Danger, Warning, Caution are not interchangeable — they are a defined hierarchy tied to standards like ANSI Z535 and ISO 3864. Translating warnings as ordinary text is where product-liability exposure hides.
- Signal words are a system, not vocabulary
- Where localized warnings go wrong
- How to handle it in the workflow
Signal words are a system, not vocabulary
On safety labels and in manuals, "Danger", "Warning" and "Caution" are not three ways of saying "be careful". They are a defined hierarchy: Danger for a hazard that will cause death or serious injury, Warning for one that could, Caution for lesser injury. Standards such as ANSI Z535 and ISO 3864 fix the tiers, their colours, and often their symbols. The words are the visible tip of a classification system.
That is why warnings cannot be translated like ordinary text. Each signal word must map to the target-language term that carries the same tier under the applicable standard — not to whichever synonym reads most naturally. Collapse Danger and Warning into one comfortable target word and you have removed a distinction the standard requires and a court may later examine.
| Signal word | Severity tier it denotes | What flattening it removes |
|---|---|---|
| DANGER | A hazard that will cause death or serious injury. | The top tier stops being distinguishable from the one below it. |
| WARNING | A hazard that could cause death or serious injury. | Readers lose the cue that the outcome is conditional, not certain. |
| CAUTION | A hazard that could cause lesser injury. | Minor hazards read as severe ones, and severe ones stop standing out. |
Each signal word must map to the target-language term carrying the same tier under the applicable standard — not to whichever synonym reads most naturally. Fix the mapping once, per language, in the termbase.
Where localized warnings go wrong
Tier flattening. The target language’s everyday vocabulary may not have three neat equivalents, so an unwary translator uses one word for two tiers. The fix is a fixed, approved mapping of signal words per language, locked in the termbase — decided once, by someone who knows the target-market standard, not improvised per document.
Losing the structure. Warning format is regulated too: signal word, then hazard, then consequence, then avoidance. Rewriting that into a flowing sentence can strip the required structure even when the words are right.
Symbol and text drift. The pictogram and the text must agree. A translated caption that no longer matches its hazard symbol is a defect, and it is easy to introduce when text and artwork are handled by different people at different times.
| Failure | What it looks like | The fix |
|---|---|---|
| Tier flattening | The target language has no three neat equivalents, so one word covers two tiers. | A fixed, approved signal-word mapping per language, locked in the termbase — decided once by someone who knows the destination-market standard. |
| Losing the structure | Signal word / hazard / consequence / avoidance is rewritten into a flowing sentence. | Review against the source layout and the governing standard, not only for linguistic quality. |
| Symbol and text drift | A translated caption no longer matches the hazard pictogram beside it. | Keep text and artwork in one review pass; the pictogram and the text must agree. |
How to handle it in the workflow
Treat signal words and standard warning phrases as protected, pre-approved terminology — supplied to the translator locked, with the target-market standard identified. This is the one place where "translate it naturally" is the wrong instruction; the right instruction is "use the approved term for this tier, exactly".
Review localized warnings against the source layout and the governing standard, not just for linguistic quality. And when the destination market or its standard is unclear, resolve that before translation — because the correct target term for "Warning" can differ depending on which standard applies, and you cannot pick the word until you know the rulebook.