33
B	I liked some of the images like the clouds, river, and flowers, but i also found it a bit distracting since i did feel i was in a virtual reality that was not real. It also made me nauseous, which made it harder to concentrate.	−0.04
Non-VR	Great experience but for module 4, I found it very boring and long so I may have lost focus there.	−0.64

, participants gave slightly more positive sentiment scores in the VR conditions compared to the Non-VR condition.

We also counted the word occurrence on all the words related to immersiveness (either expressed in single words or in phrases), such as ‘immersive,’ ‘realistic,’ and ‘made me feel like I am actually in the forest’ (see Appendix B for complete word list) from each comment. One-way ANOVA revealed significant differences in word occurrences across the conditions in Session 1, F(2, 55) = 4.39, p < 0.05, η² = 0.14 (see Figure 7). Post hoc Tukey tests revealed that Condition A (M = 0.67, SD = 0.84) had significantly higher immersiveness wording than Condition Non-VR (M = 0.11, SD = 0.32), p < 0.05, Cohen’s d = 0.87, 95% CI [0.16, 1.58]. Participants in Condition A (virtual forest only) reported a greater sense of immersiveness compared to Condition Non-VR condition with therapeutic instructions only. Condition B, which combined the forest with therapeutic instructions, showed no significant difference from either Condition A or Non-VR, as it fell between the two conditions. The therapeutic instructions in Condition B may have distracted participants from fully engaging with the immersive elements, resulting in lower immersiveness than Condition A but still higher than the Non-VR condition, where no virtual environment was present.

3.5. Restoration Outcome Scale

We applied mixed ANOVA to examine the effects of conditions and phases on restoration rating in Session 1. There was a significant main effect of phases, F(1, 109) = 27.81, p < 0.0001, η² = 0.20 (see Figure 8). Participants from all three conditions showed a significant increase in restoration rating from pre-experiment (M = 3.21, SD = 0.94) to post-experiment (M = 4.34, SD = 1.32), Cohen’s d = 0.98, 95% CI [0.59, 1.38]. This suggested both the virtual forest shown in the two VR conditions, as well as the forest sounds and forest picture in the background in Condition Non-VR, can invoke similar levels of increased restoration.

4. Discussion

4.1. Non-VR (Therapy Only) Condition Shows Potential for Reducing Social Anxiety

Only the Non-VR therapy showed a tendency to reduce social anxiety levels. Social anxiety is often deeply ingrained in individuals, and it typically requires multiple sessions over a longer period rather than the single-session intervention that most participants underwent in this study. Thus, it is not surprising that statistically significant reductions in social anxiety were not found. In contrast, scores on the ROS increased significantly after the intervention for all three experimental conditions. Consistent with the finding for social anxiety, the increase in ROS from pre- to post- tended to be greater in Condition Non-VR (see Figure 8 in Section 3.5.). A similar trend was found in a real-world setting by Korcz et al. [73], where informal education in a forest setting contributed to greater psychological restoration compared to merely walking in the forest without educational guidance. Their study highlighted the importance of incorporating structured educational tools into forest exposure to enhance psychological well-being.

Condition Non-VR might have produced better results because participants could focus on the therapy instructions without the distractions present in the VR environments (where participants could explore the forest environment). The virtual forest contained many features designed to increase engagement and the quality of the user experience (such as the checkpoint building, animals, forest sounds, and various plants and trees), but this may have also distracted participants from attending fully to the therapeutic instructions.

4.2. VR Therapy Enhances Long-Term Engagement and Learning of Therapeutic Information

The quiz performance data revealed that participants in Condition B obtained higher accuracy in Checkpoint 3 (assessing the comprehension of the therapy content) relative to those in Condition Non-VR (see Section 3.2.). Checkpoint 3 occurred after roughly 15 min of interaction task.

On the other hand, sentiment analysis of text comments made after the session showed that sentiment tended to be more positive in the VR conditions vs. therapy only (p = 0.067). While this effect was borderline significant, the trend suggests that participants might have felt more positively toward the virtual forest experience, which could have important implications for long-term engagement and treatment adherence. Participants in the VR conditions may have perceived the session as more enjoyable or engaging, potentially increasing their willingness to participate in future therapy sessions. Since treatment adherence is a major challenge in anxiety interventions, virtual forest creating a positive emotional experience in therapy settings could be an important factor in sustaining engagement over time.

Our examination of immersiveness also showed that immersiveness (as indicated by the occurrence of words related to immersiveness in participant comments) was lower in the therapy-only condition. Performance in Checkpoint 3 may have suffered in the therapy-only condition because the lack of immersion and engagement may have made the task more boring. Thus, combining therapy with a virtual forest environment may facilitate more accurate comprehension of therapy content over time, in spite of any distracting effects of the virtual forest.

These findings align with prior research suggesting that immersive virtual nature, when combined with mindfulness practices, can enhance task performance over time. Ch et al. examined the effects of virtual nature combined with audio-guided mindfulness exercises (e.g., breathing and awareness exercises) on remote workers’ creativity over a nine-week period. Their study found that participants demonstrated improved performance on convergent thinking tasks, including a higher number of correct answers, faster response times, and better focus [74]. Similarly, prior research has shown that virtual reality environments can effectively stimulate learners’ interest and enhance engagement [75,76].

There was a tendency towards more reduction in social anxiety in Condition Non-VR. However, the better quiz performance, and ratings on two of the user experience items, in Condition B suggest that VR combined with therapy may offer long-term benefits for therapeutic intervention. The immersive and interactive forest elements may help to sustain engagement and promote deeper comprehension of therapy content over successive therapeutic sessions.

Thus, while Condition Non-VR may tend to reduce social anxiety in a single session, VR conditions might provide better outcomes for long-term learning and engagement, leading to stronger therapeutic benefits overall. An effective VR therapy application would need to optimize the balance between the therapeutic content and immersive forest environment to maintain users’ interest, reduce boredom, and promote sustained learning with an accompanying reduction in social anxiety. This balance is crucial, as overly immersive environments can sometimes lead to cognitive and sensory overload, potentially hindering therapeutic effectiveness [77,78]. We recommend that future research focus on integrating elements that combine the benefits of both approaches, perhaps by incorporating periods of focused therapy content followed by immersive, engaging activities to optimize both relaxation and learning outcomes.

4.3. Forest Sound Played a Significant Role in Enhancing Restoration

Self-rated ROS increased after all three conditions in the study as reported in the mixed ANOVA reported in Section 3.5. All three conditions utilized the same background audio, which were forest sounds, including birds chirping, a river flowing, and wind rustling. Many participants commented that the forest sound made the experience more vivid and calming (see

Table 6. Participants’ comments on forest sounds.

Condition	Comments
A	“sound was very calming”
	“I really enjoyed the sounds and the graphics”
	“The sound and visual images made me feel like I am actually in the forest, so I felt calm.”
	“I feel like the visuals and the sounds were realistic enough to make my experience similar to a real-life experience.”
	“I think the environment that it built was really nice, but the soundtrack was the part that made it more alive.”
B	“An aspect that particularly stood out to me was the sound quality, which was really good (both the forest background noises and the checkpoint narrator’s voice).”
	“The aspect of the application that stood out to me was the sound quality. There was a lot of richness to the forest sounds, and the sound quality of the male narrator’s voice was also really good. It felt like it was coming from everywhere and nowhere at once.”
	“The forest is pretty and the sound effect is really realistic.”
Non-VR	“The background sounds were very relaxing and helped me visualize that I was in a forest.”
	“The sound effects were quite vivid which help immerse quickly.”

). This aligns with past research showing that exposure to forest sounds can reduce negative emotions and promote relaxation [79,80]. Moreover, Annerstedt et al. emphasize that auditory stimuli play a crucial role in enhancing the recovery aspect of VR-based nature experiences, demonstrating that physiological stress recovery is significantly improved when visual and auditory nature cues are combined. Their findings reinforce the idea that natural soundscapes are not merely supplementary but essential in fostering an effective immersive therapeutic environment [81].

Participants in Condition Non-VR were instructed to imagine being immersed in a forest while viewing a screenshot of the virtual forest as the interface’s background image. The combination of forest sounds and the imagined forest environment produced a restoration effect comparable to that observed in the VR conditions. In Table 6, one participant from Condition Non-VR used words like vivid and immersive. This suggests that the cognitive process of imagining an immersive natural environment, when paired with concordant auditory stimuli, can be as effective in promoting restoration in simulated or imagined virtual experiences. Prior research has similarly shown that guided imagery combined with nature sounds can evoke relaxation responses, reduce anxiety, and manage pain [82,83]. The equivalence in restorative effects between the VR and Non-VR conditions suggests that a cheaper and simpler Non-VR implementation could provide similar benefits in making therapeutic interventions more engaging through guided imagery and natural soundscapes.

5. Limitations and Future Research

While our study provides insights into the usability and effectiveness of combining a virtual forest with therapeutic exercises addressing social anxiety in young adults, there were several limitations. One major limitation was our sample size and the high dropout rate. Although the study initially included fifty-eight first-year psychology students with approximately twenty participants per condition, only nine participants remained for the second session. The high dropout rate limits our ability to draw robust conclusions about the longitudinal changes in the effectiveness and usability of our interventions.

Additionally, the duration and frequency of sessions were limited. Social anxiety is a deeply ingrained trait, and reductions in self-rated anxiety within a single session or even over two sessions across two weeks should probably not be expected in most participants. Typically, therapeutic interventions for social anxiety require about 14–20 weekly sessions to observe meaningful changes [84,85]. The short duration of our intervention, with a maximum of two sessions, limits the generalizability of our findings, as social anxiety therapy typically requires a more extended intervention period. Moreover, our study primarily focused on short-term effects, without assessing the long-term impact of VR therapy on social anxiety reduction. Our study’s short duration and limited number of sessions likely contributed to the minimal observed effects on social anxiety. Future research should implement extended intervention periods and follow-up assessments to examine the sustained effectiveness of virtual forest therapy over time.

The demographics of our participants may not be representative of the broader population of young adults, as the study mainly involved first-year psychology students. A more diverse sample, including individuals from various educational backgrounds and a wider age range (18–30 years), would enhance the generalizability of our findings. Furthermore, none of the participants were clinically diagnosed with severe social anxiety (a requirement of our research ethics protocol), and the relatively low initial social anxiety scores made it more challenging to detect significant reductions in anxiety levels post-intervention. Future research should include participants with varying levels of social anxiety severity and pre-screen for the level of social anxiety to balance it across conditions and provide a more sensitive measure of intervention effectiveness.

We derived a 5-item scale extracted from the Social Interaction Anxiety Scale (SIAS), which showed a strong correlation with the original 20-item scale (r > 0.9, p < 0.05), suggesting that a concise and less unwieldy version of the SIAS should capture most of the relevant information. Peters et al. developed a shortened six-item version of the SIAS, which was also applied in our analysis [72]. We calculated the Cronbach’s alpha for the six-item Peters scale and found internal reliability scores of 0.76 for the pre-session data and 0.81 for the post-session data, both of which were lower than the reliability measures we obtained for our custom five-item scale (0.87 and 0.93 for pre- and post-session data, respectively). Based on these findings, we chose to use our five-item version of the SIAS rather than the Peters scale. However, the variability in the reliability of the two shortened scales in our study suggests a need for further research to see whether either of these shortened versions, or others yet to be developed, can demonstrate stability across different samples. A short version of the SIAS should be particularly useful in settings where researchers want to use a range of instruments, and time and participant fatigue may become concerns.

More broadly, new measures for social anxiety that capture a broader spectrum of symptoms may need to be developed, since the SIAS predominantly measures interaction anxiety and lacks components that assess performance anxiety and avoidance behaviors [86].

Some participants reported experiencing nausea in the VR conditions in our study, although not to the point where they asked to stop the experimental session. Other researchers have used physical movement input, linked to movement in the virtual world, to reduce the amount of nausea experienced in VR [87]. The addition of physical movement input may be beneficial in future work on the use of virtual forests in social anxiety therapy.

Future research should explore whether the observed trends in sentiment scores in this study persist across multiple therapy sessions and whether these emotional responses influence treatment adherence or therapy effectiveness. Additionally, examining how specific VR elements (such as interactivity, nature sounds, or dynamic visual changes) contribute to these effects could help refine VR therapy design for maximum benefit.

We used one virtual forest and two paths within that forest. Future research should examine how variations in the designed features of a virtual forest affect various aspects of user experience, including immersiveness, engagement, enjoyment, and susceptibility to therapeutic instruction and resulting benefits. Relevant features of the forest that could be systematically varied in such research include the number of interactive elements, visual and auditory stimuli, and the degree of participant movement within the VR environment.

6. Conclusions

Our study explored the usability and effectiveness of combining a virtual forest with therapeutic exercises to address social anxiety in young adults. Condition Non-VR showed a tendency to reduce social anxiety in contrast to the other conditions thereby answering our first research question in the negative (in the context of a short duration study), and our second research question in the positive), which suggests that a simple application that focuses on therapeutic content, without the distractions of a rich virtual environment, may be better at reducing social anxiety, at least in the short term. For our third research question, we didn’t find that adding therapeutic instruction lowered the user experience or positive sentiment associated with the virtual forest, but we did find that therapeutic instruction reduced immersiveness as measured by the number of immersive words and phrases used in open-ended comments.

Condition B, combining virtual forest and therapeutic exercises, demonstrated benefits in enhancing long-term engagement and learning. Participants in this condition reported more positive experiences, and had greater quiz accuracy, indicating more accurate comprehension of therapy content. The virtual forest may sustain participant interest and engagement with therapeutic instruction, suggesting that the combined approach could be more effective in maintaining attention and fostering deeper comprehension of the therapy content over time. However, there may be potential for cognitive overload when therapeutic instructions are combined with an engaging immersive environment. It is possible that overload may have reduced the immersiveness of the virtual forest when therapy instructions were added (as shown by the large drop in Figure 7 in immersiveness for Condition B relative to the VR-only condition, i.e., condition A). Thus there is a need for careful design in therapeutic virtual forests to create appropriate levels of immersiveness and engagement, but without inducing cognitive load or compromising the effectiveness of therapy.

These findings highlight the importance of the delivery mode in therapeutic interventions. While the virtual forest offers an engaging and immersive experience that may enhance long-term learning and engagement, traditional non-immersive methods may facilitate a better immediate reduction in social anxiety. Future research should aim to optimize VR therapy by balancing the immersive elements of the virtual environment with features that help the participant focus on therapeutic content. One possibility may be to use simplified environments that are less immersive and distracting than VR while retaining some of the features and benefits of a forest bathing experience (e.g., a forest video, or succession of still images, on a large screen, combined with forest sounds and other ambient sound effects).

Virtual forest bathing is a promising environment for delivering therapeutic content. However, careful design is required to ensure the technology does not become distracting or interfere with the reception and understanding of therapeutic content. One key issue is how to deliver therapeutic content over time. For example, therapy could start in non-VR settings and progress to VR in later sessions to maintain participants’ interest and engagement in the therapeutic process. Another consideration is individual differences—some individuals may be particularly prone to cybersickness and may benefit from less immersive environments, such as images of forests combined with forest sounds.

Beyond general therapeutic settings, virtual forest technologies have notable practical applications. For example, these tools could be introduced to people with disabilities who are unable to physically visit a forest, elderly individuals who are homebound or live in nursing homes, leukemia patients undergoing treatment who have weakened immune systems and cannot interact with the external environment, or patients in closed institutions such as prisons undergoing rehabilitation. These populations, who may have limited access to natural environments, could greatly benefit from virtual nature exposure as a safe and effective means of promoting psychological well-being. While the challenges discussed in this paper are complex and considerable work will be required to develop virtual forest technologies for therapeutic interventions, the potential benefits of successfully implementing these technologies should far outweigh the research effort needed to develop them.