Icons shown without context are rarely guessed correctly
Aliases: guessability · isolated identification · icon comprehension · ISO 9186
What it is
Lift a glyph out of a toolbar, a tab, or a button, put it on a blank field, and ask what it does: most people get it wrong. The measure is guessability — can the picture be read on first sight, with no surrounding controls and no task to narrow the candidates. That is not learnability. A poor guess can still be remembered after a few uses. Guessability is the untaught encounter.
Why it happens
An interface icon almost never carries its own definition. It is a compressed pointer that works only when three things are present at once: the viewer has seen a similar drawing, neighbouring entries supply contrast, and the current task collapses the meaning space to one or two options. Strip those away and a silhouette remains. People map the silhouette onto household objects or interface folklore; when the mapping misses, they invent a plausible story, and the story is often wrong. The failure is not that the drawing is insufficiently realistic. It is that an isolated presentation does not supply enough constraint to disambiguate.
Studying it
The industrial skeleton is the comprehension procedure in ISO 9186: show the symbol, ask for a free-text meaning, score correct / partially correct / wrong. A matching test (pick from options) inflates the score by turning naming into recognition, and is not evidence of guessability.
Typical independent variables: product context present or absent, whether the whole set is shown together, whether participants already use the product. Typical dependent variables: free-naming accuracy, error type (near-miss versus unrelated), confidence.
In interface work the method is used to decide whether a glyph can travel without a label. One caution: isolated lab tests under-estimate performance in a real toolbar, where a task constrains meaning. If the product will actually ship the icon unlabelled at a new entry, the isolated score is an upper bound — the field will be worse.
Where it stops holding
- High-frequency entries that already have muscle memory are not blocked by poor guessability; the people who suffer are those meeting the control for the first time.
- Showing the whole set together raises accuracy through mutual contrast; a set score must not stand in for a single-glyph score.
- Comprehension scoring depends on the coding scheme. Lenient and strict rules can split the same transcripts by a wide margin, so cross-study percentages are not comparable.
- This is "the picture cannot be read", not "the same picture is assigned conflicting meanings across products". That is a different failure.
Applying it
- New, rare, or onboarding-facing actions do not pass as unlabelled icons.
- In review, flash the glyph alone, with no page, to people who did not draw it, and ask what it does. If they cannot say, add words — do not redraw the picture more carefully.
- How to check: run open naming on the candidates and record first-try accuracy. Exclude design colleagues from the sample.