Mandarin IPA

The International Phonetic Alphabet (IPA) is the universal sound key for languages: one symbol, one sound, no overlap. For Mandarin specifically, the IPA matters more than for most languages, because the standard way of writing Mandarin sounds in Roman letters, pinyin, is a Romanisation rather than a phonetic alphabet. Pinyin makes deliberate compromises so that every syllable can be typed on a keyboard with the 26 Latin letters; the IPA does not.

The result is that pinyin has letters doing jobs they would never do in any European language.

The pinyin q is not the English /kw/. The pinyin x is not the English /ks/. The pinyin c is not the English /k/ or /s/. The pinyin i sounds completely different after z, c, s, zh, ch, sh, r than it does after any other consonant.

None of these are mistakes; they are choices a 1958 committee made to keep pinyin typeable. The IPA disambiguates immediately. Read q as /tɕʰ/ and you know exactly what to do.

This page covers the 21 initial consonants of standard Mandarin in IPA, the vowels and finals with the pinyin spellings that mislead, the four tones in IPA tone-letter notation, the sandhi rules, and the erhua /ɚ/ ending. For more on pinyin itself, including the writing-system context and the typing system, see the pinyin page.

The 21 initials

Mandarin initials, the consonants that begin a syllable, organise by place of articulation and aspiration. Each row of the chart that follows shares an articulation point. Each column distinguishes by manner (unaspirated stop, aspirated stop, fricative, etc.).

Bilabials and labiodental

English uses voicing to distinguish "b" from "p" (the vocal cords vibrate during b, not during p). Mandarin uses aspiration: both consonants are voiceless but one has a puff of air after and the other does not.

Pinyin chose the letter b for the unaspirated one to match the English speaker's intuition that it sounds like "softer than p". In IPA, the unaspirated voiceless stop is /p/, the aspirated voiceless stop is /pʰ/.

Dental and alveolar

The pinyin c is /tsʰ/ in IPA, a single sound: aspirated "ts". It is not /k/ as in English "cat", not /s/ as in English "city", and not two separate consonants. It is one affricate (a stop followed immediately by its release as a fricative).

Retroflex

Retroflex consonants curl the tongue tip back so it points to or touches the area just behind the alveolar ridge (the bump behind the upper teeth). To produce /ʈʂ/, /ʈʂʰ/, /ʂ/: curl the tongue back as if you are trying to lick the roof of your mouth, then produce an English "j", "ch", or "sh" from that position. The result is darker, hollower, and unmistakably Mandarin.

The pinyin r is /ʐ/ in IPA, not the English /r/. Standard Mandarin /ʐ/ is the voiced version of /ʂ/ (sh).

It sounds halfway between English r and the s in "pleasure", with the tongue tip curled back. English speakers often substitute their own /r/ approximant; this is recognisable as foreign but understood. Some northern Chinese speakers produce /r/ as something closer to a true retroflex approximant /ɻ/, blurring the line.

Alveolo-palatal

This is the row that confuses English speakers most. The pinyin q is /tɕʰ/ in IPA. It is not the English /kw/ as in "queen". It is an aspirated affricate produced with the tongue flat against the area where the alveolar ridge meets the hard palate, lips spread (not rounded).

The pinyin x is /ɕ/ in IPA. It is not the English /ks/ as in "box". It is a soft "sh" sound with the tongue tip down behind the lower teeth and the tongue body raised toward the hard palate.

The fastest English starting point: say "she" and pay attention to the tongue. Now lower the tongue tip behind the lower teeth and let the tongue body do the work. That is /ɕ/.

The pinyin j is /tɕ/, the unaspirated partner of q. Like the English "j" in "jeep" but with the tongue forward and flat. It is not voiced like the English j; both /tɕ/ and /tɕʰ/ are voiceless. The difference between them is the puff of air after.

Velar

The pinyin h is /x/ in IPA, the same symbol as Spanish j and German Bach. It is produced further back in the throat than English /h/, with audible friction. Many Mandarin speakers (especially southern) realise it closer to English /h/, so substitution is tolerated. The standard target is the harsher /x/.

The three sibilant series, side by side

The hardest single thing about Mandarin initials for adult learners is keeping the three sibilant series straight. Each series has an unaspirated affricate, an aspirated affricate, and a fricative.

The series are mutually exclusive in their finals. The alveolo-palatal series (j, q, x) only ever combines with i, ü, or finals starting with i or ü. The dental and retroflex series never combine with i, ü, or finals starting with i or ü, only with the back vowels and the buzzy "i" (see below).

This complementary distribution is what lets pinyin spell j+i as ji without ambiguity: only the alveolo-palatal j can appear before i.

Vowels and finals

A final is what follows the initial: a vowel, a diphthong, or a vowel with a nasal coda. Mandarin has about 35 finals built from a small set of vowel sounds.

Monophthongs

The pinyin e alone is /ə/ in IPA, a schwa, not English "eh". The word gē (older brother) is /kɤ/ to /kə/, not "geh".

The pinyin o alone is a rounded back vowel /o/, only appearing after b, p, m, f and in a few other contexts.

The vowel /y/ (the umlauted u) is the front-rounded vowel of French tu and German Tür: say "ee", then round the lips while keeping the tongue forward. After j, q, x, and y the dots are dropped in pinyin spelling (ju is /tɕy/, not /tɕu/), but the sound is still /y/.

The buzzy "i"

The pinyin i has two completely different sounds depending on what precedes it.

The symbols for the second and third rows vary between transcriptions.

• After j, q, x, y and after the bilabials, dentals, etc.: the standard "ee" vowel.
• After the dental sibilants z, c, s: a buzzy continuation of the consonant, with the tongue staying in the dental position. zi is not /tsi/.
• After the retroflex sibilants zh, ch, sh, r: a buzzy retroflex vowel, with the tongue staying curled. shi is not /ʃi/.

The buzzy i is not a separate vowel sound English has any obvious equivalent for. The closest English analogy is the vocalic ending of "shh" when you shush someone: a held continuation of the fricative, with no clear vowel quality. Once you hear it, it stops being confusing.

Diphthongs

Nasal endings

The mismatches: where pinyin lies

A handful of common pinyin spellings mask what they sound like in IPA. These are the spellings most likely to mislead an adult learner.

Why each one is spelled that way:

• ian, not /ian/: the a is opened to an /ɛ/ before the n. tiān (day) rhymes with "yen", not "yan".
• ui: the e is spelled-out in slow speech, suppressed in spelling. guī (rule) is "gway", not "gwee".
• iu: same logic, the o is suppressed in spelling, pronounced in speech. liú (flow) is "lyoh", not "lyoo".
• un: /uən/ after most consonants, with the e suppressed. dūn is "dwun", not "doon".
• u after j, q, x, y is ü: ju is not /tɕu/. There is no /tɕu/ in standard Mandarin, which is what licenses the spelling shortcut.

These six spellings cause the bulk of pronunciation errors among self-taught Mandarin learners. The IPA shows you what is actually said.

The four tones in IPA

Mandarin is tonal: the pitch contour of a syllable is part of the word, not a layer of intonation on top. Get the tone wrong and you have said a different word.

IPA represents tone with tone letters, a column of vertical lines from 1 (low) to 5 (high), or with numeric superscripts. Below is each Mandarin tone in tone-letter notation, with the pinyin diacritic and a Chao tone-number representation (the system Chinese linguist Yuen Ren Chao introduced, on a 1-to-5 scale where 5 is highest).

The third tone in citation form is the famous dip: starts at 2 on the scale, drops to 1, rises to 4. In connected speech, the rising tail often disappears and the tone simplifies to a low half-third (just the 21 part).

The fifth tone (neutral) is not a separate pitch contour. It is the absence of a tonal target: the syllable is short, light, and its pitch is whatever falls naturally after the preceding tone. After a first tone, the neutral falls to a low pitch; after a third tone, it rises slightly.

It appears on grammatical particles (的 de, 了 le, 吗 ma), the second syllable of doubled nouns (妈妈 māma, 谢谢 xièxie), and many suffixes.

A full Mandarin word in IPA looks like:

• ma1 mother: /ma˥/
• ma2 hemp: /ma˧˥/
• ma3 horse: /ma˨˩˦/
• ma4 scold: /ma˥˩/
• ma0 question particle: /ma/ (neutral, pitch depends on context)

Or with the Chao numbers: /ma55/, /ma35/, /ma214/, /ma51/.

Tone sandhi in IPA

Tone sandhi rules are the changes in tone that happen when certain tones meet in connected speech. The pinyin is normally written with the citation tones, but the actual IPA realisation shifts.

Third plus third becomes second plus third. When two third tones meet, the first becomes a second tone:

• nǐ hǎo (你好, hello): citation /ni˨˩˦ xau˨˩˦/, actual /ni˧˥ xau˨˩˦/.

The character 一yī (yī, one) changes tone depending on what follows:

• Alone or at the end of a phrase: /i˥/.
• Before a fourth tone or neutral tone: becomes second tone, /i˧˥/. So yí gè /i˧˥ kɤ/.
• Before any other tone (1st, 2nd, 3rd): becomes fourth tone, /i˥˩/. So yì zhāng /i˥˩ ʈʂaŋ˥/.

The character 不bù (bù, not) changes from fourth to second tone before another fourth tone:

• Before any tone other than fourth: stays fourth, /pu˥˩/. So bù chī /pu˥˩ ʈʂʰi˥/.
• Before a fourth tone: becomes second, /pu˧˥/. So bú duì /pu˧˥ tuei˥˩/.

These are all automatic for native speakers. Learners do not need to memorise them as rules for life, only to be exposed to enough connected speech that the changes happen without conscious thought.

Erhua: /ɚ/

The character 儿 (ér) is /ɚ˧˥/ on its own (a rhotacised mid vowel with rising tone). As a suffix, it does not become a separate syllable; it modifies the end of the syllable it attaches to.

This is erhua (儿化), the rhotacisation of finals especially characteristic of northern Mandarin and the Beijing dialect.

The IPA realisation depends on what the final ends in. The tongue curls back at the end of the syllable, and the rhoticisation can swallow or modify the preceding sound:

• 花 huā /xua˥/, flower, becomes 花儿 huār /xuɚ˥/.
• 一点 yìdiǎn /i˥˩ tiɛn˨˩˦/, a little, becomes 一点儿 yìdiǎnr /i˥˩ tiɚ˨˩˦/. The n is swallowed.
• 这边 zhèbiān /ʈʂɤ˥˩ piɛn˥/, this side, becomes 这边儿 zhèbiānr /ʈʂɤ˥˩ piɚ˥/.

Standard Putonghua (the official spoken standard) tolerates erhua but does not require it. Learners can recognise it when they hear it and use it sparingly or not at all without sounding wrong. Heavy erhua is a marker of northern, especially Beijing, speech.

Why the IPA is essential for Mandarin

For most languages, the IPA is a useful tool. For Mandarin, it is closer to essential, because pinyin is built on compromises that mislead anyone who reads it as if it were English spelling.

The pinyin q read as English "kw" is unrecognisable. The pinyin x read as English "ks" is unrecognisable. The pinyin i after shi read as "ee" is wrong. None of these problems exist in IPA: /tɕʰ/, /ɕ/, and /ʂʐ̩/ all unambiguously specify what to say.

The pinyin page covers the practical, typeable system you will actually write Mandarin in. This page covers the underlying sound system the pinyin is trying to encode. Adult learners who hold both in their heads at once stop being surprised by Mandarin pronunciation traps. They have the answer key.