On Use of the Lang Attribute
Way back in October I noticed this WHATWG HTML bug (26942) where someone asked
why do these examples of I thought the answer from Hixie was a bit dismissive and not based on any data or real-world benefits of use, particularly in the context of screen readers:
<html> lack the
Why not? Realistically, few people include it. It just means the language is unknown.
At the time, I could not get the latest archive to download from WebDevData.org (though that has changed, see below), so I fell back to asking for help on why the
lang attribute is valuable.
lang Attribute on
<html> Is Used
I got lots of good bits of feedback, which I collected into a Storify (crappy PDF version since the new Storify domain owners may block WayBack). I’ve distilled all that great information to these key points:
- VoiceOver on iOS uses the attribute to auto-switche voices.
- VoiceOver can speak a particular language using a different accent when specified.
- Leaving out the
langattribute may require the user to manually switch to the correct language for proper pronunciation.
- JAWS uses it to load the correct phonetic engine / phonologic dictionary — Handy for sites with multiple languages.
- NVDA (Windows) uses it in the same way as VoiceOver and JAWS.
- When used in HTML that is used to form an ePub or Apple iBooks document, it affects how VoiceOver will read the book.
- Firefox, IE10, and Safari (as of a year ago) only support CSS
hyphens: autowhen the
langattribute is set (not from Twitter; source).
In the absence of setting a
lang attribute on the
<html> element, screen readers will fall back to the user’s default system setting (barring any custom overrides) when speaking content.
How Many Pages Use
On January 8, WebDevData.org (from a W3C Community Group) posted its latest archive (which did not error on download, woo!). It consists of the HTML from 87,000 web pages.
I pulled down the 780MB file and re-taught myself the skills necessary to parse the files. For those who are regular expression geniuses, you are welcome to suggest an alternate approach, but I used the following pattern to return all the
<html([^>]+)>. It fails for any
<html> with no attributes at all, but for what I am doing that’s ok.
Of the 84,054 pages I parsed (I excluded XML, ISO files, and so on), I found that 39,433 use the
lang attribute on the
<html> element. That’s just about 47% (46.914% if I understand significant digits correctly).
What that tells me is that instead of the case being that
few people include it, nearly half the web includes it.
There are 12,672 instances of
xml:lang, though at a quick scan they appear alongside
lang. If anyone with better regex skills would like to help me further parse, please let me know.
Why You Should Use the
lang Attribute on the
lang, you get the benefits of hyphen support in your (modern) browser that you otherwise would not get (assuming you use
hyphens: auto in your CSS).
At the very least,
lang is a benefit for screen reader users, particularly when your users don’t have the same primary language as your site. It allows proper pronunciation and inflection when the page is spoken.
The W3C Internationalization (I18n) Activity has a great Q&A on why you should use
lang, which was updated less than two months ago. I’ll reprint the start of the answer, but there is far more detail and I strongly recommend you go read it.
Identifying the language of your content allows you to automatically do a number of things, from changing the look and behavior of a page, to extracting information, to changing the way that an application works. Some of language applications work at the level of the document as a whole, some work on appropriately labeled document fragments.
We list here a few of the ways that language information is useful at the moment, however, as specifications and browsers evolve in the future there could be numerous additional applications for language information.
If you go to the WHATWG HTML5 specification today and view the page source, you’ll see the following language declaration in the code:
<html class=split data-revision="$Revision: 8877 $" lang=en-GB-x-hixie>
Not to be outdone, the W3C HTML5 spec has the same language declaration.
If anybody has the
en-GB-x-hixie phonologic dictionary in his or her screen reader, I’d love to hear it.
While technically allowed (the
-x puts it in the private use sub-tag category), it’s bad form:
Private-use subtags do not appear in the subtag registry, and are chosen and maintained by private agreement amongst parties.
Because these subtags are only meaningful within private agreements and cannot be used interoperably across the Web, they should be used with great care, and avoided whenever possible.
Update: January 1, 2015
Update: February 25, 2015
Another case where a
lang attribute is important, though in this case on a specific element, is outlined in the piece HTML5 number inputs – Comma and period as decimal marks:
<input type=”number”> will open a numeric software keyboard on modern mobile operating systems. Not every user can input decimal numbers into this convenient field without proper localization.[…]
Half the world uses a comma and the other half uses a period as their decimal mark. (In Latin scripts.) Does your web application take that into consideration? Do the browsers?
Update: April 18, 2016
The WHATWG spec documents have finally been updated (hopefully all of them) to include the
If you look at the changes on GitHub, you’ll see that the
lang attribute has been added to code examples as well.
Update: April 19, 2016
The W3C spec is also being updated with notes on how and why to use
If you have more you’d like to add to that note, you can do so over on GitHub.
Update: May 20, 2016
Steve Faulkner made a related demo.
The same English language text “What will you do when the label comes off, And the plastic’s all melted, And the chrome is too soft?” is marked up in paragraphs with es/fr/de lang attribute values. These each effect the way the text in pronounced.
test file used: http://s.codepen.io/stevef/debug/MyapoQ
Update: July 14, 2016
The W3C HTML5 validator will now try to detect the language of the page you are validating and compare it to the
lang attribute as well as the
dir attribute. If you these do not appear to match, then you will get a warning.
Update: April 16, 2018
Much of this post and a lot of new material made it into my London Web Standards talk Mind Your
Update: December 5, 2021
I should have been more explicit in here that you should avoid region sub-tags (
en-us) unless users have a genuine need for a specific dialect.
From a best practice perspective (emphasis theirs):
You always start by choosing a primary language subtag, and often this is all you’ll need for your language tag.
Always bear in mind that the golden rule is to keep your language tag as short as possible. Only add further subtags to your language tag if they are needed to distinguish the language from something else in the context where your content is used.
As for screen reader support, region sub-tags are generally ignored (emphasis theirs):
If a screen reader supports the regional differences—such as having language voices installed for both Peninsular and Mexican Spanish—then the screen reader may switch to the appropriate dialect.
However, region subtags are typically ignored, especially if the screen reader’s default language matches the primary language specified. This is because it is presumed the user will prefer and better understand their default dialect over a different dialect of that same language. The numerous screen reader users in Great Britain, for example, will typically hear Great Britain English on U.S. web sites, even if the page has
lang="en-US"specified and the user has the US English language voice installed.
Only use region subtags when it is necessary to differentiate content in different dialects that may not be mutually intelligible. A web site that provides content in both Mandarin and Cantonese (one of which may not be understood by speakers of the other) would typically differentiate them using
Script and and extended language sub-tags have poor support in screen readers.
Better regex to select HTML tags
Jake, was there a regex example in your comment? I am not seeing one.
I had a screenreader user on complaining that my website made his screenreader read out oddly: turned out the en-gb tag was switching his TTS. Confused him utterly! Most users won't have multiple TTS variants, I suspect, but might be best to stick with "en" rather than "en-gb" and "en-us" so as not to surprise users?
Was the content in British English? For example, did the content include spellings like "colour" instead of "color?" Punctuation inside or outside of quotes?
I have no idea if those would cause a difference in experience, but that's certainly where I'd start looking. I'll also ping some folks and see if others are familiar with this.
Better <html> regex: <htmlb.*?>
b is a break
.* is anything. ? afterwards makes it "not greedy"
Thanks, I will give that a shot and see what I get.
Document language is typically easy to set and maintain, but doing so simply by HTTP header can be more practical. Changes in language are more costly to set and maintain, and detection of such changes may be best delegated to software (I don’t think this is covered here, however). Our whole way of working with language may not be ideal .
I was able to scare up some additional reasons to use
- Default font selection for CJK languages (very politically important, there’s still hurt feelings over how Unicode handled them)
- The quotation marks around
<q>change with the language
::first-letterpseudo-element can behave badly in non-English languages without
lang. Other upcoming CSS typographical niceties like hanging punctuation are also likely to require it.
spellcheckattribute needs it to function properly, especially for multilingual users, as browser heuristics aren’t perfect
- Input types for dates and times are also need it, because around the world there is absolutely no de facto standard for writing them up. This huge variance is partly why browsers have been so slow to implement these types.
Taylor, your comment came to mind when I read this post and I figured it might be useful for those who don’t know the CJK font issue: Localization Gotchas for Asian Languages (CJK) (I did localization in that part of the world over a decade ago, and I wasn’t clear on it).