The Indiana University Institutional Review Board approved all recruiting and study procedures (protocol number 17398). Students provided informed consent before study participation. Privacy and confidentiality were maintained for all participants, and compensation was awarded in the form of class credit.

A total of 57 healthy undergraduate students were recruited from an urban midwestern university via online listings through the university’s Sona Systems human subjects pool; course credit was provided for participation. Students provided informed consent before study participation. Exclusions included poor sense of smell, extreme sensitivity to odors or volatile chemicals, chronic or current asthma, pregnancy or nursing, or the use of a nasally administered medication (excepting steroids). Physiological data collection devices were unavailable in the beginning of the study, so those data were not collected from 21 participants. Technical challenges additionally limited data collection for heart rate and respiration (n=1), heart rate only (n=2), and respiration only (n=1). At least 1 participant did not feel well enough to complete the study and provided only demographic and personality data.

Prior to the VR experience, participants provided demographic information and self-reports of anxiety, height-specific anxiety (major component of acrophobia), and impulsive sensation seeking. All self-report inventories were administered via Qualtrics. Participants also completed the behavioral sensation-seeking task. Participants were then guided through an immersive fear-inducing VR simulation of height exposure, with physiological recording initiated before the VR experience and continuing until the VR task was completed. Self-reported state anxiety was measured again near the end of the fear induction while in VR. Figure 1 illustrates the experimental procedures in a timeline format.

The Aroma Choice Task (ACT) is a validated behavioral test of sensation seeking that measures the relative preference for an intense, novel, varied, risky option versus a mild, safe, “boring” option, with odorants delivered in real time. Participants are instructed:

Choice ratio, the percentage of “Varied” choices out of a total of 20 binary choice trials (range: 0%‐100%), yields a single behavioral index reflecting behavioral sensation seeking (designed after self-reported sensation-seeking trait descriptions) [30,31]. The original ACT was developed with an air dilution olfactometer [36], but a simpler, manual version yielded analogous results [38]. We further modified the task to deliver 20 trials instead of the original 40, as our prior work indicated that the first 20 trials accurately capture the trait with lower participant burden [36].

Heart rate, respiration, and skin conductance were collected for 36 participants using the BioRadio (Great Lakes Neurotechnologies; Cleveland, OH, USA) and skin electrodes plus a respiratory inductance plethysmography belt, with data logged on a laptop computer. At least 1 researcher closely assisted the participant during VR to prevent collisions and stumbles. A second researcher entered live event markers (single keystrokes) that were logged in the BioRadio data stream as the participant progressed through the phases of the VR experience. Standardized breathing exercises (eg, Balban et al [48]) prior to VR were intended to mitigate physiological variability between participants. The “Pre-Plank” measurements included the anticipatory elevator ride up to altitude and the 20 seconds before stepping onto the plank (mean 0.74 [0.24] minutes; max 1.70). The “Plank” measurements comprised stepping onto the plank after the 20 seconds was complete and then walking the length of the plank and off into freefall (mean 1.88 [0.60] minutes; max 2.85). Skin conductance data were not analyzed.

The fear of falling from heights is an “innate” fear, that is, nonreliant on associative conditioning [29] or locomotion experience [49], and nearly universal [50], meaning it is highly generalizable and reliable for eliciting potent fear responses. An immersive height exposure simulation was delivered on the Meta Quest 2 head-mounted VR display. The Meta Quest 2 features 6 degrees of freedom, high refresh rate (up to 120 Hz), and adjustment for interpupillary distance [51,52]. These features, combined with limited movement and short VR duration (mean 2.90 [0.58] minutes, max 3.90), minimized the risk of VR sickness [53]. Nonetheless, participants were informed about the risk of VR sickness and were encouraged to notify the researcher at any time to end the simulation. No participants complained of VR sickness or feeling unwell due to the VR experience.

The immersive height exposure simulation was the “Richie’s Plank Experience” [44] delivered with a Meta Quest 2 head-mounted VR display. This simulation is widely used in VR research on fear responses and behavior [43,54,55]. It has also been used to study risk-taking [56] and suicide willingness [57] and is even proposed as a psychological preparation for psychedelic experience [58]. The paradigm simulates an elevator ride to the top of a tall building, and upon the door opening, presents a cityscape view from heights (~80 stories). Participants walk onto, down the length of, and off the end of a wooden plank protruding from the skyscraper. The height illusion is conveyed by audio and visual simulation of extreme exposure to open space and reified with wind noise and birds flying below the participant. Immersion was maximized with haptic feedback from a real wooden plank (6’ long “2×8” [2 m×4 cm×19 cm]) spatially registered to the virtual plank. The wooden plank was slightly warped and thus creaked and shifted with human weight. Although verbal instructions were provided as needed, researchers refrained from unnecessary verbal or physical interaction during the experience to preserve presence. A simulated participant (the first author) walks on the plank (Figure 2A) in the center of partitioned office space (Figure 2B), with the participants’ view at heights illustrated in Figure 2C.

Physiological data collected by the BioCapture (Great Lakes Neurotechnologies; Cleveland, OH, USA) software were modeled in VivoSense (version 3.4). VivoSense calculated the means of heart rate and respiration rates in 10-second bins. Changes in physiological measures were calculated as the average rate during anticipation of stepping on the plank (Plank) minus the baseline average (Pre-Plank). All statistical tests were performed in SPSS v29.0.2.0 (IBM). Analyses were stratified by sex as prior work demonstrated important interactions by sex between anxiety and sensation seeking [26] and by sex and elements of sensation seeking [30,59]. VR-induced increases in anxiety and physiological arousal were tested with paired t tests (2-tailed) between baseline “Pre-Plank” and the moment of stepping off the plank “Plank.” Change scores were calculated for self-reported state anxiety and physiological measures (Plank minus Pre-Plank scores). These were tested for correlations with behavioral sensation seeking, each other, and height-specific anxiety. The potential effect of changing the STAI-S assessment method (computer vs verbal inside VR) was tested with independent-samples t tests (2-tailed). We tested for differences in participant characteristics of those with physiological recording versus without to assess potential effects on interpretation. t tests (2-tailed) of inhomogeneous variance (Levene test) were performed using the Welch t test (2-tailed). The false discovery rate was controlled with the Benjamini-Hochberg procedure [60], and limited to 5% (q<.05), by setting α=.0278 for individual tests in a priori hypotheses.

Tests evaluating baseline sex differences (χ² for nominal, t test [2-tailed] for continuous) reported were uncorrected and descriptive in nature. All in-text data are reported as mean (SD). Women reported higher baseline levels of anxiety than men did (neuroticism-anxiety, state anxiety, and height-related anxiety, P<.03; Table 1). No other sex differences were detected.

The VR experience increased anxiety in both men (t₁₆=5.29, P<.001, 28.59 [7.87], and 44.76 [17.27], q<.05, Pre-Plank and Plank, respectively) and women (t₃₈=9.50, P<.001, q<.05, 34.26 [9.31] and 54.05 [14.51]; Figure 3A). The increase in anxiety was not significantly different before versus after the data collection method change (for more details, see the Anxiety section; P=.75).

In the 36 participants from whom physiological data were collected, the VR experience increased heart rate in both sexes (t₈=3.84, P=.005 for men t₂₃=6.99, P<.001 for women; q<.05; Figure 3B). Similarly, the VR experience increased respiration in both sexes (t₈=3.06, P=.02 for men and t₂₄=5.35, P<.001 for women; q<.05; Figure 3C). Of note, characteristics (Table 1) did not differ between participants with physiological data collected versus those without, q>.05 (11 tests); P=.53, P=.046, P=.24, P=.51, P=.84, P=.88, P=.62, P=.60, P=.25, P=.50, and P=.61, corresponding to childhood income, age, impulsive sensation seeking, aggression-hostility, activity, sociability, neuroticism-anxiety, AQ-anxiety, STAI, sex, and race, respectively.

Fear (evoked state anxiety) did not correlate with changes in heart rate (P=.60 and .79 for men and women, respectively) nor respiration (P=.47 and .48) in the participants from whom physiological data were collected. Baseline anxiety was also uncorrelated with changes in heart rate (P=.70 and .29) and changes in respiration (P=.71 and .17).

Fear of heights (acrophobia) was correlated with evoked state anxiety in men (r(15)=.606, P=.01) and women (r(38)=.410, P=.009, q<.05; Figure 4).

High intensity preference (ACT scores) negatively correlated with evoked state anxiety (Δ STAI-S) in men (r(15)=−.559, P=.02, q<.05), but not in women (P=.67; Figure 5). Preference for high intensity was not correlated with baseline STAI-S scores in women (P=.61), although there was a trend in men (r(15)=−.465, P=.06).

High intensity preference (ACT scores) negatively correlated with increased heart rate in men (r(7)=−.771, P=.02, q<.05), but not women (P=.54; Figure 6). Increased respiration was uncorrelated in both sexes (P>.28).

High intensity preference did not correlate with self-reported sensation seeking in men (P=.47) or women (P=.72); collapsing across sex did not permit detection of an association (P=.89). See Table 2 for correlations of personality and behavioral metrics (uncorrected).

We found support for our hypotheses concerning VR’s capacity to evoke fear (self-reported changes in state anxiety and physiological arousal) and the association between evoked fear and fear of heights. We found support for the negative association between evoked fear and behavioral sensation seeking in men, but not women. This unexpected finding is potentially due to higher trait anxiety scores in women, that is, imposing a ceiling effect and making associations between elicited anxiety and other traits difficult to detect. Interestingly, evoked fear measures were uncorrelated. Behavioral and self-reported sensation seeking were uncorrelated. These findings suggest that VR experiences possess sufficient ecological validity to elicit subjective and objective fear responses mirroring responses to real-world scenarios. Moreover, responses to the digital experience reflect associations with other traits (acrophobia and sensation seeking).

This study demonstrates the potential of VR for neuroscience and clinical research. VR continues to offer considerable promise for clinical and research applications alike. The expanding reach of VR is facilitated by better immersion technology, decreasing cost, creative applications, and wider adoption. While VR has long been used to administer exposure therapy [61-63] and treat pain [64-66], emergent applications target increasingly abstract constructs [67,68]. Germane to this study, VR applications effectively treat various anxiety disorders (eg, specific phobias, social anxiety disorder, panic disorder) in randomized controlled trials, producing effects comparable to conventional treatment and significantly better than passive controls [69]. In addition to promising efficacy data, providers and patients appear eager to adopt VR methods, as suggested by 2 recent reports on integrating VR in clinical practice [70,71]. The immersive nature of this nascent technology permits new avenues of investigation and permits research on humans that would be dangerous or impractical to study with real stimuli. We believe that the potential of VR to unite human and animal paradigms heralds a new translational era wherein strictly controlled animal neuroscience experiments can be accurately replicated in humans. For example, the elevated plus-maze—the gold standard in rodent anxiety research—can be instantiated for human participants [19] to connect basic neuroscience in rodents with human behavioral data. Other behavioral paradigms widely used in rodents, such as conditioned place preference, can now be accurately reproduced in humans using VR [72], producing convergent findings in both research contexts. Thus, the role and relevance of VR for both laboratory research and clinical practice are expected to grow substantially in the near future. The use of VR permits testing extant theoretical knowledge in more lifelike settings and experiences, and when paired with behavioral tests, yields increasingly objective outcomes.

VR can add knowledge to mature bodies of research, such as approach-avoidance, by presenting realistic simulations in humans. Approach-avoidance describes behavior that orients organisms toward positive and away from negative stimuli, respectively [73]. Approach-avoidance tendencies are likely rooted in evolutionary factors, primarily through sexually divergent selection pressure [74]; that is, reproductive fitness optimized by exploratory behaviors in men [75,76] and harm avoidance in women [77,78]. The extremes of the approach-avoidance continuum are marked by exaggerated attention on reward or threat cues [79]. These tendencies emerge from overactive brain reward or motivational systems [80] and underregulated brain threat systems [81] for approach, and conversely, overactive brain threat systems [82] and underactive brain reward systems [83] for avoidance. Sensation seeking and fear represent aspects of approach and avoidance, that is, opposing processes that modulate threat responses, such that high sensation seekers are less physiologically responsive to threat stimuli than low sensation seekers. One study testing fear responses found that high sensation seekers showed no response to threatening stimuli (versus control stimuli), whereas low sensation seekers produced an 8-fold increase in electromyographic response to threats [25]. No group difference in self-reported emotional reactivity was detected, further supporting the value of objective measurements.

Men are higher sensation seekers than women (particularly thrill or adventure seeking and disinhibition) [84,85], but the relationship between sensation seeking and fear appears to differ by sex. Investigating this relationship as an interaction with sex, Blankstein [26] found that the Sensation Seeking Scale (SSS) [23] total score negatively correlated with anxiety reactivity (Activity Preference Questionnaire) total and subscales (Social and Physical) at r>.43, P<.01 in men, but not in women (r<.07). However, a similar study found a number of negative correlations between the Sensation Seeking Scale and anxiety-related items (S-R Inventory) in both sexes [86], indicating mixed results in detecting sex interactions with approach-avoidance correlations. The lack of consilience in prior work might be explained by either (1) dependence on self-report inventories, with self-reported fear and sensation seeking often incongruent with objective measures [38,87], or (2) high anxiety and low sensation seeking in women producing restricted ranges (ceiling and floor effects), making correlations difficult to detect. Both factors potentially contribute together to the divergence. Future well-powered studies, ideally using precise behavioral tasks, should clarify these possible associations.

We did not detect correlations between self-reported fear and physiological responses to height exposure. While this is perhaps a surprising result, prior work suggests that self-reported fear does not necessarily reflect biological responses. In a real-world test, participants’ self-reported fear of crime did not differ between walking down a dimly-lit path (vs well-lit control), but the dimly-lit path participants’ heart rate increased by 17% (P=.002), with that in the controls remaining unchanged [87]. Even patients with anxiety disorders do not accurately report the degree of physiological responses to stress in laboratory tests [88,89]. This lack of concordance may be explained by individual differences in interoceptive ability [90]. The disconnection between self-reported traits and objective measures is also found in behavioral assessments of impulsivity [91], empathy [92], and risk preference [93], suggesting that the incongruence extends well beyond fear and anxiety. A recent report on associations between interoceptive ability and autobiographical memory [94] indicates that interoceptive perception (physical self-awareness) relates to episodic recall (cognitive self-awareness) and suggests the intriguing possibility that individual differences in these domains may be governed by some larger self-awareness meta factor.

Behavioral seeking and self-reported sensation seeking were uncorrelated in this sample. Existing findings offer scant data on this association due to the absence of behavioral sensation seeking tasks. Related traits such as impulsivity and risk-taking reveal low agreement between behavioral and self-reported assessment; for example, various measures of impulsivity are correlated at r=~0.1 in meta-analysis [91], and practically no relationship is observed between risk-taking measures [95]. Low reliability is observed in behavioral task data across domains [96]; in parallel, self-reported data suffer from various serious forms of bias [97]. The “jingle fallacy” (conflating interpretation of 2 measures because they have the same name) [98] exacerbates this problem. While the lack of convergence between behavioral and self-reported findings is perhaps unintuitive, this divergence between self-reported and behavioral data represents opportunities to discover additional features of personality traits. That is, important features of a given trait may not be fully captured by any single task or inventory.

Some limitations should be acknowledged. First, the sample would benefit from more power. The homogeneity of the sample—reflecting typical undergraduates—is predominately female and White, precluding well-powered direct comparisons by sex and potentially limiting generalizability. Generalizability would be enhanced by a community sample with a more even sex distribution and a larger range of age and socioeconomic status. Finally, the truncated sample of subjects providing physiological data was suboptimal, as physiological data were only collected from 36 of the 57 participants, although the absence of significant differences between the subsamples somewhat mitigates this concern.

The current report demonstrates the potential use of VR for neuroscience and clinical research. Beyond research and education, VR is now well established as a clinically valuable tool in health care. The combination of using VR, objective measures (physiological recordings and behavioral tasks), and subjective measures (self-report) to investigate a behavioral health topic allows more rigorous investigation than any one of these approaches alone. Through these measures, we confirmed associations between self-reported experience and physiological fear in response to heights, in addition to behavioral patterns and personality related to sensation seeking. Further evidence of the disconnection between objective and self-reported methods was found, although this was perhaps unsurprising. We expect ever-wider adoption of VR applications and objective measures in the clinic and continued expansion in the laboratory research domain.