Physiological monitoring.

The CGX Aim II Physiological Monitoring device (CGX, Inc.), using ECG electrodes (Skintact Inc.), measured the following bio peripherals: Electromyography (EMG; 2 electrodes on the L/R base of the neck on the sternocleidomastoid muscle), Bio-Impedance-Based Respiration Rate (2 paddles with 2 electrodes each on the L/R pectoralis major), Heart Rate, and Galvanic Skin Response (GSR; 2 electrodes on the palm of the non-dominant hand).

Lab recorder.

Lab Streaming Layer with LabRecorder [6,10,11] was utilized to temporally synchronize our peripheral and experimental time series (experience start and end timestamps) within an XDF file format.

Trait measures.

1. a. Dispositional Positive Emotion Scale (D-PES;[53]) – The Awe Subscale of D-PES measures an individual’s dispositional propensity to feel awe toward the world, consisting of 6 items.
2. b. Modified Tellegen Absorption Scale (MODTAS; [54]) – Adapted from the Tellegen Absorption Scale (Tellegen & Atkinson, 1974), uses a Likert scale to assess the imaginative involvement and the tendency to become mentally absorbed in everyday activities.
3. c. NEO Five-Factor Inventory (NEO-FFI-3; [55]) – Measures five broad personality dimensions – neuroticism, extraversion, openness to experience, agreeableness, and conscientiousness.

State measures.

1. a. State-Trait Anxiety Inventory (STAI;[56]) – Assesses state anxiety (temporary/situational) and trait anxiety (enduring propensity).
2. b. Profile of Mood States 2^nd Edition (POMS 2; [57]) – Evaluates current mood state affective dimensions like anger, confusion, depression, fatigue, tension, and vigor.
3. c. Valence and Arousal – Measures emotional state on two Likert scales from 0 to 10: Valence (0= very unpleasant,10 = very pleasant) and Arousal (0 = totally calm, 10 = totally excited/agitated).

Outcome measures.

1. a. Toronto Mindfulness Scale (TMS; [58]) – A 13-item administered post-meditation to assess state mindfulness, while differentiating between reflective awareness and ruminative attention. The TMS contains two subscales: Curiosity, gauging the interest in one’s experiences, and Decentering, reflecting the ability to view thoughts and feelings as transient mental events.
2. b. Awe Scale (Awe-S; [59]) – An 8-item self-report scale measuring individual differences in the tendency to experience awe, using a 7-point Likert scale to rate their agreement with statements regarding feelings of awe and perceptions of vastness.
3. c. Body Size Estimation (BSE; [60]) – Assesses body image perception, from a selection of nine progressive figural drawings that depict varying body sizes to estimate their own perceived body size, both in height and width. This measure has been linked to the intensity of awe and immersion in non-ordinary experiences [61].
4. d. Ego-Dissolution Inventory (EDI; [62]) – Consists of 16 items assessing altered states of ego-consciousness, using a visual analog scale ranging from 0% to 100%. It includes 8 items relating to ego-dissolution, representing the core of the inventory, and 8 to ego-inflation, contrasting experience of heightened self-assuredness.
5. e. Igroup Presence Questionnaire (IPQ; [63]) – Measures sense of presence, immersion, engagement, and connection in virtual environments.
6. f. Modified Immersion Experience Questionnaire (adapted per [64]) – Measures depth and quality of immersion/engagement in a particular context or virtual environment.
7. g. Motion Sickness Assessment Questionnaire [65] – A 16-item scale assessing multiple dimensions of motion sickness (gastrointestinal, central, peripheral, and sopite-related symptoms).

Behavioral tasks.

1. a. Trail-Making Test (TMT; [66]) – A neuropsychological assessment, taken on the computer, evaluating visual attention, scanning, processing speed, and cognitive flexibility. Participants were instructed to connect a series of 25 dots in sequential order, alternating between numbers and letters (1-A-2-B, and so on), aiming for speed and accuracy.
2. b. Architex [67] – A validated, computerized neuropsychological assessment tool designed to objectively measure multiple cognitive domains and their interactions, including attention, working memory, processing speed, affect recognition, planning, organization, impulsivity, and verbal and nonverbal abstract reasoning.

Content delivery systems.

The first immersive content-delivery system was MetaQuest 2 (Meta Platforms, Inc.) VR headset, equipped with an Elite Strap (see Fig 1A), presenting stimuli through its native video player at a resolution of 5376x2688, scaled to the maximum allowed resolution of 1920x3664, and a refresh rate of 120Hz. All videos were captured using a 360 camera (GoPro Max 360) placed inside the MindGym during the programmed experience, with the audio content subsequently overlaid. During the experiment, the VR is set to cast through an iPad Air (iPadOS 16.6.1) and navigated using a wired mouse (Razer Inc.).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Participant setup.

(A) VR; (B) MindGym.

https://doi.org/10.1371/journal.pmen.0000269.g001

The second system was the MindGym (Lumena, Inc.), a 7’ isotropic MindGym lined with reflective mylar on its interior walls, while its floor and ceiling are equipped with mirrors. (see S1 Appendix). The MindGym incorporates WS2815 LEDs, with 121 pixels per edge between the vertices, totaling 1452 pixels across 12 edges. It utilizes SMD5050 RGB LED chip, offering RGB color with 256 grayscale levels and an output of 990-1080 lumens per meter. The LEDs are operated at less than half of their maximum capacity. Additionally, the system features a color temperature of 5500K. Participants in the MindGym group were provided over-ear wireless, noise-canceling headphones (Sony Group Corp.) to wear during the experience (see Fig 1B).

Stimuli.

The first piece of content was “4-8 Breathwork” (see Fig 2A, Fig 2C), led by Dr. Jannell MacAulay’s guided audio and utilizing two primary lighting setups. The beginning featured the “Mindful Minute” mindfulness exercise with multiple colors and slowly upward-moving LEDs, designed to encourage mind wandering among participants. MacAulay then introduced the breathwork technique and its underlying rationale, aided by “Galaxy” visuals of blue LEDS, around 6000K, embedded in the walls, giving the sensation of being surrounded by stars. The guided breathwork audio was complemented by upward and downward cascades of vertically aligned LEDs, synchronized with the inhalation and exhalation periods (see S2 Appendix).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Screenshot of stimuli.

(A) MindGym (Breathwork); (B) MindGym (Rain); (C) VR (Breathwork); (D) VR (Rain).

https://doi.org/10.1371/journal.pmen.0000269.g002

The second piece of content was “Rain” (see Fig 2B, Fig 2D), an immersive environment created for reflection. The audio begins with complete darkness as the sound of rain starts to trickle in. This is complemented by a downward cascade of vertically aligned blue LEDs, around 6500K, against a black background, simulating rain visuals falling from each vertex. The participant remains immersed in the light and sound environment for the remainder of the reflective experience.

Overview.

Before each session, all participants provided signed consent forms. They began the session by filling out pre-experience questionnaires and completing cognitive tasks, followed by CGX setup. Subsequently, each group underwent a 10-minute experiential phase, after which they completed post-experience questionnaires and tasks (see Fig 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Procedure overview.

The total participation time was roughly 80 minutes.

https://doi.org/10.1371/journal.pmen.0000269.g003

Randomization.

Subjects were iteratively assigned to one of two groups using a MATLAB-based randomization algorithm that maintained gender balance, with recruitment ceasing for each gender category upon reaching the predetermined quota. The experimenter remained blind to group assignment until launching the Python script, which initiated the group-appropriate content delivery system and stimuli by entering the participant’s ID number. Participants were randomly assigned to one of four content delivery systems and stimuli pairings: (1) MindGym (Breathwork), (2) MindGym (Rain), (3) VR (Breathwork), and (4) VR (Rain).

Procedure.

Upon arrival, participants were seated on an office chair facing a 35.5” x 55” desk and given an overview of the session (pre-questionnaires, equipment setup, 10-minute experience, post-questionnaires). They were informed their “experience” would involve sitting for 10 minutes in an immersive audiovisual environment.

For approximately the first 30 minutes, participants used the iMac to complete a set of questionnaires (NEO-FFI-3, D-PES, MODTAS, STAI, POMS-2, Valence, and Arousal) on Google Forms, followed by a brief neurophysiological assessment via the TMT and Architex. Participants were instructed to silence their phones and remove jewelry and bulky items for comfort, before being told which experiential group they were assigned to. Using a cotton pad and 91% isopropyl alcohol, the experimenter cleansed the relevant areas (see CGX materials above) before outfitting the participant with biosensors and peripherals.

Participants assigned to VR sat in the same room, with the room shades lowered. The VR headset was positioned and adjusted with the tension dial on the Elite strap for clear vision and comfort. The experimenter cast the VR to the iPad and adjusted the settings, such that the audio was played through the headset. Upon connecting to Lab Recorder, the stimuli are launched concurrently with the initiation of the LSL trigger, by the Python script.

If assigned to the MindGym, participants entered the MindGym and sat centrally, while holding the CGX device and wearing headphones. The Lab Recorder was connected, and the experience began within 15 seconds of closing the MindGym door.

Across groups, participants were told to sit still and relax during the 10-minute experience without falling asleep. Participants were shown how to remove the VR or exit the MindGym if they decided to end the experience early, otherwise, the experimenter would return to assist them once the experience ended.

After the experience, biosensors were removed and participants used the restroom if needed before completing post-experience behavioral assessments (STAI, POMS-2, Valence, Arousal, IPQ, MSAQ, TMS, Awe-S, BSE, and EDI) and cognitive tasks (TMT and Architex). Finally, participants were compensated and dismissed.

Dependent variables.

Seven primary outcome measures collected at both timepoints were used as dependent variables (DV) in separate models. There were four mood state measures: Anxiety (STAI), Arousal, Valence, and Total Mood Disturbance (POMS). There were also three cognitive measures: two that tracked performance on the Architex task (Total Speed and Total Score) and one that tracked performance on the TMT using the ratio of reaction time divided by accuracy, here called TMT RTACC, which decreases with increasing performance due to increased accuracy and/or decreased reaction time).

Eleven DVs were collected at the Post timepoint only, which included Awe, BSE, EDI, Immersion (based on IEQ), overall Motion, Curiosity and Decentering (both from Toronto), and four IPQ variables (General, Involvement, Spatial Presence, Experienced Realism).

There were also four physiological variables – Breath Rate, EDA, HR, and HRV – that were collected throughout the duration of each experience. These measures were averaged or computed in either two 5-minute timepoints (for HRV), five 2-minute timepoints (for EDA and Breath Rate), or ten 1-minute timepoints (for HR).

Independent variables.

The primary independent variables (IV) were Time or Timepoint (Pre, Post) as a categorical, within-subjects factor, for any variables collected at both timepoints, as well as both Tech (MindGym, VR) and Stim (Breathwork, Rain) as categorical, between-subjects group factors. Several continuous or ordinal IVs were tested for potential moderation or mediation of the Tech and Stim effects. There were eight psychological variables: DPES, Immersion, IPQ (General), IPQ (Involvement), IPQ (Spatial Presence), IPQ (Experienced Realism), MODTAS, and Openness. There were seven physiological variables: Motion (Overall), initial breath rate (Breath Rate, first 2 min), breath rate change (Breath Rate change, last 2 min – first 2 min), initial heart rate (HR, first 1 min), heart rate change (HR change, last 1 min – first 1 min), initial heart rate variability (HRV, first 5 min), and change in heart rate variability (HRV change, last 5 min – first 5 min). The demographic variables of age and sex were also included in all moderation models to control for their potential effects.

Nonparametric tests.

For all DVs collected at both timepoints (pre, post), changes over time were tested with nonparametric t-tests using Wilcoxon signed-rank tests in JASP [68]. Effect sizes were reported as the matched rank biserial correlation (RBC), an appropriate measure for nonparametric t-tests which ranges between –1 to 1 and indicates the relative proportion of positive and negative rank scores while ignoring ties (e.g., RBC = 0.60 means that 60% more subjects showed an increase instead of a decrease). Timepoint tests were performed separately for each of the four experimental groups (MindGym (Breathwork), MindGym (Rain), VR (Breathwork), VR (Rain)) as well as the combined MindGym and VR groups, the combined Breathwork and Rain groups, and all groups combined.

For all DVs, differences between groups were tested with nonparametric Mann-Whitney U tests in JASP, which included RBC effect sizes. Six separate group comparisons were tested: MindGym (Breathwork) vs MindGym (Rain), VR (Breathwork) vs VR (Rain), MindGym (Breathwork) vs VR (Breathwork), MindGym (Rain) vs VR (Rain), MindGym vs VR, and Rain vs Breathwork. For any DV collected at both timepoints, the difference score (Post – Pre) was used to compare group differences in changes over time. For the physiological measures, the first timepoint, last timepoint, and the difference between first and last timepoints were separately used as DVs to compare group differences in initial, final, or change in physiological response.

Moderation.

Moderation of the experimental effects on mood or cognition was tested as interactions between each covariate and the Tech and Stim group factors in separate models for the three-way interaction (Tech x Stim x Moderator) and two-way interactions (Tech x Moderator, Stim x Moderator), as well as a separate model with only main effects. Demographic covariates (age and sex) were also included as main effects in all models. For any outcome measure collected at both timepoints, the difference score (Post – Pre) was used as the DV. Because we did not have any a priori hypotheses about differential moderation effects between nested levels of Tech and Stim, we conducted post-hoc t-tests for only the interaction terms that were deemed significant.

Moderations were analyzed with generalized linear models (GLM), using the glmmTMB [69] and emtrends [70] packages in RStudio [71]. Based on extensive model diagnostics, which were conducted with the DHARMa [72] package, it was determined that either gaussian or t-family distributions (both with the identity link) were optimal for model fitting to ensure no substantial violations or problems with linearity, independence of errors, homoscedasticity, dispersion, zero-inflation, outliers, or within-group normality of residuals.

Separate models were tested for the hypothesized moderators, which included eight psychological variables (DPES, Immersion, IPQ General, IPQ Involvement, IPQ Spatial Presence, IPQ Experienced Realism, MODTAS, and Openness) and seven physiological variables (overall motion, initial Breath Rate, Breath Rate change, initial HR, HR change, initial HRV, HRV change). All moderation models were tested separately for 10 outcome variables as DVs: Arousal, Anxiety (STAI), Awe, BSE, EDI, Total Mood Disturbance (POMS), Valence, Architex (Total Score), Architex (Total Speed), and TMT (RTACC).

Repeated measures ANOVA.

For the physiological measures collected during the experiences, repeated measures ANOVAs (rmANOVA) were performed in JASP to assess changes over time as a main effect of Timepoint (two timepoints for HRV, five timepoints for Breath Rate and EDA, ten timepoints for HR) and to assess experimental effects on changes over time, with separate three-way (Tech x Stim x Timepoint) or two-way interactions (Tech x Timepoint, Stim x Timepoint). Levene’s test was used to ensure no violations of homogeneity. The assumption of sphericity was tested (using Mauchly’s test) and corrected (with Greenhouse-Geisser) if necessary. Effect sizes for all main or interaction effects were reported as generalized eta-squared, which appropriately adjusts for mixed design with within-subject (Time) and between-subjects (Tech, Stim) factors. Effect sizes for post hoc t-tests, if an interaction was significant, were reported as Cohen’s d.

Follow up analyses.

Three outcome variables - STAI, POMS (Total Mood Disturbance), and a discrete version (10 total responses ranging from ‘not at all anxious’ to ‘extremely anxious’) of the generalized anxiety visual analog scale (GA-DS) – were collected at the 1-week follow-up timepoint. For the two variables (STAI and POMS Total Mood Disturbance) collected at all three timepoints (pre, post, and follow-up), rmANOVAs were performed in JASP to assess changes over time as a main effect of Timepoint (pre, post, follow-up) and to assess experimental effects on changes over time, with separate three-way (Tech x Stim x Timepoint) or two-way interactions (Tech x Timepoint, Stim x Timepoint). One variable (GA-DS) was collected only at the follow-up timepoint, so an ANOVA was performed to compare groups. Levene’s test was used to ensure no violations of homogeneity. The assumption of sphericity (for models including Timepoint) was tested (using Mauchly’s test) and corrected (with Greenhouse-Geisser) if necessary. Effect sizes for all main or interaction effects were reported as generalized eta-squared (for models including Timepoint) or partial eta-squared (for models without Timepoint). Effect sizes for any relevant post hoc t-tests were reported as Cohen’s d.

Multiple comparison correction.

We used the false discovery rate (FDR) method of correcting P values for multiple comparisons (significance threshold chosen as a=.05, P<.05) to control Type I and Type II errors [73,74]. FDR correction was performed across DVs and separately for 1) the nonparametric tests of timepoint effects for each group (single or combined), 2) the nonparametric tests of group differences, separately for each group comparison and separately for the measures collected at both timepoints, the measures collected at only the Post timepoint, and the physiological measures collected during the experience, as well as 3) the moderation tests, separately for each interaction or main effect. FDR correction of the rmANOVAs was performed across the different models testing interaction or main effects and separately for each measure.

Mediation.

Mediation analysis was conducted in JASP based on the lavaan [75] package in R. Missing values were excluded listwise. Separate models were tested for each potential mediator, which included the same variables that were used as potential moderators in the moderation analysis, and for each DV that was also used in the moderation analysis.

Data acquisition and preprocessing.

Physiological data, including electrocardiogram (ECG), electrodermal activity (EDA), electromyography (EMG), and respiration, were recorded using the CGX AIM Phys. Mon. AIM-0106 device. The data were stored in XDF format, which is a standardized format for multi-channel time series data [76]. The XDF files were loaded using the pyxdf library, and the relevant data streams were extracted based on their names and labels.

ECG processing.

The ECG data were first sanitized to remove any invalid or missing data points [77]. A notch filter was then applied to remove powerline interference at 60 Hz. The ECG signal was further cleaned by applying a bandpass filter with a low cut-off frequency of 0.5 Hz and a high cut-off frequency of 45 Hz using a 4^th-order Butterworth filter [78]. The filtered ECG data were then filtered again to remove any values exceeding a predefined amplitude threshold of 1500 mV.

Baseline wandering, a low-frequency artifact in the ECG signal, was computed by measuring the peak-to-peak amplitude of the ECG signal. If the baseline wandering exceeded a predefined threshold of 1700 mV, the ECG data were considered unacceptable for further analysis.

EDA processing.

The EDA data were processed in conjunction with the EMG data to identify and remove motion artifacts. First, the EMG signal was normalized. A sliding window approach was then used to identify windows of EMG data that contained a high proportion of outliers, which were considered to be indicative of motion artifacts. The Local Outlier Factor (LOF) algorithm [79] was used to detect outliers within each window (implemented using the scikit-learn library) [80].

The EDA signal was smoothed using a median filter with a kernel size of 8 seconds. A low-pass Butterworth filter with a cut-off frequency of 5 Hz was then applied to remove high-frequency noise. The contaminated windows identified in the EMG signal were used to guide the interpolation of the EDA signal. Linear interpolation was performed to replace the artifact-contaminated segments of the EDA signal with values interpolated from the neighboring clean segments.

The quality of the cleaned EDA signal was assessed by checking for values outside the acceptable range (0.05 - 60 μS) and for rapid changes exceeding ±10 μS/sec within a 1-second interval [81]. The proportion of data points violating these criteria was used to compute a quality index ranging from 0 to 100, with higher values indicating better signal quality. If the quality index was below 50, the EDA data were considered unacceptable for further analysis.

Respiration processing.

The respiration data were preprocessed following the methods of [82]. We first applied a high-pass Butterworth filter with a cut-off frequency of 0.1 Hz to remove low-frequency drift. The filtered signal was then resampled to a fixed sampling rate of 100 Hz. A sliding window approach was used to estimate the respiration period within each window using the Fast Fourier Transform (FFT). The estimated respiration period was used to guide the application of a moving average filter to smooth the respiration signal within each window. The smoothed signal was then resampled back to the original sampling rate.

Inhalation and exhalation onsets were detected by identifying the intercepts between the smoothed respiration signal and its moving mean. The intercepts were then classified as inhalation or exhalation onsets based on the direction of the slope at each intercept. A minimum amplitude threshold of 0.25 was used to filter out small fluctuations that do not correspond to true respiratory events.

The number of breaths within the first and last 5 minutes of the experiment was computed using the inhalation and exhalation onsets to define individual breaths.

Heart rate variability analysis.

Heart rate variability (HRV) analysis was performed by computing the low-frequency (LF) and high-frequency (HF) components of the ECG signal using the Welch’s method for power spectral density estimation [83]. The LF component was defined as the power in the frequency range of 0.04-0.15 Hz, while the HF component was defined as the power in the range of 0.15-0.4 Hz. The ratio of LF to HF power (LF/HF ratio) was computed as a measure of sympathovagal balance [84]. The LF/HF ratio was computed separately for the first and last 5 minutes of the experiment.

EDA tonic component analysis.

The tonic component of the EDA signal, which reflects the slow-varying level of skin conductance, was extracted by applying a low-pass filter to the EDA signal [77]. The mean value of the tonic component was computed within non-overlapping 2-minute intervals. These mean values were used as a measure of the overall level of arousal throughout the experiment.

IRB.

The Advarra (Columbia, MD) Institutional Review Board approved all recruitment and testing procedures before initiating enrollment (Pro00070581). Following ethical standards and legal requirements, all study participants provided written informed consent on a document, hosted through DropboxSign. They were afforded ample time to seek clarification from the Principal Investigator and study personnel. The informed consent process incorporated the California Experimental Research Bill of Rights as mandated by Health and Safety Code Section 24172. Additionally, the study adhered to the principles outlined in the Declaration of Helsinki. Furthermore, all laboratory personnel possessed and maintained up-to-date certifications in Good Clinical Practice and the Protection of Human Research Participants through online training.

Timepoint effects within group.

Analysis of pre-post changes revealed significant improvements across various measures when considering all groups combined and within specific conditions (see Table 4 and S1 Tables A-I for detailed results).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Summary of results from the nonparametric (Wilcoxon Signed-Rank) t-tests of timepoint (pre vs post) differences for the different groups or group combinations.

https://doi.org/10.1371/journal.pmen.0000269.t004

When considering all groups combined, there was evidence of a decrease in TMT (RTACC) (W=5443.00, Z=4.32, p.fdr<.001, RBC=.45), Architex (Total Speed) (W=6044.00, Z=7.82, p.fdr<.001, RBC=.84), Anxiety (STAI) (W=5156.50, Z=4.86, p.fdr<.001, RBC =.52), Valence (W=1626.00, Z=-2.14, p.fdr=.042, RBC=-.26), and POMS (W=5659.00, Z=5.77, p.fdr<.001, RBC=.61).

Examining specific conditions, the MindGym (Breathwork) group demonstrated improvements in Architex (Total Speed) (W=396.00, Z=3.36, p.fdr<.001, RBC=.70) and POMS (W=331.00, Z=2.92, p.fdr=.014, RBC=.63). The MindGym (Rain) condition showed enhancements for TMT (RTACC) (W=411.00, Z=3.19, p.fdr<.001, RBC=.66), Architex (Total Speed) (W=421.00, Z=4.40, p.fdr<.001, RBC=.94), STAI (W=353.50, Z=2.94, p.fdr=.005, RBC=.63), and POMS (W=442.00, Z=3.80, p.fdr<.001, RBC=.78).

The VR (Breathwork) group exhibited improvements in Architex (Total Speed) (W=355.00, Z=3.99, p.fdr<.001, RBC=.88), STAI (W=405.00, Z=3.55, p.fdr<.001, RBC=.74), and POMS (W=406.00, Z=3.10, p.fdr=.005, RBC=.64) (see S1 Table C), while the VR (Rain) condition only showed a significant improvement in Architex (Total Speed) (W=379.00, Z=4.01, p.fdr<.001, RBC=.87).

When comparing the combined Breathwork and Rain conditions, both showed significant improvements across multiple measures, with slight variations in effect sizes. Similarly, the combined MindGym and VR conditions significantly enhanced cognitive performance, anxiety reduction, and mood improvement.

Group differences in timepoint effects.

Analysis of group differences in timepoint effects revealed no statistically significant differences between experimental conditions after FDR correction for multiple comparisons (all p.fdr>.05) (see S2 Tables A-F). While all interventions yielded positive outcomes across various measures, including improvements in cognitive performance, anxiety levels, and mood, the magnitude of these changes did not differ significantly between experimental conditions when accounting for multiple comparisons. This lack of significant between-group differences indicates that no single condition demonstrated superior efficacy over the others in terms of the measured variables.

Outcome measures as a function of group.

No statistically significant differences were observed between the group comparisons for the post-intervention measures (all p.fdr>.05) (see S3 Tables A-F). The closest result to reaching statistical significance with a notable effect size was found for IPQ (Involvement) when comparing Breathwork to Rain (W=2501.50, p.fdr=.12, RBC=.26), suggesting a potential trend towards higher involvement in the Breathwork condition compared to the Rain condition.

Group differences.

Analysis of physiological measures revealed significant differences between experimental conditions, particularly in breath rate patterns, the most notable being between the Breathwork and Rain conditions (see S4 Tables A-E for all results).

At the end of the session, Breathwork participants had significantly lower final Breath Rate compared to Rain (W=269.50, p.fdr<.001, RBC=-.70). The overall change in Breath Rate from beginning to end of the experience was significantly different between the two conditions (W=198.50, p.fdr<.001, RBC=-.78), indicating a more substantial decrease in breath rate for the Breathwork group.

Similar patterns were observed when comparing the MindGym (Breathwork) and MindGym (Rain) conditions. MindGym (Breathwork) had a significantly lower final Breath Rate (W =77.00, p.fdr<.001, RBC=-.68) compared to MindGym (Rain), and the overall change in Breath Rate showed greater decrease, with a significantly different Breath Rate (W=68.50, p.fdr<.001, RBC=-.72) with MindGym (Breathwork).

VR (Breathwork) and VR (Rain) showed significant differences in breath rate patterns, with significantly lower Breath Rate at the end of the experience (W=55.00, p.fdr<.001, RBC=-.73) compared to VR (Rain). The overall change in Breath Rate was significantly different (W=37.00, p.fdr<.001, RBC= -.82), with VR (Breathwork) demonstrating a more substantial decrease.

These results consistently demonstrate that participants in the Breathwork conditions, regardless of the delivery method (MindGym or VR), experienced a decrease in breath rate throughout the session compared to those in the Rain conditions. This pattern suggests that the Breathwork interventions effectively modulated participants’ breathing patterns, potentially leading to a more relaxed physiological state by the end of the session.

No other physiological measures (EDA, HR, HRV) showed significant differences between conditions after FDR correction.

Changes over time.

We employed repeated measures ANOVA (rmANOVA) to examine changes over time and the potential effects of experimental conditions. We observed significant temporal changes in HR, Breath Rate, and EDA levels, with only Breath Rate showing interactions between time and technology. No significant main effects of time or interactions with groups were observed for HRV after FDR correction.

The results revealed a significant main effect of time on HR (F_{5.17, 398.38}=8.78, p.fdr=.004), indicating a gradual increase throughout the experiment (see Table 5). EDA levels revealed a significant main effect of time (F_{2.35, 162.37}=17.11, p.fdr=.004), characterized by an initial increase followed by a plateau around the third timepoint.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Results from repeated measures ANOVA on physio measures.

https://doi.org/10.1371/journal.pmen.0000269.t005

Breath Rate showed a significant main effect of time (F_{2.02, 171.33}=25.31, p.fdr=.002), characterized by an initial decrease followed by a plateau around the third timepoint. A significant interaction between time and Stimuli was also observed (F_{2.48, 208.49}=28.49, p.fdr=.002). Post hoc paired-sample t-tests (with Holm correction of P values) were conducted between timepoints within Stimuli groups. Within the Rain groups combined, breath rate did not reliably change between any timepoints (all p.holm>.05). Within the Breathwork groups combined, breath rate significantly decreased from the first to second timepoint (ΔM=-3.55, SE=.49, t=7.19, p.holm<.001, d=1.08), remained similar between second and third timepoints (ΔM=-1.12, SE=0.49, t=2.26, p.holm=.416, d=.034), decreased again from third to fourth timepoints (ΔM=-2.28, SE=.49, t = 4.62, p.holm <.001, d=.69), without any further reliable decrease from fourth to fifth timepoints (ΔM=.77, SE=.49, t=1.56, p.holm=1.000, d=.23).

Mediation.

Mediation analyses were conducted to test if hypothesized variables mediated the group effects on the outcome scores. No mediation analyses passed the threshold for significance after FDR correction (all p.fdr>.05).

Moderation.

Our moderation analysis revealed significant effects of both psychological and physiological factors on participants’ experiences and performance during the immersive interventions (see S5 Tables A-J for detailed results). Among the psychological trait measures, MODTAS emerged as a significant predictor of both Awe (b=.37, SE=.11, Z=3.28, p.fdr=.004) (see Table 6) and EDI (b=.03, SE=.01, Z=2.98, p.fdr=.015), with higher MODTAS scores associated with greater reported Awe and EDI. DPES also showed a significant main effect on Awe (b=13.89, SE=3.48, Z=4.00, p.fdr<.001), indicating that individuals with higher dispositional positive emotions were more likely to experience awe during the intervention.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Results from GLM tests of moderation.

https://doi.org/10.1371/journal.pmen.0000269.t006

Psychological state measures also played a significant role in moderating participants’ experiences. Immersion showed strong main effects on both Awe (b=1.16, SE=0.22, Z=5.17, p.fdr<.001) and EDI (b=0.07, SE=0.02, Z=4.55, p.fdr<.001), with higher levels of immersion consistently associated with greater reported awe and emotional depth. IPQ (General) was also positively associated with Awe (b=6.18, SE=1.72, Z=3.59, p.fdr=.002) and EDI (b=0.32, SE=0.13, Z=2.49, p.fdr=.048). Only IPQ (Involvement) followed this same pattern with Awe (b=2.24, SE=0.76, Z=2.96, p.fdr=.009) and EDI (b=0.18, SE=0.05, Z=3.24, p.fdr=.009).

Turning to physiological moderators, we found the change in heart rate throughout the intervention (HR-diff) showed a significant interaction effect with the type of technology used on Architex (Total Score) (F_df1=10, p.fdr=.023). Specifically, an increase in heart rate was associated with decreased accuracy for participants in the MindGym groups, but not for those in the VR.

Follow up analyses.

Follow-up analysis results are fully reported in the Supplemental Materials (S6 Tables A-C; S6 Fig). Both STAI and POMS (Total Mood Disturbance) showed significant changes over time, without any interactions between groups, such that scores significantly decreased immediately after the experience (consistent with the nonparametric results reported earlier) but then significantly increased one week later to a similar level (POMS) or even higher level (STAI) compared to before the experience. Effect sizes of pairwise group comparisons were small to moderate (ranging from 0.09 to 0.68). For the measure of anxiety (GA-DS) collected only at the 1-week timepoint, there were no significant main effects of Tech groups (F(1,85) = 0.31, p = 0.577, partial η² = 0.004), Stim groups (F(1,85) = 0.11, p = 0.738, partial η² = 0.001), or interaction between Tech and Stimuli groups (F(1,83) = 0.88, p = 0.352, partial η² = 0.010).