Translations of the same content vary substantially in length across languages

What it is

Cross-language text expansion is the change in characters, glyph width, line count, and screen area when equivalent interface content is expressed in another language. It is not simply evidence of a verbose translator. Morphology, grammatical marking, writing system, and conventional phrasing can make one language encode in several words what another expresses as a compound or leaves implicit.

Why it happens

An interface lays out shaped glyphs, not abstract character counts. Word length, spaces, capitalization, ligatures, average glyph width, and line-breaking rules all affect the rendered measure. Grammar may also require explicit number, case, subject, or politeness information absent from the source. Similar character counts therefore need not occupy similar widths. In a narrow component, one additional word can trigger a whole new line, producing a discontinuous jump in height.

Studying it

Use the same message key, equivalent meaning, and translation version as the unit of analysis, then render production translations under controlled font, size, viewport, and wrapping conditions. Outcomes include width, line count, component height, overflow, occlusion, malformed breaks, and action visibility. Means hide the failing tail, so report percentiles and worst cases by locale and content type. Parallel corpora without UI context describe linguistic variation but cannot establish component failure rates.

Where it stops holding

There is no universal expansion percentage for a language. Source length, terminology, translation strategy, font metrics, and component width all alter the result, and translation direction is not symmetric. Measurements from one font or desktop viewport do not transfer directly to another font, a narrow screen, or enlarged text. Flowing prose and single-line controls also have different failure modes.

Applying it

• Render real translations in production fonts instead of estimating from character counts.
• Treat copy as variable data: let containers grow, wrap, or reflow, and document the few places that genuinely require one line.
• Combine long translations, narrow viewports, and enlarged text in tests covering actions, navigation, validation messages, and notifications.
• Use automated screenshots and overflow checks, then manually inspect semantic line breaks and whether critical actions remain fully visible.

Related

• Same group: S1.01.2 Short labels expand proportionally more than long passages · S1.01.3 Containers fixed to source-language length inevitably overflow · S1.01.4 Layouts must be tested against the longest language, not the average · S1.01.5 Truncation hides overflow; it does not solve it
• Adjacent: S1.03 Character sets, font coverage, and missing glyphs · S3.04 High- and low-context cultures and information-density preferences
• Search terms: text expansion · localization testing · glyph width