Change blindness
People miss large changes to a scene when the change coincides with a visual disruption.
A substantial change to a visual scene goes unnoticed if it happens during a blink, a saccade, a flicker or an occlusion. Without a motion transient to capture attention, detection depends on where attention already was. The failure is about comparison across a disruption, not about the change being too small to see.
How it shows up in software
Any change that happens off-screen, during a page transition or under a spinner is a candidate to be missed: a silently updated total, a filter that reset, a price that changed between the cart and the checkout. Single page applications make this worse, because the full page reload that used to signal a change is gone.
Using it well
- Animate changes into place rather than swapping them, so the motion transient does the signalling.
- Put confirmations next to the thing that changed, not in a corner toast that appears where nobody is looking.
- After a state change, say what changed in words: three filters cleared, total updated to include shipping.
- Never change a number the person has already reviewed without an explicit acknowledgement step.
Where it turns manipulative
- Adjusting a total, a delivery date or an interest rate under a loading state, so the new value is technically shown but practically unseen.
- Silently adding an item, a warranty or a donation to a cart during a transition.
- Changing the destination of a button between the moment a person reads it and the moment they click.
Where you have seen it
Google Docs
Remote edits arrive with the collaborator's cursor and a coloured highlight, so the change carries its own transient.
GitHub
If the base branch moves while a pull request is open, the page states that the diff has changed rather than silently re-rendering.
Linear
Status changes animate the item between columns instead of re-sorting the board underneath the pointer.
What the research says
- Rensink, O'Regan and Clark, 1997Well evidenced
Alternating two images separated by a brief blank field made large changes very hard to spot, sometimes taking dozens of alternations, though they were obvious once found.
- Simons and Levin, 1998Well evidenced
An experimenter asking for directions was swapped for a different person behind a passing door. Around half of participants did not notice the substitution.
Small field samples. The flicker paradigm result is the more heavily replicated of the two.
Grades are a judgement about the evidence, not about how useful the idea is. Plenty of contested effects are still worth knowing, as long as you do not cite them as settled.