In ten children and adolescents who stutter, virtual classrooms produced anxiety and rated stuttering severity comparable to a live audience

Background

Classroom presentations are among the most anxiety-provoking experiences for young people who stutter. Avoidance of these situations can affect academic participation and social confidence. If virtual classrooms could reliably evoke the same emotional and communicative responses as real ones, they could provide a practical way for young people to build familiarity and confidence through repeated, controlled practice.

What the researchers did

Ten children and adolescents who stutter (eight boys, two girls; ages 9-17, mean 12 years), all currently in stuttering treatment at the Marie Enfant Rehabilitation Center in Montreal, gave 4-minute speeches under four conditions across two sessions (real-audience session and virtual-reality session, with order counterbalanced):

• A baseline conversation with the examiner alone
• An empty virtual apartment (control condition, viewed through an Oculus Rift head-mounted display)
• A “neutral” virtual classroom - seven avatar students and a teacher who looked attentively at the speaker but were neither distracting nor supportive
• A “challenging” virtual classroom - same setup but with avatar reactions (yawning, falling asleep, laughing at the speaker, and an unhappy teacher) triggered by the experimenter at fixed timepoints
• A small live audience - four other participants who stutter plus two adults conducting the study (NOT a real classroom or peer group)

Self-reported anxiety was measured on a 0-10 Subjective Units of Distress Scale (SUDS) at the beginning of each speech, after 30 seconds, and at the end of every minute. Stuttering was measured three ways: a 0-9 stuttering severity rating (SR) by a blinded speech-language pathologist viewing the video, a 0-9 SR by the participant themselves, and percent syllables stuttered (%SS) on the first 200 syllables of each speech. Sense of presence and cybersickness were measured with the ITC-SOPI questionnaire.

The virtual classroom software was developed by the research team in collaboration with InVirtuo, Inc. and ran on an Oculus Rift headset.

What they found

The Wilcoxon signed-rank test found no significant difference between the virtual-classroom condition and the live-audience condition on SUDS ratings (Z = 0.18, p = .86) or on SLP-rated stuttering severity (Z = 0.68, p = .49). Both conditions produced significantly higher anxiety than the empty virtual apartment (SUDS: Z = 2.50, p = .013). The Spearman rank-order correlation between virtual-classroom SUDS and live-audience SUDS was very strong (rho = 0.92, p < .001) - and notably, virtual-classroom SUDS did NOT correlate significantly with the SUDS reported while talking to the examiner alone (rho = 0.49, p = .16), suggesting that the virtual classroom is specifically eliciting audience-related anxiety, not generic VR-immersion anxiety.

The more conventional behavioral measure of stuttering - percent syllables stuttered - did not differ significantly across any of the conditions (Friedman Chi-squared = 2.46, p = .29). The authors note this is consistent with the wider literature showing mixed results for whether %SS responds to changes in speaking situation.

Anticipatory anxiety - SUDS ratings during the 3-minute preparation phase before each speech - was significantly higher for the live audience (M = 2.7, SD = 2.31) than for the virtual classroom (M = 1.59, SD = 1.81; Z = 2.52, p = .01). The authors flag this as a potential clinical advantage of VR-based exposure: it may lower the barrier to engagement for young people who would otherwise refuse to attempt the situation.

Contrary to expectations, the neutral and challenging virtual classroom conditions did not differ significantly from each other on any measure. The authors hypothesize this may be due to (a) participants having low overall social anxiety in this sample (LSAS-CA mean 30/144), (b) the fixed presentation order with neutral always before challenging (which may have allowed fear extinction before the challenging condition), and (c) the limited number of audience reactions in the 4-minute speech. They note this finding is consistent with other studies that have struggled to differentiate audience attitudes in VR.

Participants reported acceptable presence and immersion (ITC-SOPI subscales above the midpoint) and low negative effects; 70% rated the virtual classroom as very useful (7 or more on a 0-10 scale).

Why this matters

This is the first study to extend Brundage et al.’s VR-and-stuttering work from adults to school-age children and adolescents. The finding that virtual classrooms produced self-reported anxiety and rated stuttering severity comparable to a small live audience supports the feasibility of using virtual classrooms as an exposure context for cognitive-behavioral therapy in young people who stutter. The lower anticipatory anxiety in the virtual condition may also be clinically useful - it could reduce avoidance and make it more likely that a young person attempts a feared situation.

The null finding on %SS is also clinically relevant: clinicians using VR exposure with children who stutter should not expect or rely on changes in percent syllables stuttered as a primary outcome of within-session exposure. Rating-based stuttering measures and self-reported anxiety are more sensitive to situational effects in this population.

It is important to read the study as a validity/feasibility study, not as evidence that VR-based exposure reduces anxiety over time. The study did not include follow-up assessments or compare VR exposure against any treatment.

Limitations

The authors flag the following in their own discussion:

• Small sample (n = 10) limits how broadly findings should be interpreted; the authors call for larger samples.
• Low overall anxiety in the sample. Participants did not rate themselves as having high social anxiety, which may have contributed to small absolute differences between conditions and the lack of a neutral-vs-challenging classroom effect.
• Fixed order of neutral before challenging virtual classroom (not counterbalanced within the virtual session). Pre-tests suggested this order was clinically and ethically more acceptable, but it means a fear-extinction explanation cannot be ruled out for the absence of a neutral-vs-challenging difference.
• No physiological measures of arousal (e.g., heart rate variability, skin conductance, cortisol) to complement the self-report SUDS.
• Anxiety ratings collected every 30-60 seconds during the speech, which may have interfered with immersion in the virtual environment.
• The control condition was a virtual empty apartment, not an empty virtual classroom. This means the study cannot fully separate “presence of audience” from “presence of a classroom context”, though the strong virtual-versus-live correlation and the lack of correlation with the examiner-only condition partially address this.
• The “real audience” was not a real classroom - it was a group of four other PWS plus two researchers. Generalization to actual classroom presentations remains untested.
• No control group of children who do not stutter would be needed to determine whether the responses observed are specific to the PWS population.
• No follow-up was conducted to determine whether VR exposure leads to lasting reductions in anxiety or stuttering.