Mandarin vs Cantonese: how much do they really overlap?

About half the basic words match on paper

How much the two overlap depends on which layer you compare. The formal writing standard is shared, and about half the most basic words are identical on paper. Everyday speech drifts much further apart.

To measure the words, we used the Leipzig–Jakarta list: 100 basic meanings that languages rarely borrow words for, like "fire", "eat", "he" and "not". For each meaning, we took the everyday Mandarin word, the most frequent candidate in a corpus of Mandarin film subtitles.

We paired it with the everyday Hong Kong Cantonese word, checked against words.hk, CC-Canto and Beijing University's 1964 survey of dialect vocabulary. Then we sorted the pairs.

Our verdict: 50 use the same written word, 21 a related one and 29 a different one. Here they are, side by side:

Same on paper doesn't mean same in sound. Even the identical words are said differently, and the related ones usually just drop a Mandarin suffix:

Source: our core-100 comparison, with readings checked against CC-CEDICT, words.hk and CC-Canto.

A Mandarin learner can usually guess the related words. Further down, you'll see why the sounds differ, and why they differ by rule.

The everyday words are the ones that change

The 29 different words include some of the most frequent in any conversation:

They split three ways:

• Ten are words a Mandarin reader knows from formal writing. Cantonese often keeps the older word that Mandarin now uses only in writing. Cantonese "eat" is the first character of Mandarin "food", 食物(shíwù), and Cantonese "drink" is the first character of Mandarin "drinks", 饮料(yǐnliào).
• Four are false friends: the same character with a different everyday meaning.
• Fifteen use characters or words standard Mandarin doesn't have, like the Cantonese words for "he or she", "not", "at" and "give" above.

Here are the four false friends:

Fourteen of our calls are borderline, meaning a native speaker might reasonably pick the other word. Mouth is one: Cantonese uses both 嘴(zeoi2), the same word as Mandarin, and 口(hau2).

If every borderline call went the other way, the "same word" count would land between 47 and 52. So "about half" is the honest summary.

Why other counts say 37% or 85%

Published word lists give very different answers, and the reason is useful. Here's how two expert surveys compare with our list:

Sources: our counts on the Lexibank editions of Beijing University's 1964 survey, 903 concepts, and Liu et al.'s 2007 survey, 203 basic words, and our core-100 list.

The low figure records speech. Beijing University's fieldworkers wrote down colloquial Guangzhou speech in the 1950s.

The high figure mixes in writing. Several of the 2007 survey's Guangzhou entries are formal written Chinese rather than speech. It gives the written words for "who" and "not", 誰(shéi) and 不(bù), where people actually say the Cantonese words in the table above.

Count what people write, and Cantonese looks like Mandarin with a Southern accent. Count what they say, and it doesn't.

The grammar shifts too

Word order changes in some common patterns:

The explorer above sets nine everyday phrases side by side, including "Have you eaten?" and "Grandma, I miss you."

Cantonese speakers understand Mandarin, not the reverse

Without study, a Mandarin speaker won't follow Cantonese. In a controlled test, Mandarin listeners caught 3% of the target words in Cantonese sentences, while Cantonese listeners caught 90% of Mandarin ones.

The test comes from a 2009 study in Lingua by Tang Chaoju and Vincent van Heuven. They recorded words and sentences in 15 Chinese varieties, played them to listeners from each one and collected 47,250 answers.

Listeners did two tasks:

• Words: sort a single spoken word into one of ten meaning categories. Guessing would score 10%.
• Sentences: write down the last word of a spoken sentence.

Here's how the two groups did:

On single words, Cantonese listeners got 63% and Mandarin listeners 34%. In whole sentences, the gap became a cliff.

For comparison, each group scored 92–98% on sentences in its own variety.

Why: Cantonese speakers grow up hearing Mandarin

Exposure explains the gap, not similarity. The listeners were rural speakers, chosen to be as monolingual as possible. Even so, 63% of them said they could speak Standard Mandarin.

The authors put Mandarin's head start down to the standard language reaching every part of China through school and the media. That exposure keeps growing:

Sources: China's Ministry of Education for 2025 and 2020; Hong Kong's 2021 census, residents aged 5 and over.

Nothing comparable pushes Cantonese north. If you learn Mandarin, expect Cantonese to sound like a foreign language.

Hong Kong speaks Cantonese but writes standard Chinese

You can read Hong Kong's formal writing, because it follows Mandarin. Written colloquial Cantonese is another matter.

Hong Kong people mostly write Standard Written Chinese, the Mandarin-based standard, in Traditional characters. They speak Cantonese. Written Cantonese turns up in informal places: chat, some columns, ads and some subtitles.

To measure the gap, we used the Universal Dependencies Hong Kong treebanks. They pair 1,004 transcribed Cantonese sentences with their written versions, from three student films and a 2016 Legislative Council meeting.

Here's a typical pair. Both lines mean "What are you looking for?", and only the word for "you" survives:

Only 34.0% of the 11,189 spoken words reappear in the written version of the same sentence. Explore eight real pairs, and the words that change most:

One caveat: subtitles and the official record also trim and reword speech. The written side has 8,135 words to the spoken side's 11,189, so these numbers measure the distance from speech to its written form, not vocabulary alone.

The words that vanish most often are grammar words: "is", the possessive marker, "not", the completed-action marker, and "he or she".

Formal writing is mostly readable

A Mandarin reader can follow the news pages of a Hong Kong newspaper. Expect two hurdles:

• Traditional characters. They're easy to learn with Simplified as a base, and our guide to Simplified vs Traditional shows which ones change.
• Hong Kong vocabulary. Some words are local to Hong Kong.

SayMei can show its Mandarin lessons and stories in Traditional characters. It uses Taiwan's character forms, not Hong Kong's, which differ in a few characters.

Written Cantonese: 20 characters cover most of the gap

We counted how often written Cantonese uses characters outside Mandarin's 3,000 most common. Those cover 99.8% of Mandarin film subtitles, so anything outside them is rare for a Mandarin reader.

Here's how often each kind of text leaves that list:

Source: our counts over 15,428 characters of transcribed speech, 2.47 million characters of written Cantonese in the words.hk corpus, and 12,080 characters in the Standard Written Chinese version of the same sentences. Mandarin's top 3,000 come from the SUBTLEX-CH subtitle corpus.

The gap is concentrated. Twenty characters account for 81.2% of these Cantonese-only characters, so learning them buys a lot. Here they are, most common first:

The same character gets a predictable new sound

Both languages descend from Middle Chinese, and each changed in its own regular way. That means a Mandarin learner can predict many Cantonese sounds.

We lined up 29,969 words that appear in both an open Mandarin dictionary, CC-CEDICT, and a Hong Kong one, words.hk. Then we compared them syllable by syllable and counted the patterns:

Cantonese kept the endings Mandarin lost

About one Cantonese syllable in six ends in -p, -t or -k. Across the words, that's 16.4% of syllables; weighted by how often you hear each word, it's 12.1%.

Cantonese also keeps an -m ending: 99.2% of its syllables that end in -m end in -n in Mandarin. Here are five common words that show both:

Mandarin lost the stop endings centuries ago and moved those syllables into its four tones. They landed everywhere:

Source: our alignment of the 29,969 words in CC-CEDICT and words.hk.

So Mandarin gives you no clue which words have them. You learn them word by word.

Consonants shift by rule

Cantonese has no curled-tongue sounds, so Mandarin zh, ch and sh become z, c and s almost every time. Mandarin r becomes j, said like English y, in 98.2% of syllables: "person", 人(rén), is 人(jan4).

Some shifts even keep a distinction Mandarin merged. "Capital", 京(jīng), and "essence", 精(jīng), sound the same in Mandarin. In Cantonese they're 京(ging1) and 精(zing1).

The sound map above lists every rule, with its share and an example.

Tones line up, mostly

Each Mandarin tone has a usual Cantonese partner, as the tone grid above shows. Tone 1 stays tone 1 in 82.5% of syllables, and tone 2 becomes tone 4 in 75.0%.

Tones 3 and 4 split two ways, but the first sound of the syllable sorts most of them out:

Source: our alignment of CC-CEDICT and words.hk.

So "I", 我(wǒ), is 我(ngo5), and "bitter", 苦(kǔ), is 苦(fu2). The tone 4 split follows a distinction Mandarin lost.

These are tendencies, not rules to apply blind. But they make Cantonese pronunciation far less random than it first sounds.

How many tones does Cantonese have: 6 or 9?

Six that change meaning, or nine by the traditional count. The traditional count treats three tones on -p, -t and -k syllables as separate. The Linguistic Society of Hong Kong's Jyutping table shows both views.

The six tones are clearest on one syllable, si:

The "entering" tones are the three on syllables ending in -p, -t or -k:

They use the same pitches as tones 1, 3 and 6 but are cut short, which is why Jyutping gives them the same numbers. Count them separately, and you get nine.

Mandarin has four tones plus a light neutral tone: "mother", 妈(mā); "hemp", 麻(má); "horse", 马(mǎ); "scold", 骂(mà); and the question particle 吗(ma).

For Mandarin practice, our tone-pair trainer drills the pairs that trip learners up. Our guide to learning Mandarin tones has a tone lab.

Jyutping, the Cantonese pinyin

Jyutping does for Cantonese what pinyin does for Mandarin. The Linguistic Society of Hong Kong published it in 1993, and it writes tones as numbers after each syllable.

So "eat" is 食(sik6), and "he or she" is 佢(keoi5). Most current dictionaries use Jyutping, including words.hk and CC-Canto. Some older textbooks use Yale romanization instead.

Who speaks each language, and where

Mandarin is China's national language, and almost nine in ten people on the mainland could use it by 2025. Cantonese is the home language of most Hong Kongers. It's also spoken across much of Guangdong, in Macau and in older Chinese communities abroad.

Here's what the counts say:

Worldwide totals for Cantonese alone are guesses. The articles we read gave anything from 70 to 100 million, mostly without a source.

What language do they speak in Hong Kong?

Cantonese, overwhelmingly: 88.2% of residents usually speak it at home. Hong Kong's Basic Law makes "the Chinese language" and English official, without naming a spoken variety.

In practice, spoken Chinese there means Cantonese, and written Chinese means the Mandarin-based standard in Traditional characters. Mandarin is rising, as the chart above shows. English is the usual home language of 4.6% of residents.

Should you learn Mandarin or Cantonese?

Learn the language of the people you want to talk to. If that's a Cantonese-speaking family or a life in Hong Kong, learn Cantonese. Four quick questions will point you one way or the other:

When Cantonese is the right first choice

• Your family speaks it. A grandparent who grew up in Hong Kong or Guangzhou will feel the difference immediately. Mandarin won't get you there: remember the 3% sentence score. If you already understand some Cantonese from home, you're building on it, not starting over.
• You live, or will live, in Hong Kong. Office jobs often run in English, but daily life runs in Cantonese. A survey of 2,260 Hong Kong employers in 2012–13 found Cantonese was the default spoken language with local colleagues and customers. Putonghua was the default for dealings with mainland China.
• You love Hong Kong films, Cantopop or TVB. They're in Cantonese.

When Mandarin is the better start

• Your people are in mainland China, Taiwan or Singapore, or you don't have specific people in mind yet.
• You want a certificate. The main exams, HSK and TOCFL, test Mandarin.
• You want choice. Mandarin has far more learning resources: courses, apps, graded readers and tutors.

Mandarin also gives you the writing standard that Hong Kong uses for formal text. If you add Cantonese later, about half the core words and many sound patterns carry over.

Difficulty won't decide it for you. The US Foreign Service Institute puts both in its hardest category: about 88 weeks of full-time study, or 2,200 class hours, to reach professional speaking and listening.

Where to learn Cantonese, and where SayMei fits

SayMei teaches Mandarin only: we have no Cantonese course, no Jyutping and no Cantonese voice. If the tool pointed you to Cantonese, these are reasonable places to start. We haven't tested them, and we checked them in October 2026.

• Dictionaries:
  • words.hk, a crowd-written Hong Kong Cantonese dictionary with example sentences.
  • CC-Canto, a free Cantonese–English dictionary.
  • Pleco, a dictionary app with Cantonese readings built in.
• Pronunciation: jyutping.org, which teaches Jyutping with audio.
• Courses:
  • CantoneseClass101, with audio and video lessons.
  • Pimsleur Cantonese, for audio-first speaking.
  • LingQ, for learning from real Cantonese media.
• People:
  • online tutors, such as those on Preply
  • university courses, such as the Yale-China Chinese Language Centre at the Chinese University of Hong Kong

Duolingo's Cantonese course is offered only to people who already speak Mandarin, says Duolingo's blog, updated in June 2026.

What to do next

• Name the people you want to talk to. Family, city and favourite films should decide this, not difficulty.
• Start with Mandarin if no one comes to mind. It has the most courses, the main exams and the formal writing standard.
• Start with Cantonese for Cantonese-speaking family or Hong Kong. Mandarin won't carry a conversation there.
• Lean on the sound rules if you add the other language. Many shared words arrive with a predictable pronunciation.
• Learn the common Cantonese-only characters before reading Cantonese chat. A short list does most of the work.