Chinese dialects and accents: how close are they? We measured

How many Chinese dialects are there?

Linguists count about ten groups, with hundreds of local varieties inside them. Chinese calls them all regional speech, 方言(fāngyán), which English usually translates as dialects.

Whether a group is a dialect or a language is partly a matter of politics. By the usual test, whether speakers understand each other, the southern groups behave like separate languages. Our guide to Mandarin and Cantonese works through that question for Cantonese.

Here are the ten groups, with the speakers each has:

Mandarin has two in three of all speakers. Source: the Language Atlas of China, 2012, via Wikipedia; first-language speakers.

The names you hear in daily life map onto these groups. Cantonese is the best-known Yue variety, Shanghainese is Wu, and Hokkien and Teochew are Min. Taiwanese, widely spoken in Taiwan, is a form of Hokkien too.

We compared how the same words sound

Comparing dialects by ear is hard, so we let the characters do the work. Chinese varieties write a shared word with the same character, even when they say it very differently.

Take person, 人(rén). Beijing starts it with a curled r. Shenyang says something closer to yín, Shanghai starts with ny, Cantonese with a y sound, and Hokkien with an l.

For every pair of 22 cities, we compared the characters both use, a median of 574 per pair. We scored two things for each character:

• The first consonant: the same, or not.
• The rest of the syllable: how many sounds you'd have to change to turn one into the other.

We kept tones out of the score and show them separately. In a listening experiment, matching tone pitch didn't predict how well people understood each other, but consonants and vowels did.

One check keeps us honest. When two surveys recorded the same city, its two versions scored 92% alike on average. So any pair near 90% is about as close as two transcriptions of one city.

Min splits off first in the family tree

When we grouped the cities by how alike they sound, the tree split the southeast coast off first. Here's the family tree our numbers produce:

Read it from the left. The further right two branches join, the more different they sound. Four branches stand out:

• Min forms the first split. Xiamen and Chaozhou, home of Hokkien and Teochew, sound 78% alike, and Min's three cities sit far from all the rest.
• Wu comes next. Shanghai and Suzhou pair up tightly, while Wenzhou stands apart even inside Wu.
• Cantonese, Hakka and Gan cluster on sound. Linguists have long paired Hakka with Gan.
• Mandarin cities huddle together, with Changsha's Xiang and Taiyuan's Jin close by.

Changsha and Taiyuan aren't Mandarin, but both have absorbed a lot of it. Changsha's speech has been significantly influenced by Southwest Mandarin, and linguists filed Jin under Mandarin until the 1980s.

If you speak Mandarin, the branches far from yours are the ones you'll need to learn almost from scratch.

On the map, Mandarin is one tight cluster

A tree forces every city onto one branch. A map shows the same distances in two directions at once. Here, distance means difference, not kilometres:

Two directions stand out. Left to right runs roughly north to south, from the Mandarin cities to Cantonese and Min. Top to bottom separates Wu from everyone else.

That matches the first large-scale computer analysis of the Linguistic Atlas of Chinese Dialects, built from fieldwork in 2000–2008. In it, Huang and colleagues found the same two main patterns: northern features spreading from the North China Plain, then a second wave up the Yangtze.

The Mandarin points sit close together even though they stretch across thousands of kilometres. That's the shape of Chinese: one huge, fairly uniform north, and a south packed with very different speech.

Compare any two varieties

Pick two cities to see how alike they sound, which parts differ and how many basic words they share. Where a listening test covered both, you'll also see how much each side understood:

Start with Beijing and the city your family comes from. Then try two cities in the same group, like Shanghai and Suzhou, to see how much variety hides inside one name.

Mandarin speakers catch little of the south

The numbers match what listeners actually understand. In a 2009 experiment, Tang Chaoju and Vincent van Heuven played sentences in 15 Chinese varieties to rural listeners from each one.

Listeners wrote down the last word of each sentence. Here's how many words listeners from Beijing caught:

Source: Tang & van Heuven (2009), Lingua 119(5), table 3. Listeners aged 40 to 65; each figure is 60 answers.

Understanding runs one way. Listeners from every city understood Beijing speakers well, 87% of words on average, because they hear standard Mandarin at school and on TV.

Our sound score lines up with these results. Across the 91 pairs both studies cover, varieties that sound more alike by our measure were understood better.

So if you learn Mandarin, expect to follow most of the north and almost none of the south without study. Our Mandarin and Cantonese guide has the full picture for Cantonese.

Words drift less than sounds

We found that most basic words share a root across Chinese, even where the sounds have moved far apart. Beijing and Guangzhou share roots for about three in four basic meanings, yet sound only half alike.

The words that do change are some of the most common ones. Here's how three cities said five of them in the 1964 survey:

Guangzhou shares none of these five words with Beijing, and Xiamen only one. Source: Beijing University's 1964 survey, readings in the IPA with tones left out; southern words appear as the survey wrote them.

Cantonese and Hokkien both say eat with the first character of Mandarin food, 食物(shíwù). Hokkien's see is Mandarin's word too, just pronounced very differently.

That's why reading carries further than listening. Formal written Chinese follows Mandarin everywhere, so a southern speaker who reads a newspaper is reading Mandarin's words.

The south kept old sounds the north lost

Southern varieties kept more of Middle Chinese, the Chinese of the Sui and Tang dynasties, and have more tones to show for it. By the traditional count, Beijing has four tones and Guangzhou nine:

Source: our counts from the 1964, 2004 and 2007 surveys, checked against standard descriptions. How we counted.

Two more old features survive in the south:

• Stop endings. Cantonese, Hakka and Hokkien still end syllables in -p, -t or -k, about one syllable in six. Shanghainese and Fuzhou keep only a short catch in the throat, a glottal stop, and Beijing lost them all.
• Voiced b, d and g. Wu kept these old voiced consonants. About one syllable in five in our Wu cities starts with one, against almost none in Beijing.

Mandarin learners notice the tones first. In the listening test, though, it was the consonants and vowels that tracked understanding. Ten, 十(shí), ends in a p in Cantonese and in a glottal stop in Shanghainese.

Beijing's own accent is the exception

Inside Mandarin, Beijing has two habits most other cities don't share. Standard Mandarin took Beijing's sounds as its model, so learners meet both.

The first is erhua, an r sound added to the ends of words. Where, 哪儿(nǎr), is the classic example. Here's how often everyday words carried it in the 1964 survey:

Source: our count over the Beijing University 1964 survey. No city outside Mandarin used it.

The second is the curled-tongue zh, ch and sh. They're in about one Beijing syllable in six. In the same survey, Shenyang in the northeast and Chengdu in the southwest have almost none: both say them as z, c and s.

Other accent markers vary by region. Chengdu merges n and l, so Southwest Mandarin speakers may say south, 南(nán), much like blue, 蓝(lán). The four tones also take different pitches from city to city.

So when a Mandarin speaker from Chengdu or Taiwan drops the r and flattens zh to z, that's a regional accent shared by millions, not a mistake.

Taiwan and mainland Mandarin: one language, two standards

Mainland China and Taiwan both teach a standard built on Beijing's pronunciation. Since 1949 they've drifted apart in accent, some readings, new words and script, but speakers still understand each other without much trouble.

The biggest difference is how people actually speak. Here's how Taiwan's everyday Mandarin compares with the mainland standard:

Taiwan's everyday Mandarin softens the curled-tongue sounds and drops almost all erhua. Sources: Chen (1999); Kubler (1985); Fon, Chiang and Cheung (2004), via Wikipedia. Not every speaker has every feature.

The official dictionaries also disagree on some readings, usually just the tone. We checked the 3,000 most common characters in CC-CEDICT, the open Chinese–English dictionary. It notes a different Taiwan reading for 85 of them, about one in 35.

Most of them are rare enough that you'd meet one in about every 200 characters of speech. Week, 星期(xīngqī), is xīngqí in Taiwan, and trash, 垃圾(lājī), is lèsè.

Everyday words that differ

Vocabulary differs most for things invented or named after 1949. Here are words you'd use on a first day out:

Most of the differences are words for things that arrived in the 20th century. Source: CC-CEDICT, release of 6 October 2026, which marks the Taiwan usage of each.

Watch the false friends, words written the same with different meanings:

• Potato or peanut: 土豆(tǔdòu) is a potato on the mainland and a peanut in Taiwan.
• Computer or calculator: 计算机(jìsuànjī) is a computer on the mainland and a calculator in Taiwan.

Script is the other visible difference. Taiwan writes Traditional characters, while mainland China and Singapore use Simplified. Our guide to Simplified and Traditional Chinese shows which characters change and how fast you can learn the other set.

Taiwan also speaks more than Mandarin. In its 2020 census, about two in three people named Mandarin as their main language, and about three in ten named Taiwanese Hokkien.

Singapore and Malaysia: Mandarin with a southern accent

Singapore's Mandarin is written like the mainland's, in Simplified characters. In speech it carries the sound of Hokkien, Teochew and Cantonese, the dialects many Chinese Singaporeans grew up hearing.

A study of Singaporean speech found that adults front their curled-tongue sounds toward z, c and s, and that the neutral tone is largely missing. Tone 2 has a low, level start that Beijing and Taiwan don't share.

Local words fill the gaps. Taxi is 德士(déshì), from English, and percent can be 巴仙(bāxiān). Malaysia's Mandarin shares many of these words.

The dialects themselves are fading at home. In 2000, three in ten Chinese Singaporeans mostly spoke one at home. By 2025 it was about one in fifteen, Singapore's statistics office reports.

If you learn Mandarin from mainland materials, you'll follow Singaporean speakers. Just expect every syllable to keep its full tone, and words you won't find in your textbook.

Which variety should you learn?

Learn the variety of the people you want to talk to. For most learners that's standard Mandarin, which every Chinese-speaking place teaches and understands.

• No particular people in mind: learn standard Mandarin. It's the one every school, exam and app teaches.
• Family from Guangdong or Hong Kong: learn Cantonese, or add it after Mandarin. Our Mandarin and Cantonese guide helps you choose.
• Family that speaks Shanghainese, Hokkien or Hakka: start with Mandarin for reading and wider use, then learn the family variety from your family.
• Living in Taiwan: learn standard Mandarin with Traditional characters. Your accent will adjust once you're there.

One caveat: none of these varieties is "broken Mandarin". Each has its own sounds, words and grammar, with a long history of its own.

What to do next

• Compare your family's city. Use the tool above to see how far it is from standard Mandarin.
• Learn standard Mandarin first if you're unsure. It's understood in every Chinese-speaking place.
• Don't chase a Beijing accent. Clear tones matter more than erhua.
• Learn both character sets if you'll read in Taiwan. Traditional is easier once you know Simplified.
• Expect local words. Taxi changes between Beijing, Taipei and Singapore, but people will understand you.