Ecological Validity
Research exploring why virtual speaking situations feel real and produce communicative responses equivalent to real-world interactions.
The most fundamental question in VR for speech therapy is whether the experience feels real enough to matter.
Ecological validity research examines whether virtual environments produce the same communicative responses (physiological, emotional, and behavioral) as equivalent real-world situations. If they do, then practice in VR transfers meaningfully to everyday life. If they do not, VR is just an expensive distraction.
The evidence strongly supports ecological validity. The strongest finding comes from Brundage and Hancock (2015), who demonstrated a near-perfect correlation (rho = 0.99) between speech behavior during a challenging virtual audience speech and a live audience speech. Communication apprehension and speaker confidence showed similarly strong correlations.
Subsequent studies have reinforced this finding across different populations and contexts, demonstrating that vocal intensity adjusts to virtual listener distance, that anxiety responses in VR match those in equivalent real situations, and that people consistently report feeling genuinely present during virtual speaking situations.
A useful corrective to the assumption that “more photoreal equals more therapeutic” comes from work on layered realism. Communicative realism (how a character moves, gestures, looks back, and responds in real time) tends to matter more than anthropometric realism (how anatomically accurate the body proportions are), which in turn tends to matter more than photorealism (how lifelike the surface textures look). Cartoonists figured this out decades before VR did. For clinical practice, this means a virtual speaking situation does not need to look like a film studio to produce genuine communicative responses; what matters is whether the audience appears to be paying attention, looking back, and reacting in real time.
This body of work provides the foundation for everything else in the Evidence Hub. VR works for speech therapy because the experiences it creates are real enough to produce genuine communicative behavior.
21 Studies
RCT (n=47) - three short VR sessions helped autistic teens and adults respond more effectively in live police encounters, vs matched video
An RCT randomized 47 verbally fluent autistic participants (ages 12-60) to Floreo's VR Police Safety Module or BeSAFE The Movie video modeling: three 45-minute sessions each, with ~12 minutes of active VR practice per session. The VR group gave significantly more appropriate responses and showed calmer body language during live post-intervention interactions with real officers; the video-modeling group did not. Both groups reported greater knowledge and comfort with police interactions after training.
First scoping review of immersive VR in speech-language pathology maps two decades of research
A scoping review in AJSLP synthesized 11 peer-reviewed studies (2007-2025) using immersive VR with people who have communication differences. Across populations, immersive VR elicited communicative and emotional responses comparable to real-world contexts, with consistent presence and engagement findings. Sample sizes ranged from 3 to 36 participants, ages 9-81 years. The review describes the field as still early-stage and disproportionately focused on stuttering (5 of 11 studies).
Pilot of Immersive VoiceSpace VR (N=17, vocally healthy plus people with dysphonia) - participants scaled loudness and pitch across graded virtual restaurant conditions
A within-subjects pilot of Immersive VoiceSpace (IVS), a custom VR voice-training platform developed by the sole author. Seventeen adults (10 vocally healthy speakers and 7 people with dysphonia) completed a menu-ordering task in a virtual restaurant under four conditions - a baseline plus three graded IVS levels manipulating avatar distance, voice-activation thresholds, and walkaway timeouts. Sound pressure level and mean speaking f0 increased significantly across IVS levels in both groups; pitch flexibility was more constrained in the dysphonia group. Feasibility ratings were good overall (4.0/5), with comfort and safety excellent (4.5/5) and no cybersickness reported.
Virtual audiences trigger real anxiety and comparable voice responses
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.
Brief VR voice therapy with clinician feedback elicited teaching-style prosody in pre-professional teachers - but also significantly increased reported vocal discomfort
A single-session within-subjects pilot with 10 pre-professional teachers (9 analyzed). Both a teaching-style mock lesson and a clinician-controlled VR teaching intervention elicited teaching-style prosody vs a conversation control. CTT-style clinician feedback inside VR produced short-term modulations in SPL, fo, and Dt%. Critically, the VR condition also significantly increased self-reported vocal discomfort vs control (+20.5 VAS, p=.023) - a caveat that should travel with any clinical citation.
Pilot RCT in youth who stutter: real-time photorealistic-avatar VR is well-accepted and elicits arousal, but one session did not outperform SLP role-play
A pilot RCT randomized 12 children and adolescents who stutter (ages 9-18) to one of two training conditions before facing an unknown actor-teacher: a conversation with a photorealistic virtual teacher in VR controlled live by their own SLP via facial motion capture (n=6), or face-to-face SLP role-play (n=6). The VR system was well-accepted (high presence, low cybersickness). Skin conductance was elevated from baseline in the VR group; SLP role-play raised self-reported anxiety more. A single session did not outperform role-play on self-efficacy or post-task in vivo anxiety.
Virtual room size and listener distance influence how people use their voice
Using the Rooms module of Therapy withVR, this study found that speaker-to-listener distance in virtual environments significantly influenced vocal intensity and mean fundamental frequency, with trained singers showing more systematic vocal adjustments across distance conditions than untrained speakers.
Engineering + user-reception study (Computers & Graphics 2025) of a speech-controlled VR system for voice and public-speaking training: extracts pitch / timbre / speech rate from 529 utterances by 15 students for real-time virtual character response
An engineering and user-reception study published in Computers & Graphics special section on XRIOS 2024. Polish-British collaboration (AGH Krakow, SWPS Warsaw, Polish Academy of Science, Kielce University of Technology, University of Cambridge). The system is built on a speech recordings corpus of 529 utterances during presentations by 15 students. Voice parameters extracted: pitch, timbre, speech rate. Six expert annotators evaluated stress levels per presentation. The multi-parameter analysis selects features for real-time animation of virtual characters responding dynamically to speech changes. The contribution is design and user-reception evaluation rather than clinical efficacy.
Within-subjects study in 31 vocally healthy adults: auditory, visual, and audiovisual room cues in immersive VR all measurably change self-perceived vocal loudness, effort, comfort, and acoustic output
Thirty-one vocally healthy men and women were tested under 18 sensory-input conditions in immersive virtual reality - two auditory rooms with different reverberation times, two visual rooms with different volumes, and audiovisual combinations - each with and without background noise. Speakers performed counting, sustained vowels, an all-voiced CAPE-V sentence, and a Rainbow Passage sentence. Self-perceived vocal loudness and effort INCREASED, and self-perceived vocal comfort DECREASED, as room volume, speaker-to-listener distance, audiovisual richness, and background noise increased. Sound pressure level (SPL) and spectral moments (mean, SD, skewness, kurtosis) showed concomitant changes. Visual and audiovisual input - not just auditory - measurably shaped voice production.
Three-arm RCT (n=51) of stand-alone 360° video VR exposure therapy for public speaking anxiety: both audience-content and empty-room content produced significant pre-to-post reductions (partial η² up to .90) versus no-treatment control
Fifty-one participants with high public speaking anxiety were randomly allocated to one of three conditions: 360° video VRET incorporating audience stimuli (n=17), 360° video VRET incorporating empty-room stimuli (n=16), or no-treatment control (n=18). Outcomes measured at five timepoints. Mixed-ANOVA revealed a significant time × intervention-group interaction for PSA, social anxiety, and fear of negative evaluation (FNE). Both 360° VRET groups showed large pre-to-post reductions; for PSA, partial η² = .90 (audience) and .71 (empty room). Active intervention participants showed continued significant improvement out to 10-week follow-up. The study also addresses whether 360° video content (audience vs empty room) affects VRET outcomes - both worked.
In ten children and adolescents who stutter, virtual classrooms produced anxiety and rated stuttering severity comparable to a live audience
Ten school-age children and adolescents who stutter spoke under three conditions: an empty virtual apartment, a virtual classroom (neutral and challenging variants), and a small live audience. Self-reported anxiety and clinician-rated stuttering severity in the virtual classroom did not differ significantly from the live audience, and virtual-classroom anxiety correlated strongly with live-audience anxiety (Spearman rho = 0.92, p < .001).
Stuttering adaptation is stronger in VR than in real-world settings
This study examined whether people who stutter show the expected decline in stuttering across repeated readings in VR compared to real-world settings. Twenty-four adults completed tasks in both environments, and the adaptation effect was actually stronger in VR.
A VR classroom successfully brings out how teachers really use their voice when teaching
A THREE-condition experiment: teachers delivered lessons in a real classroom (in vivo), in a virtual classroom (in virtuo), and in a free speech control situation. Voice parameters (intensity, pitch, intonation) and temporal measures (pause duration) were compared across all three. The virtual classroom successfully elicited teaching-voice characteristics matching in-vivo teaching, providing validation that VR can substitute for real classrooms in voice research and support.
Umbrella review: clinical VR has matured into a viable tool, with caveats clinicians should know
An umbrella review by two senior figures in clinical VR examined the breadth of evidence across psychological and neurocognitive applications, concluding that VR is ready for routine clinical use in many contexts while flagging implementation challenges that practitioners should plan for.
Stuttering and anxiety responses in virtual audiences closely correspond to those in live audiences
A foundational study in ten adults who stutter showing that stuttering frequency during a challenging virtual audience speech correlated at Spearman rho = 0.99 with stuttering during a live audience speech, and that anticipatory apprehension and confidence measured before the virtual condition correlated strongly with the same measures before the live condition (rho = 0.82 and 0.88 respectively). The neutral virtual audience also correlated with the live condition, but less strongly (rho = 0.82 for stuttering frequency).
Can VR elicit SAD-typical distress? In 21 SAD adults + 24 non-anxious controls, VR public-speaking task elicited significant physiological + subjective distress over baseline - but less than the in-vivo task; no SAD-vs-control group differences on physiology
Twenty-one adults with social anxiety disorder (SAD) and 24 non-anxious controls each gave an impromptu speech in front of an in-vivo (live) audience AND in front of a virtual reality audience. Outcomes: heart rate, electrodermal activity, respiratory sinus arrhythmia, and self-reported distress, plus sense of presence ratings. Results: VR significantly increased heart rate, electrodermal activity, RSA, and subjective distress over baseline - but less than the in-vivo task. Participants reported moderate presence in VR, but significantly less than in-vivo. NO significant SAD-vs-control group differences on physiological measures. The study addresses VR's realism-and-validity for SAD clinical work.
A virtual library task picks up executive function differences after TBI that paper-and-pencil tests miss
A newly developed Virtual Library Task assessed seven components of executive function in 30 adults with traumatic brain injury and 30 matched controls. The TBI group performed worse across multiple components, with the virtual task showing better real-world correlations than traditional measures.
First-person perspective alone transfers body ownership in VR - synchronous touch not required
A 2×2×2 experiment with 24 male participants showed that first-person perspective (seeing through the virtual body's eyes) alone was sufficient to produce body ownership over a virtual female body - synchronous visuotactile stimulation added little in first-person conditions. The synchronous-touch manipulation only mattered when perspective was third-person. The study established that perspective position is the primary driver of virtual body ownership.
Three validation studies (n=40) - stuttering behavior, anxiety, and cortisol in VR mirrored real speaking situations
This paper makes the conceptual and empirical case for integrating VR into stuttering assessment and support. It presents three validation studies showing that stuttering behavior, anxiety, and physiological stress in VR are comparable to real-world speaking, positioning VR as a bridge between the therapy room and everyday life.
VR job interviews show interviewer style affects stuttering frequency; %SS in VR correlates strongly with %SS in a clinical SSI-3 interview
Twenty adults who stutter completed virtual job interviews in two conditions (challenging and supportive). %SS was significantly higher in the challenging condition. %SS in both VR conditions correlated strongly with %SS in a separate clinical SSI-3 interview task (r=.90+). The comparison was VR-vs-clinical-interview, NOT VR-vs-real-world job interview - the authors are explicit that real-world comparison was future work.
Early VR study (n=10) - adults with TBI completed fewer virtual errands than controls; virtual performance matched real multitasking
Adults with traumatic brain injury and matched controls completed a Virtual Errands Test set in a virtual university building. The TBI group completed significantly fewer errands, and virtual performance correlated with real-world errands at the same site.
See the Software for Yourself
Whether you have questions, want to see the software, or are ready to get started - help is always available.
Get in TouchNo obligation - see the software before you commit