Why most Chinese topolects have no written standard

Everyone in China reads the same characters, which is why people assume everyone writes the same language. They do not. Written Chinese is Mandarin, and a Cantonese or Shanghainese speaker writing it is translating.

Updated 22 September 2026 · 5 min read

Tap 🔊 to hear an example. Shanghainese examples play the recordings from Verbavia’s course, made with a Shanghainese voice; Mandarin examples play a Mandarin voice.

The thing nobody mentions

Modern written Chinese — 白话文, the written vernacular that replaced classical Chinese in the early twentieth century — is built on Mandarin. Wikipedia states it plainly: written Chinese “most nearly represents the vocabulary and syntax of Mandarin”. The reformers who created it “used Lower Yangtze and Beijing grammar and vocabulary in order to make their writing understandable to the majority of readers”, and the effect was to standardise written Chinese across the country.

That was a real achievement. It also means that when a Cantonese speaker writes standard Chinese, they are writing a language they do not speak, and then reading it back aloud in their own pronunciation. The characters are shared. The grammar and the everyday vocabulary underneath them are Mandarin’s.

Native speakers of Wu and Yue do this from childhood and mostly stop noticing. It is the single biggest reason outsiders think these varieties are closer to Mandarin than they are — the page always agrees, whatever the mouth is doing.

Cantonese is the exception, and it is a real one

Written Cantonese exists, it is developed, and it is used in public. It has its own characters — 佢 for he or she, 唔 for not, 係 for to be, 嘅 for the possessive, 冇 for to not have, 哋 for the plural, 睇 for to look, 咗 for a completed action — and none of those is standard Mandarin.

Where it appears, according to Wikipedia’s account: newspapers run their news in standard Chinese but may carry editorials and columns in Cantonese; it is “quite popular in certain tabloids, online chat rooms, instant messaging, and even social networking websites”; it turns up in advertising and on billboards; and cinema audiences “regularly encountered written Cantonese at the cinema” in subtitles. Hong Kong even maintains an official character set for it, the Hong Kong Supplementary Character Set.

Why Cantonese and not the others? Wikipedia gives the short answer: “Cantonese-speaking Hong Kong used to be a British colony isolated from mainland China before 1997, so most HK citizens do not speak Mandarin. Written Cantonese has developed as a means of informal communication.” A separate jurisdiction, a mass-market press, a film industry and a population that could not default to Mandarin — that is the recipe, and no mainland variety had it.

How you invent a character

This is the genuinely interesting part, because writing a language that has never been written means solving a problem thousands of times. Three strategies, all attested:

1. Dig up a forgotten one. Writers hunted for Old Chinese words “when their original forms have been forgotten”, using ancient rhyme dictionaries such as the Guangyun to find a character that once had the right sound and sense.
2. Build a new one, usually with a mouth. The commonest method adds the mouth radical 口 to a character used for its sound, marking the result as a colloquial word: 咗, 唔, 咁. If you see a lot of 口 on the left-hand side of characters in a text, you are reading something written to be spoken.
3. Give up and write letters. Latin characters get used inline — D standing in for 啲, or a syllable written in romanisation because no character exists or the one that does cannot be typed.

Shanghainese uses the same repertoire on a much smaller scale, and the choices it makes are on the page about writing Shanghainese.

Why Shanghainese did not

It is not for want of trying, and the history is older than most people expect. Protestant missionaries romanised Shanghainese in the nineteenth century — Joseph Edkins among them — producing Bibles and dictionaries, and the American missionary Tarleton Perry Crawford went further, devising a set of phonetic symbols shaped like Chinese characters for the purpose.

There is even a literary monument. The Sing-song Girls of Shanghai (海上花列传), published by Han Bangqing in 1892, narrates in literary Chinese and puts all of its dialogue in Wu. Hu Shih later argued that the heavy Wu was precisely why readers struggled with it, though the scholar David Der-wei Wang has disputed that reading.

What never arrived was a standard. Wikipedia’s summary of the position today is that Chinese characters are often used to write Shanghainese and “there is no formal standardisations, there are characters recommended for use”. Recommended, not fixed. Two writers will spell the same word differently and neither is wrong.

The places that did get a standard

Two counter-examples, both outside the mainland’s policy environment, and both instructive about what it takes:

• Taiwanese Hokkien. Pe̍h-ōe-jī, a Latin alphabet devised by missionaries in the nineteenth century, still holds the richest body of written work. Then in 2006 Taiwan’s Ministry of Education officially promoted the Tâi-lô romanisation, and in 2007 it published a standard character set, the Taiwanese Southern Min Recommended Characters — now taught in schools. That is a genuine official orthography, which almost no topolect has.
• Hakka. Written with characters and no single standard, but Guangdong’s education department did create an official romanisation of the Meixian variety in 1960.

The pattern is hard to miss: a written standard follows a government that wants one. Hong Kong’s separation, Taiwan’s post-1987 language policy. Where the institutional default is Mandarin, the writing stays improvised.

The Unicode tax

Inventing characters has a modern cost. Many topolect characters ended up in CJK Unified Ideographs Extension B, which spans more than forty thousand code points on Unicode’s supplementary plane, and font coverage there is patchy — the block was so large that the original charts used a single glyph for all regions, and support has been uneven ever since.

The practical result is a blank box. Shanghainese has a purpose-made character for its yes-or-no question particle, 𠲎 at U+20C8E, which sits in that block — so writers substitute the ordinary 伐 va ‘(question particle)’ instead, as this site does. Details on the typing page.

The legal background, stated plainly

China’s constitution provides that the state promotes the nationwide use of Putonghua, and teaching non-Mandarin varieties to non-native speakers is discouraged by law. In 2012 Guangdong’s language regulations moved to promote standard Chinese in broadcast and print media in place of Cantonese and other local varieties.

These are promotion and discouragement rather than blanket prohibition, and the distinction is worth keeping. But for a writing system the effect is much the same: an orthography needs schools, publishers and input methods, and those follow official status. The wider picture is on the page about whether the topolects are dying.

Why this matters if you are learning one

Because it explains the thing that will frustrate you most. There is no spelling to look up, so dictionaries disagree; search engines index the Mandarin form; input methods assume Mandarin readings; and a text you find online may use three conventions in one paragraph.

The workaround is to lean on romanisation for learning and characters for reading, rather than expecting either to be authoritative. Which romanisation, and why this course picked the one it did, is on the romanisation page.

Sources

1. Wikipedia: Written Chinese — modern written Chinese “most nearly represents the vocabulary and syntax of Mandarin”
2. Wikipedia: Written vernacular Chinese — Lower Yangtze and Beijing grammar and vocabulary as the basis of the standard
3. Wikipedia: Written Cantonese — where it is used, the Hong Kong Supplementary Character Set, the character-creation strategies, and the colonial explanation
4. Wikipedia: Written Shanghainese — Edkins’s romanisation, Crawford’s New Phonetic Character, and the absence of a formal standard
5. Wikipedia: The Sing-song Girls of Shanghai — Han Bangqing, 1892, with all dialogue in Wu; Hu Shih and David Der-wei Wang
6. Wikipedia: Written Hokkien — Pe̍h-ōe-jī, Tâi-lô officially promoted from 2006, and the 2007 Taiwanese Southern Min Recommended Characters
7. Wikipedia: Hakka Chinese — the 1960 Guangdong romanisation of the Meixian variety
8. Wikipedia: CJK Unified Ideographs Extension B — the block, its size and its font-coverage problems
9. Wikipedia: Protection of the Varieties of Chinese — the constitutional promotion of Putonghua and the 2012 Guangdong regulations