How Chinese compares to other languages: a linguistic analysis

Is Chinese related to Japanese, Korean or Vietnamese?

No. Mandarin is a Sino-Tibetan language. Japanese and Korean each belong to small families of their own, Vietnamese is Austroasiatic, Arabic is a Semitic language in the Afroasiatic family, and English, Spanish, French and German are Indo-European, according to WALS's language data.

What links Mandarin with Japanese, Korean and Vietnamese is two thousand years of borrowing from written Chinese, not a common ancestor.

You can see the difference between borrowing and ancestry in basic words. ASJP keeps a 40-word list for thousands of languages, with words like "I", "two", "eye", "water", "fire" and "name" that languages rarely borrow.

We compared Mandarin's list with each language's using LDND, the normalised edit distance of the Wichmann method. On its scale, about 100 means the words are no more alike than random pairs.

Only Cantonese, Mandarin's sister language, scores clearly below chance, at 80.5. Korean, Vietnamese and Japanese all score near 100, about as far from Mandarin as English, French and Arabic. The chart in the next section has every score.

For scale, English and German score 75.4, and English and Spanish 94.1, by our computation on ASJP's word lists.

Japanese, Korean and Vietnamese sit at chance level because their core words are native. Here's "eye" in all four:

The Chinese loans come in a later, more educated layer of the vocabulary. That's where the next sections find them.

Grammar puts Vietnamese, Korean and Japanese closest

On grammar, the order is the same whichever way we count. Vietnamese, Korean and Japanese are closer to Mandarin than any of the European languages or Arabic.

Thai, an unrelated language we added as a reference, lands right next to them. That's the first warning about what "close" means here:

We measured grammar distance as the share of WALS features whose values differ between two languages. WALS, the World Atlas of Language Structures, codes up to 192 features per language, from "does the language have tones?" to "does the relative clause come before the noun?". We used two feature sets:

• 36 learner features: the ones a learner notices, chosen before we looked at results. They cover tones, syllables, verb endings, gender, plurals, cases, tense and aspect, measure words, word order and questions.
• All shared features: every WALS feature coded for at least seven of the nine languages, minus a handful of word-meaning features, up to 143 per pair.

English is the interesting case. On the 36 features a learner feels, it differs from Mandarin on 78%, near the bottom.

On everything WALS codes, it differs on just 46%, statistically tied with Japanese: when we resampled the features, their ranges overlapped almost entirely. English and Mandarin share a lot of quiet structure, like subject-verb-object order, no case endings on nouns and few verb forms next to Spanish or Korean.

What they don't share is everything you hear in the first lesson.

Measured from English, German is closest, differing on 18% of learner features, and French and Spanish come next. Mandarin differs on 78%, the most of the nine.

Tones, short syllables and a few new sounds

Mandarin's sound system isn't unusually big. PHOIBLE, a database of sound inventories, lists 24 to 27 consonants for Mandarin across its four descriptions.

That's the same range as English across its nine, and WALS puts both in its "average" class. The differences are which sounds you use, and what pitch does.

Pitch is the big one. Of the nine languages, only Mandarin and Vietnamese use it to tell words apart, WALS finds; so do Cantonese and Thai, our two reference languages. Here's how each language uses pitch:

Vietnamese speakers already work this way. Everyone else builds a new habit, and our tone pair trainer plays all 20 tone combinations to help. Our guide to learning Mandarin tones explains how adults get them right.

Mandarin syllables are short, too. They end in a vowel, -n or -ng: 中(zhōng), 南(nán), 好(hǎo).

By our count, CC-CEDICT's single-character readings use just 419 distinct syllables, or 1,291 once tones are counted. English, French, German and Arabic allow far heavier consonant clusters.

Most speakers meet the same short list of new sounds. Here's how each language starts, based on PHOIBLE's sound inventories:

Source: SayMei reading of PHOIBLE inventories 2175, 2210, 2182, 2184, 2196, 2197, 2233 and 2157 (CC BY-SA 3.0) plus WALS 11A and 13A. Not yet reviewed by a phonetician.

The pinyin chart has audio for every syllable, if you want to hear these.

The grammar drops endings and adds three habits

Mandarin removes most of what makes European grammar heavy. Then it adds three things of its own: measure words, aspect particles and a strict "modifiers first" order.

Here are all 36 learner features, side by side. Filled cells work the same way as in Mandarin:

What Mandarin drops

Verbs never change. "I eat" is 我吃(wǒ chī), "he eats" is 他吃(tā chī), and "we eat" is 我们吃(wǒmen chī).

WALS counts zero to one inflectional category per Mandarin verb. Spanish, French and Japanese have four to five, and Korean and Egyptian Arabic six to seven.

There's no grammatical gender. "He", "she" and "it", 他, 她 and 它, differ only in writing, and all three are pronounced tā.

Nouns don't take plural endings: "one book" is 一本书(yì běn shū), and "three books" is 三本书(sān běn shū). The suffix 们(men) is used only for people, and optionally, exactly the WALS value for Japanese.

There's no past tense, either. Time words carry the time: "I ate yesterday" is 我昨天吃了(wǒ zuótiān chī le), and "I'll eat tomorrow" is 我明天吃(wǒ míngtiān chī).

What Mandarin adds

That 了(le) in "I ate yesterday" is an aspect marker. It says the action is complete, not that it happened in the past.

WALS codes Mandarin, Spanish, French, Korean and Arabic as marking aspect grammatically, and English, German, Japanese and Vietnamese as not. So Spanish and French speakers already have the concept, as in comí versus comía, or j'ai mangé versus je mangeais, even though 了(le) isn't a past tense.

Every counted noun needs a measure word:

• "a book": 一本书(yì běn shū)
• "a sheet of paper": 一张纸(yì zhāng zhǐ)
• "a cat": 一只猫(yì zhī māo)

Japanese, Korean and Vietnamese speakers have counters already; English, French, German and Arabic speakers don't. The measure word finder lists the 36 you'll use most.

Adjectives also work like verbs, so "I am busy" has no "am": 我很忙(wǒ hěn máng).

What Mandarin rearranges

The basic order is subject-verb-object, as in English, Spanish, French and Vietnamese. But everything that describes a noun goes before it, however long, and only Japanese and Korean share that habit. Place, time and questions follow their own rules:

• Descriptions first: "the book I bought yesterday" is 我昨天买的书(wǒ zuótiān mǎi de shū), literally "I yesterday buy DE book".
• Place before the verb: "I eat at home" is 我在家吃饭(wǒ zài jiā chīfàn). Time goes there too.
• Yes-or-no questions keep statement order: "Are you a student?" is 你是学生吗？ nǐ shì xuésheng ma?, which adds 吗(ma) at the end, the way Japanese adds か ka.
• Question words stay put: "Where are you going?" is 你去哪儿？ nǐ qù nǎr?, with "where" left where the answer goes.

Borrowed words give some learners a head start

An English speaker learning Spanish recognises información and problema on day one. An English speaker learning Mandarin gets almost nothing, and the data shows why.

Mandarin borrows less than any language in the World Loanword Database. By our count, just 26 of the 2,130 words in its list are clearly or probably borrowed.

That's 1.2%, the lowest of the database's 41 languages. The next lowest, Old High German, borrowed 5.6%.

English contributed two of Mandarin's loans: "lemon", 柠檬(níngméng), and "bus", 巴士(bāshì). For new things, Mandarin usually builds words from its own parts: a computer is 电脑(diànnǎo), literally "electric brain". So friendly loanwords like "coffee", 咖啡(kāfēi), and "sofa", 沙发(shāfā), stay a short list.

Japanese, Korean and Vietnamese went the other way. They borrowed so much from written Chinese that linguists call the shared layer Sino-Xenic vocabulary:

Sources: Mandarin forms checked against CC-CEDICT; Japanese, Korean and Vietnamese forms against Wiktionary. Not yet reviewed by native speakers.

How big is that layer? It depends on what you count:

Sources: our counts on WOLD v4.2; for Korean, the Standard Korean Language Dictionary as counted by Jeong 2000, NIKL.

The WOLD lists cover about 1,460 everyday meanings, while the Korean figure counts a full dictionary, technical terms included.

That's why figures of 50–70% circulate online: a dictionary count weighs in thousands of rare, formal nouns.

The Korean dictionary also shows where the layer sits:

Source: Jeong 2000, NIKL.

The words transfer; the grammar doesn't.

Even shared words need care. The sounds have drifted for centuries, and some meanings moved: in Japanese, 手紙, read tegami, is a letter, while Mandarin 手纸(shǒuzhǐ) is toilet paper.

One word did travel everywhere. In all nine languages, the word for tea comes from Chinese, WALS shows:

• Mandarin, Japanese, Korean, Vietnamese and Arabic took cha, as in Mandarin 茶(chá).
• English, Spanish, French and German took Min Nan te. English got it through Dutch thee, according to WOLD.

Only Japanese still writes with Chinese characters

Mandarin is written in characters, each usually a syllable with a meaning, and pinyin is the Latin-letter bridge for learners. Of the eight languages, only Japanese still writes everyday text with Chinese characters.

Here's how each one writes today:

Sources: Agency for Cultural Affairs, 2010; Kim 2015; Vietnamese Nôm Preservation Foundation.

For Japanese readers, shared characters often keep their meaning, but the readings differ completely, and many characters were simplified differently in China and Japan. "Library" is 图书馆(túshūguǎn) in Mandarin and 図書館, read toshokan, in Japanese.

Korea's list of basic hanja was last adjusted in 2000. Knowing hanja helps with Mandarin, but daily reading in Korean doesn't use them.

Vietnamese was written for centuries in Classical Chinese characters and in chữ Nôm, a character script built for Vietnamese, before the Latin-based quốc ngữ replaced them. So Vietnamese learners get the vocabulary but not the writing.

How many characters does a Mandarin reader need? In film and TV subtitles, 1,111 characters cover 95% of running text.

To cover 99%, you need 2,076, as our characters guide found using SUBTLEX-CH. For the two script standards in China and Taiwan, see Simplified vs Traditional Chinese.

Distance only partly predicts learning time

These distances line up only partly with official learning times, and those times exist only for English speakers. Here's how long the US Foreign Service Institute budgets for professional speaking and listening:

Source: US Department of State, FSI language categories, checked October 2026. Cantonese and Arabic share Mandarin's 88 weeks, which come to 2,200 class hours.

Set that next to the grammar distances. Vietnamese is the closest of the eight to Mandarin in grammar, yet for an English speaker it's in Category III, while Japanese and Korean are in Category IV like Mandarin. Writing systems and vocabulary weigh at least as much as grammar.

FSI publishes no estimates for speakers of other languages, so there's no official number for a Korean or Spanish speaker learning Mandarin. Our guides to whether Chinese is hard to learn and how long it takes to learn Mandarin cover difficulty from the learner's side and the hours.

Two caveats on the FSI page itself:

• Its hours don't multiply out. 88 weeks at the stated 23 class hours a week is 2,024 hours, not 2,200.
• It changed recently. Between late 2024 and 2026, its goal changed from speaking and reading to speaking and listening, and so did its hours for Categories I to III.

What these numbers don't mean

• Distance isn't difficulty. A feature you lack can be easy, like having no gender to learn, or hard, like tones. The cards label each one, and the labels are our reading of the data.
• The feature list changes the answer. English moves from near the bottom to the middle depending on which WALS features you count. We publish both sets and the 95% ranges.
• WALS codes the dominant pattern of a standard variety. Arabic here is Egyptian Arabic, the best-covered Arabic in WALS, with 145 features against 30 for Modern Standard Arabic. Modern Standard Arabic puts the verb first and question words at the front.
• LDND is coarse. Scores in the 90s don't separate distant relatives from unrelated languages: English and Spanish, both Indo-European, score 94.1.
• Loanword shares depend on the word list. WOLD's everyday list and a full dictionary give very different percentages for the same language.
• No learner data. None of this measures how fast real learners progress.

What to do next

• Pick your language in the tool above. See which features will feel familiar and which will need rewiring.
• Train tones from the start, unless you speak Vietnamese. Of the eight, only Vietnamese has lexical tone, as Mandarin does.
• Use your borrowed words if you speak Japanese, Korean or Vietnamese. Expect familiar meanings in new sounds, and watch for false friends.
• Expect the first lessons to feel new, even from English. The shared structure is quiet; tones, measure words and word order aren't.
• Hear it for yourself. The quickest way to feel the difference is to hear the tones and say a few sentences out loud.