Gaze-Dwell Selection

Gaze-dwell selection is an XR interaction technique in which a user activates a button, menu item, or interactive element by directing their gaze at it and holding their gaze there for a defined dwell time — typically 0.6 to 2.0 seconds — without any additional input gesture. It is the interaction primitive of last resort for hands-free operation: requiring no controller, no hand gesture, no voice command, only the direction in which the user is looking.

Mechanism

When eye or head gaze enters the bounding area of an interactive element, a timer begins. Visual feedback — typically an animated progress ring or filling bar on the target element — shows the user how much of the dwell time has elapsed. When the timer completes, the selection fires. If gaze leaves the target before the timer completes, the timer resets.

Head gaze dwell uses the head's forward direction as the pointing vector, driven by the headset's IMU rather than dedicated eye tracking cameras. It is lower precision than eye gaze (approximately ±2–5 degrees vs. ±0.5–1.5 degrees) but is available on every headset including those without eye tracking hardware. Microsoft HoloLens 1 used head gaze dwell as its primary hands-free interaction mechanism before HoloLens 2 introduced hand tracking.¹

Eye gaze dwell uses the precise gaze direction from eye tracking hardware, enabling smaller targets and faster selection times — but the precision increases the difficulty of casual browsing, as eyes naturally scan across content and may trigger unintended dwells.

The Midas Touch Problem

The fundamental design tension of any gaze-based activation system is the Midas Touch problem, named by Robert Jacob in a 1990 CHI paper: whatever you look at gets selected.² Eyes move constantly during normal reading and scene exploration, and the activation system must distinguish intentional dwell (the user wants to select this) from incidental fixation (the user is simply looking at it). Dwell time thresholds are the primary mechanism: a sufficiently long dwell is unlikely to be accidental. But longer dwell times slow down the interaction and increase frustration; shorter dwell times increase accidental activations.

Practical design guidance converges on 0.6–1.0 seconds as the sweet spot for standard button selection, with longer thresholds (1.5–2.0s) for irreversible or destructive actions.¹

Feedback Design

Because dwell selection fires without any deliberate physical action from the user, clear progress feedback is essential — it is the only signal that the system has received the gaze and is counting toward activation. The animation must be legible at the target size and reading distance, not obscure the target's label, and complete at a rate that feels responsive without feeling rushed. A filling arc around the target perimeter is the most common pattern; a colour shift or scale animation serves as a secondary confirmation cue at completion.¹

Use Cases

Gaze-dwell is most appropriate in:

Hands-free industrial AR: factory workers with gloved hands or occupied hands using AR for guided assembly, inspection, or navigation
Accessibility: users with limited hand mobility for whom gestures and controllers are impractical
Sterile environments: surgical or laboratory settings where controllers cannot be used
Passive content consumption: navigating menus while hands are otherwise engaged

It is least appropriate as a primary interaction mechanism in content-rich interfaces where the user's gaze moves frequently across many candidate elements, or in fast-paced interactive applications where the dwell delay introduces unacceptable latency.

Combined with Voice

The most effective hands-free interaction model combines gaze dwell for targeting with voice for confirmation: gaze at a target to highlight it, say "select" to activate. This separates the targeting from the activation, eliminates the Midas Touch problem entirely (voice confirms intent), and enables faster selection than waiting for a dwell timer. Apple Vision Pro uses the converse — gaze for targeting, hand pinch for confirmation — which is equally effective for users with full hand mobility.³