Standards · WCAG 2.2

SC 3.1.2 Level AA WCAG 2.0

Language of Parts

When a passage or phrase on the page is in a different language from the page default, mark it with a lang attribute on its container — so screen readers switch voice and pronunciation for that fragment.

What it asks

If a sentence, paragraph, blockquote, or inline span on a page is in a language other than the page’s declared default, that fragment must carry its own lang attribute. Exceptions: proper names, technical terms, undetermined-language words, and words that have become part of the surrounding language (“rendezvous” in English).

The intent is that screen readers switch pronunciation engines on the fly — reading the German blockquote with a German voice, returning to English for the surrounding commentary.

How to meet it

  • Wrap the foreign-language passage in any element (often <span>, <blockquote>, <q>, <p>) and add lang.
  • Example: <p>The motto is <span lang="la">carpe diem</span>.</p> — except “carpe diem” is borrowed and exempt; the rule applies to genuinely foreign passages.
  • For long quotations, use <blockquote lang="fr"> around the whole quote.
  • For mixed-language headings, mark each segment.
  • For right-to-left fragments inside LTR pages, pair with dir="rtl".
  • When pulling localized strings from a CMS, ensure the locale field flows through to a lang attribute on the rendered fragment.

Common failures

  • A multilingual page with the correct root lang but no per-fragment annotations — every foreign quote, song title, or testimonial read with the wrong voice.
  • Marking proper names with lang (over-application) — the SC explicitly exempts these.
  • Using lang="english" or other non-BCP-47 values inside the fragment.
  • Bilingual product pages with side-by-side translations where only one column has a lang attribute.
  • News sites that quote foreign politicians without wrapping the quote.

Why it matters

Without per-part language switching, screen readers pronounce a French phrase like “merci beaucoup” using English phoneme rules, which is either funny or unintelligible. For users who rely on synthesized speech as their primary text channel, this matters more than it looks — long news articles, academic content, and multilingual marketing copy become exhausting to parse.

This SC is widely under-implemented even on sites that pass 3.1.1, because it requires editorial discipline at content-creation time, not just a template change.