Elements that enter and leave the view with head motion need hysteresis or they flicker

What it is

A world-locked label sitting on the rim of the view will, with each tremor of the head, cut between seen and not seen. A single threshold turns that rim into a switch: in means draw, out means delete. Visibility hysteresis replaces one line with two — the element must travel further in to appear, and further out to vanish — so edge content stays stable through physiological jitter instead of flashing with every micro-movement. The content is not wrong; the enter/leave rule is too thin.

Why it happens

A “still” head still carries a few degrees of tremor and pose-estimator noise, and the field boundary is a geometrically sharp edge. Once a content centroid sits on that edge, in-view flips every frame, taking drawing, enter animation, and spatial audio enable with it. The periphery is especially sensitive to flicker, so the rim is noisier than the centre. Hysteresis is a Schmitt trigger: the appear threshold sits further inside the field than the disappear threshold, and the band between keeps the last state. A short time delay can be stacked on top so a fast sweep does not play every label’s entrance. Fade without hysteresis is not enough — alpha near the threshold is still reopened every frame, only more softly.

Studying it

Aim people at a label parked at an adjustable eccentricity. Record show/hide counts and entrance-animation triggers while the head is still and while it makes small deliberate yaws.

Independent variables: split versus single threshold, angular width of the split, time delay on or off, entrance animation duration. Dependent variables: flips per second, rated flicker, false entrance plays, late shows (head already on target, label still absent).

Joint eye–head recording separates a true sweep across the boundary from noise flipping the threshold. Compare instant switching, fade only, angular hysteresis, and angular-plus-temporal hysteresis.

Where it stops holding

When content must appear the frame it enters (a safety alarm), the appear threshold may sit on the rim, but disappear should still carry hysteresis so the alarm does not flash itself. Fully head-locked elements never enter or leave the view; hysteresis has no object. On a very wide field, with labels much smaller than tremor, the chance of sitting on the boundary is low and hysteresis is almost unmeasurable. Too much hysteresis and people stare at empty space while the label withholds itself — it will be read as loss, not as anti-flicker. Jumps from tracking loss are a different chain; hysteresis will not repair them.

Applying it

• For world-locked or body-locked elements that can sweep the field boundary, use two eccentricity thresholds with a gap larger than still-head tremor.
• Play entrance animation only after a transition from hidden to shown has been stable for tens of milliseconds. Do not retrigger in the threshold band.
• Safety alarms may appear fast and disappear slow; the two thresholds are allowed to be asymmetric.
• How to check: park a label on the rim, hold the head still for ten seconds, then yaw it slightly. If the label flashes or the entrance sound retriggers, hysteresis is short. If it still withholds after the head is centred, hysteresis is too wide.

Related

• Same group: N3.07.1 Edge content needs a head turn to be seen clearly · N3.07.2 Critical information must not sit at the edge · N3.07.3 The edge is fit for presence cues · N3.07.4 Pull edge elements inward to leave margin for optical falloff · N3.07.5 A ring layout fits the shape of the field of view better than a rectangle · N3.07.6 Content outside the binocular overlap is seen by only one eye
• Nearby: N1.03 Field of View · N3.08 World-locked, Body-locked, and Head-locked
• Search terms: visibility hysteresis · Schmitt trigger · HUD flicker