Virtual audiences trigger real anxiety and comparable voice responses
How this was rated
Within-subjects design with matched real and virtual conditions; multiple independent outcome measures converging on the same conclusion. Downgraded from high because the sample is non-clinical university students and a single speaking context, limiting direct transfer to communication-difference populations.
Ratings use a simplified four-tier scheme (High, Moderate, Low, Very Low) informed by the GRADE working group. Learn more about how studies are rated.
Sixty university students gave presentations to a real audience, a virtual audience, and an empty virtual room. The virtual audience triggered anticipatory anxiety and heart rate increases similar to the real audience, and voice measures were largely equivalent across conditions.
The evidence suggests that carefully designed virtual audiences can serve as a meaningful stand-in for real audiences in speaking practice, producing comparable subjective, physiological, and voice responses in non-clinical adults.
Key findings
- Virtual avatars induced anticipatory anxiety: SUDS was significantly higher for virtual audience vs. empty virtual room
- Heart rate during speaking did not significantly differ between real and virtual audience conditions
- Voice parameters (f0 median and SD) were largely equivalent across real and virtual audience conditions
- Filled pauses percentage was significantly higher for real audience than virtual audience (T=3.64, p<.001) - the only significant between-condition difference in oral communication parameters
- More anxious individuals with higher presence scores showed responses closest to real-audience performance
Background
For VR to be useful in communication support, it needs to produce responses that genuinely resemble those triggered by real audiences. While earlier studies have measured subjective anxiety and stuttering frequency, few have simultaneously examined voice acoustics and physiological arousal in the same participants across matched real and virtual conditions. A multi-measure validation was needed.
What the researchers did
Sixty French-speaking university students (34 female / 26 male, mean age 21.03, SD 6.4) from the University of Liege, Belgium, each completed three speaking conditions: presenting to a live audience, presenting to a virtual audience displayed through an HTC Vive Pro Eye Office 1 head-mounted display, and speaking in an empty virtual room with no audience present. The virtual environment featured a Meeting Room setting with 8 virtual agents (4 male / 4 female) seated around a rectangular table - not an auditorium.
Heart rate was recorded continuously via a Polar Verity Sense wrist-worn device. Self-reported anxiety was measured using the Subjective Units of Distress Scale (SUDS) before and during each condition. Acoustic voice parameters including fundamental frequency (f0) median and standard deviation were analyzed. Presence in VR was measured with the ITC-SOPI scale. Six main measures were used, with Bonferroni-corrected significance threshold set at alpha=0.008 to control for multiple comparisons. Oral communication parameters (disfluency, filled pauses, total disfluencies percentage) were also assessed.
What they found
The virtual audience successfully triggered anticipatory anxiety, with SUDS ratings significantly higher before speaking to virtual avatars compared to the empty virtual room. Heart rate during the speaking task did not differ significantly between the real and virtual audience conditions, indicating comparable physiological activation. Voice measures - specifically f0 median and its standard deviation - were largely equivalent whether participants spoke to real or virtual listeners.
On oral communication parameters, filled pauses percentage was significantly higher for the real audience than the virtual audience (T=3.64, p<.001), representing the one notable difference: participants produced more filled pauses (e.g., “um,” “uh”) in front of a real audience. Total disfluencies percentage (TD%) was equivalent across conditions. Participants who scored higher on trait anxiety and who reported greater feelings of presence in VR showed the closest match to their real-audience responses.
Why this matters
This study strengthens the case for VR as a valid stand-in for real-world speaking situations. The convergence across subjective, physiological, and acoustic measures is particularly compelling, as it suggests VR does not merely feel stressful but actually engages the same vocal and bodily systems that real audiences activate. The finding that more anxious individuals respond most strongly to VR is encouraging, since these are precisely the people who stand to gain the most from graded, low-risk speaking practice.
Limitations
The sample consisted entirely of non-clinical university students, so the results may not directly extend to people with communication differences such as stuttering. All participants were native French speakers, and the single speaking context (academic presentation) limits generalizability to other situations like conversations or phone calls.
Implications for practice
The virtual audience provides a scientifically validated, lower-threshold entry point for graded speaking practice. Individuals with higher anxiety levels appear to benefit most from VR-based practice.
Implications for research
Replication in populations with communication differences is needed, along with work across different speaking contexts beyond formal presentations. Longer-term transfer effects from VR rehearsal into everyday speaking situations remain an open question.
Where this connects to Therapy withVR
The study above is independent research and does not endorse any product. The notes below are commentary from withVR on how the themes in this research relate to features of Therapy withVR. The research findings are not claims about Therapy withVR.
Meeting Room environment
This study validated virtual audiences using a Meeting Room setting (8 virtual agents around a rectangular table), not an auditorium. Therapy withVR's Meeting Room environment directly matches this validated speaking context.
Real-time clinician control
The clinician shapes the experience from their laptop while the person inside VR is immersed. The session screen is the interface where that real-time shaping happens.
Cite this study
If you reference this study in your work, the canonical citation formats are:
@article{bettahi2026,
author = {Bettahi, L. and Remacle, A. and Schyns, M. and Etienne, E. and Etienne A-M and Leclercq A-L},
title = {Validating virtual reality for public speaking research and intervention: comparing anxiety, voice, and fluency responses to real and virtual audiences},
journal = {Frontiers in Virtual Reality},
year = {2026},
doi = {10.3389/frvir.2026.1755571},
url = {https://withvr.app/evidence/studies/bettahi-2026}
}TY - JOUR
AU - Bettahi, L.
AU - Remacle, A.
AU - Schyns, M.
AU - Etienne, E.
AU - Etienne A-M
AU - Leclercq A-L
TI - Validating virtual reality for public speaking research and intervention: comparing anxiety, voice, and fluency responses to real and virtual audiences
JO - Frontiers in Virtual Reality
PY - 2026
DO - 10.3389/frvir.2026.1755571
UR - https://withvr.app/evidence/studies/bettahi-2026
ER - Know of research that should be in this hub? If a relevant peer-reviewed study is not listed here, send the reference to hello@withvr.app. The hub is kept up to date as the literature grows.
Funding & independence
FNRS grant 40021892 to first author Lamia Bettahi (FRESH grantee). No withVR BV involvement in funding, study design, or authorship. Summary prepared independently by withVR using the published paper.