Gaze points; pinch commits
Aliases: look and pinch · gaze nominate pinch confirm · eye-hand split
What it is
The eyes rest on a button across the room. Thumb and index close at the hip. The button takes. Gaze and pinch splits one selection into two channels: gaze points, pinch commits. The eye is the pointer, the hand is the key, and they need not aim at the same place.
This is not “click with the eyes,” and not “point with the hand.” The eyes only nominate; the hand only promises. The split is the input.
Why it happens
Gaze is already a high-rate angular pointer: the globe turns faster than the arm, costs little energy, and the target is almost always near the fovea. What it lacks is a discrete “this one” — gaze sweeps many objects continuously and has no natural press. Pinch supplies the press: two fingers closing is a boolean the hand tracker can fire, without the hand reaching the target, without the hand even being in view.
Far-field selection then need not take an arm-cast ray, nor move controls inside arm’s length. The eye aims in angular space; the hand does the key’s job nearby. The channels run on different clocks: gaze can already be parked on the next target, pinch fires on the prepared frame. It is the mouse split — move / click — with move given to the eye and click given to a small action of the hand.
Studying it
On the same far-field targets, compare gaze-plus-pinch, head-plus-pinch, and hand-ray-plus-button. Gaze needs an eye tracker; head uses headset pose in place of gaze, to separate foveal pointing from face direction.
Independent variables: pointing channel (gaze / head / hand), commit channel (pinch / button / dwell), target visual angle. Dependent variables: selection time, error rate, hand displacement, the eye–hand lag (gaze arrives, pinch follows).
Eye–hand lag is the signature that the split happened. If pinch consistently fires tens to hundreds of milliseconds after gaze arrival, people are using two channels. If they first raise the hand to aim and then pinch, they are pointing with the hand and gaze is idle.
Where it stops holding
When eye tracking is inaccurate, calibration has drifted, or the user has strabismus or monocular vision, the pointing channel fails first and a perfect pinch still hits the wrong target. Devices without hand tracking cannot pinch and must fall back to dwell or a key. In reading and search, gaze is the task; using the same gaze to hit controls fights itself — the eyes must scan text, not double as a pointer. Pinch needs two fingers visible or at least estimable; hands in pockets delete the commit channel. Sparse, clean lab targets make gaze selection look startlingly fast; crowded widgets on a moving background eat the advantage with pointing noise.
Applying it
- For sparse far-field choices, nominate with gaze and confirm with pinch. Do not also require a hand ray on the same target.
- Let the pinch complete near the body; the hand should not have to rise into view, or the split collapses into “the hand both points and presses.”
- Drive pointing feedback (highlight, cursor) from gaze, not from hand position, or people will raise a hand to hunt the pointer.
- How to check: record eyes and hands. In a successful selection, gaze should arrive first and the hand should still be near the torso at pinch. If the hand is already reaching toward the target before the pinch, the split was not used.