No grip is needed, but there is no haptic confirmation
Aliases: no tactile confirm · empty-hand input · ungrounded hands
What it is
The fingers close in empty air. The virtual button should have registered. The skin reports nothing. Bare-hand tracking needs no grip, but there is no haptic confirmation: the hand is free to take a cup or reseat a headset, and that same empty hand receives no cutaneous evidence of “you crossed the line.” Confirmation is dumped onto eyes and ears.
This is not a tracking-quality claim. The hand can be seen steadily and the pinch classified steadily; the gap remains. There is no key travel, no motor, no surface normal. Freedom of an empty hand is paid for with the tactile loop.
Why it happens
A physical key does two things at once: displacement past a threshold, and pressure plus restoring force on the pad. Bare hands keep only the displacement. Vision is drafted as the judge — did the button recess, did the highlight fire — and that judge samples slowly and occupies the same gaze that is still aiming. People over-pinch: squeeze harder, hold longer, even drive a finger into the panel, recruiting proprioception to impersonate the missing reaction force.
The gain is real. With no shell in the palm, the hand can handle a real object at once, or take shapes no controller key can express. The gain is bound to the loss: any action that needs “I know I hit it” is missing its shortest path. Putting confirmation only in vision asks an already busy eye to do quality control as well.
Studying it
On the same targets, compare bare-hand pinch with controller click, scoring not speed but what evidence people use to decide “it took.”
Independent variables: presence of mechanical travel, presence of a short sound, amplitude and delay of a visual press animation. Dependent variables: time spent adding force after the pinch crossed threshold, repeat triggers, glances away from the target to inspect feedback, verbal reports of “how I knew it registered.”
Add a condition that briefly hides the virtual hand during the click. If confirmation collapses once the visual hand is gone, the loop was almost entirely visual. That locates the missing channel better than a satisfaction score.
Where it stops holding
When a fingertip actually meets a desk or the other hand, reality supplies the confirmation and the gap locally closes — which is why some people “click” virtual buttons against a real tabletop. A vibrating ring or thimble adds a weak channel, with different bandwidth from a controller motor; it is not a full substitute. Sound-for-skin fails when hearing is impaired or the room is loud. If every action is cheap and reversible, people can work from outcomes (“the door opened”) rather than instant confirmation; high-cost commits cannot.
Applying it
- Bare-hand commits need a state flip readable within a frame, not a slight shift of a 3D model.
- Pair the commit with a short sound locked to the visual change, so the eyes can stay on the next step.
- Do not hang a high-cost action on a single pinch; add a visible second step (pinch again, drag into a slot) so “did it take?” becomes a process rather than an instant sensation.
- Do not imply the hand will meet resistance. A panel the finger sails through, or a button that does not follow the pad, is worse than no tactility — it promised a channel it cannot keep.
- How to check: mute audio and shrink the press animation to just-noticeable. Count repeat pinches and how often someone asks “did that register?” Each is a haptic gap no other channel caught.