Von Restorff effect
An item that differs from its neighbours is remembered better than the ones that match.
Put one red word in a list of black words and the red word is recalled more often. The advantage comes from relative distinctiveness within the set, not from the property itself, so it disappears once everything is red. It is a memory effect first and an attention effect second.
How it shows up in software
Primary buttons, recommended plan badges, unread dots and error states all work by isolation. The effect sets a hard budget: one isolate per view. A page with a highlighted plan, a pulsing badge, a coloured banner and a contrast-heavy button has no isolate at all, because nothing differs from its neighbours any more.
Using it well
- Allow one visual isolate per screen and make it the action you want taken. If a second candidate appears, demote one.
- Isolate on the dimension that matters to the task. A shape or position change survives colour blindness and dark mode better than hue alone.
- Pair isolation with a label. If the highlighted plan says why it is highlighted, the emphasis is information rather than decoration.
- Audit emphasis on a quarterly cadence. Highlight creep is what turns a system with one isolate into a system with none.
Where it turns manipulative
- Isolating the option that serves the business while the option serving the user is rendered as low-contrast text. This is the confirmshaming pattern in a different costume.
- Using isolation for urgency that does not exist, such as a permanent red badge on a non-urgent notification, which trains people to ignore red.
- Carrying emphasis by colour alone, which removes the cue for anyone with low vision or colour vision deficiency.
Where you have seen it
Slack
Channels with unread mentions get a bold name and a numeric badge while every other channel stays uniform, so the isolate is scarce by construction.
GitHub
The merge button is the only green element on a pull request page, which makes the terminal action the single isolate.
Notion
Pricing tables mark a single plan as popular with a border and label while the other columns stay identical.
What the research says
- von Restorff, 1933Well evidenced
An isolated item in an otherwise homogeneous list was recalled far more often than items in the homogeneous background.
The original work is in German and its samples were small by modern standards. The core result has been reproduced many times since.
- Hunt, 1995Well evidenced
A review argued the effect comes from distinctiveness processed against a shared context, and that the isolate gains while the surrounding items lose a little.
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.