A hand clipping through a virtual object is the most glaring class of interpenetration

What it is

A virtual ball hangs in front of the face. A real hand reaches, does not stop at the surface, and the fingers sink in and come out the other side. Virtual hand clipping is the most glaring class of interpenetration: the hand is one’s own, it is almost always in view, and everyone knows a hand does not go through a solid. A wall or a table clipped once can be filed as “tracking drifted”. A hand through an object is immediately “this thing is fake”.

Why it happens

Body ownership treats the real hand as a near-field solid that cannot be overruled. Once the seen hand — flesh or a rendered mesh — sinks into a volume that looks solid, two high-confidence cues collide: hand pose is nailed by proprioception, the object’s surface by contour and occlusion. People do not doubt where the hand is; they demote the object to a ghost. That is worse than two panels cutting each other — panels do not carry “this is my limb”. Clipping also happens on the beat of a grasp: the hand follows its closure trajectory, the object has no force to rewrite that trajectory, so penetration is the default geometry when force feedback is missing, not a rare render bug. Nearer, larger, more graspable objects glare more; a mountain in the far scene swept by a hand is barely noticed.

Studying it

Grasp and touch tasks with objects that look solid and hands that have no physical resistance. Record penetration depth and duration, and a rating of whether the object still counts as solid. Compare visual stopping (a mesh constrained to the surface) against vibration or sound at contact.

Independent variables: visual surface constraint on or off, contact vibration on or off, object size and distance, whether the real hand is seen (see-through) or only a virtual hand. Dependent variables: peak penetration depth, time from contact to visible penetration, ownership / solidity ratings, whether people switch to a tap instead of reaching in.

Video see-through with a visible real hand glares more than a rendered virtual hand only; report them separately. Do not let a whole-session presence score bury the hand — mean IPQ can still look fine after local credibility at the hand has already dropped.

Where it stops holding

Energy shields, water, holographic scans that are already understood as passable volumes: a hand inside them matches the semantics and is not a failure. A hand deliberately drawn as a ghost cursor makes penetration cursor behaviour. With gloves or controllers and the real hand hidden, glare drops to “the controller clipped”, still ugly, no longer an attack on body ownership. Ungraspable far decoration swept by a hand is lower priority than graspable objects.

Applying it

• For objects marked solid and graspable: at contact, visually stop the real or virtual hand on the surface (offset mesh or surface projection). Do not let it sink freely.
• Add vibration or a short sound on contact so “you hit the edge” is not carried by vision alone.
• If surface constraint is impossible, make the object semantically passable, or replace the grasp with a nearby snap — avoid “looks solid, feels edgeless”.
• How to check: reach for every object labelled solid and watch the hand from the side. Any frame in which fingers emerge from the far side has already bankrupted that object’s solidity claim.

Related

• Same group: N3.10.2 Disabling the depth test to force a panel on top wrecks the whole scene · N3.10.3 Environment-depth accuracy sets how bad real–virtual clipping becomes · N3.10.4 Snapping a panel to a surface removes several clipping classes at once
• Nearby: N3.03 Depth Conflict · N2.09 Direct Grab and Virtual Touch · N5.02 Virtual–Real Occlusion
• Search terms: virtual hand clipping · interpenetration · body ownership