VR self-modeling improved conversational stuttering but had limited effects on prompted speech and anxiety

Background

Self-modeling - watching edited video of yourself performing a skill successfully - has a long evidence base in areas such as sports and social skills development. Earlier research by Cream and colleagues showed that traditional video self-modeling could improve self-perception for people who stutter, but those videos were flat and viewed on a regular screen. This doctoral study asked whether immersing someone in 360-degree VR footage of themselves speaking fluently might produce even stronger effects.

What the researchers did

Three adults who stutter were filmed speaking in personally meaningful and challenging situations - the kinds of everyday contexts where stuttering typically increased for them. The footage was carefully edited to remove moments of stuttering, creating 360-degree VR videos that showed each person communicating fluently. Participants then watched these videos through an Oculus Go headset at home over a series of sessions. The entire process was delivered remotely, with no in-person clinician involvement required. Stuttering frequency was measured repeatedly before, during, and after the viewing period, using a multiple-baseline design.

What they found

Effects varied meaningfully across outcome types - a distinction the dissertation abstract states directly: “all participants demonstrated clinically meaningful reductions in their conversational stuttering severity, however limited treatment effects were found among prompted speech and anxiety data.”

Conversational (unprompted) stuttering was the clearest success: PND values of 93.75%, 95.83%, and 100% for the three participants indicate uniformly large and reliable reductions. Prompted speech showed more variable results: Participant 1’s data had high variability with an upward trend that exceeded baseline levels during the intervention and follow-up phases, making that finding questionable; Participants 2 and 3 showed PND of 100%.

State anxiety results were highly uneven: only Participant 3 showed meaningful reduction (PND 100%). Participant 1’s anxiety actually increased during the intervention (PND 13.33%). Participant 2 showed minimal change (PND 4.55%). Client satisfaction followed a similar pattern - only Participant 3 was uniformly satisfied; Participants 1 and 2 had more mixed experiences.

Importantly, data collection took place during the COVID-19 pandemic and a period of significant racial unrest in the United States. The author explicitly identifies these external events as likely confounders, particularly for anxiety data, as participants reported that these events affected their emotional state during the study period.

Why this matters

This study is valuable precisely because it reports mixed findings honestly. The strong effects on conversational speech suggest VR self-modeling can meaningfully change day-to-day communication. The limited and variable effects on anxiety and prompted speech provide important calibration for clinical expectations. The remote delivery model - Oculus Go headset at home without clinician presence during sessions - demonstrates practical feasibility, but the between-participant variability in anxiety response suggests that not all people who stutter will benefit equally from this approach.

Limitations

With only three participants, the findings cannot be generalized. The one-week follow-up is too short to assess durability. The COVID-19 and racial unrest context is an acknowledged confounder. The editing process to create fluent VR footage requires specialized effort. Participant 1’s inconsistent response across outcome types raises questions about individual factors that moderated treatment response.