Eye-tracking VR helps people who stutter improve gaze during conversation

Background

Gaze avoidance - looking away from a listener’s face during moments of disfluency - is a common experience among people who stutter. It can affect how conversations feel for both speakers and listeners, and is often something people want to change. However, measuring gaze behavior has traditionally relied on external video analysis or subjective clinician ratings, both of which are time-consuming and imprecise. With eye-tracking sensors becoming available inside VR headsets, Walkom saw an opportunity to capture gaze data automatically and use it as a feedback tool.

What the researchers did

Twelve adults who stutter completed three VR sessions each lasting approximately six minutes, spaced across several weeks. A matched group of fluent speakers completed the same task as a baseline comparison. The VR system used a Fove 0 head-mounted display, one of the first consumer-grade HMDs with integrated eye-tracking capability. In each session, participants spoke to a virtual avatar in a one-to-one conversational scenario while the Fove 0’s eye tracker recorded fixation data in real time. After each session, participants were shown a visual summary of their own gaze patterns - where they had looked, for how long, and how their gaze shifted during the conversation. Self-reported distress was also measured at multiple points.

What they found

The gaze data revealed clear changes over the three sessions. People who stutter spent significantly more time looking at the avatar’s face by the final session, rising from 48% to 68% of total speaking time. Gaze avoidance, measured by time spent fixating the back wall, dropped markedly from 17% to just 5%. Blink frequency decreased by 31% and blink duration by 41%, suggesting a reduction in prolonged eye closures that often accompany moments of stuttering. Self-reported distress also decreased across sessions. Fluent speakers showed relatively stable gaze patterns across sessions; the divergence from the stuttering group’s trajectory over the three sessions is consistent with genuine behavioral change rather than pure familiarization, though the quasi-experimental design limits causal inference.

Why this matters

This research demonstrated that eye tracking embedded in VR headsets can replace subjective gaze observation with precise, quantifiable data. Perhaps more importantly, it showed that sharing this data with participants created a feedback loop - people could see their own avoidance patterns objectively and track their progress over time. This approach respects individual autonomy by giving people concrete information about their own behavior, empowering them to set and monitor their own goals.

Limitations

The sample size was modest at 12 participants per group, and the study ran for only three sessions, so it is unclear whether the gaze improvements would persist long term. The conversational scenario involved a single avatar, which may not capture the complexity of multi-person interactions. The study did not assess whether changes in gaze behavior carried over into real-world conversations outside the VR environment.