Status changes must move, not snap
Aliases: animated state change · snap update · flicker
What it is
The status goes from A to B, the pixels already changed, and the person may still believe it is A. Often the display did change; what failed is the transition — the old value was replaced in place, with no displacement, no fade, no few frames of contrast. That detectable brief disturbance is a status change transient. A snap codes “then” and “now” as two unrelated stills; a transition codes them as one event.
The claim is how a change is seen, not which patch of the UI holds the status, and not whether the system declined to change the picture at all. If the picture did change but it looks like a different still, this already failed.
Why it happens
Vision is tuned to sudden motion transients: an edge moving, a luminance step, a local flicker orients attention there. A label that flips from “Saved” to “Syncing” in the same pixels, with no transient, has to be caught by memory comparison — last glance versus this glance. Memory comparison is slow and brittle; if the eyes were elsewhere, or a saccade sat in the middle, comparison fails. The product of that failure is change blindness: the change happened, the report is “nothing changed”.
The transition is not decoration. It translates “the value swapped” into a signal memory comparison cannot replace: 150–300 ms in which old and new are co-present (cross-fade, slide, numeral roll). Short enough and it does not become a second event; zero transient and a status control that never leaves its home can still swallow the change.
Studying it
Run a UI version of the flicker paradigm: insert a brief full-screen flash on the frame the status label switches, or time the switch into a saccade, and ask whether save-state changed. Contrast that with the same switch given a 200 ms cross-fade. Changes at the centre of interest are easier to catch, so run a second condition with the status control off the fixation point — closer to use: the person is reading the body, the chrome changes on its own.
Independent variables: presence of a transition, its duration, whether the change sits at current fixation, presence of a visual interruption. Dependent variables: detection rate, time to notice, false confidence that “it was always this”.
Lab participants know a change is coming and hunt for it; detection runs higher than in a product. Product snaps happen while someone reads or fills a form. Paradigm percentages are not an acceptance line.
Where it stops holding
Alarms, safety interlocks, a permission yanked: a transition spends a couple of hundred milliseconds, and soft motion can be read as “still in progress” rather than “already flipped”. Those switches hard-cut, then cover detection with a channel that cannot be ignored. A high-frequency heartbeat (online count refreshing every second) that plays 200 ms of motion each time turns the transient into noise; people learn to ignore the whole channel. What should change then is the update cadence, not a mandatory animation on every tick. Colour may carry no transient under colour-vision deficiency; a change of position or shape is more reliable than hue. Users who suppress motion, or who are vestibular-sensitive, take the transition near zero; detection then rides a brief textual diff (“went from Saved to Syncing”) or a non-visual cue, not a insisted-on displacement.
Applying it
- Give decision-changing status switches a 150–300 ms visible transition: save, connection, sync, permission, conflict. Do not only swap the string.
- The transition must start on the beat the value changes. An animation that waits for a full-page redraw is a second event, not this change’s transient.
- Numerals and short labels at the same locus: roll or cross-fade. Do not “also” change status by refreshing the whole page.
- How to check: cut the network from up to down, then step the recording frame by frame on the connection cue. If two adjacent frames already show the new word and no frame holds both, that is a snap. Ask a colleague who was not staring at the control what just happened. “Didn’t notice”, while the pixels did change, means the transient was too thin.