Shanghainese tones and tone sandhi, explained properly

Shanghainese has five tones, which sounds like more work than Mandarin. In practice it is less, because inside a word only the first syllable keeps its tone, and the rest of the word follows a melody that syllable sets.

Updated 13 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 five tones

Said on its own, a Shanghainese syllable carries one of five tones. Linguists describe tones with pitch numbers from 1, the lowest, to 5, the highest, so 53 is a fall from high to mid and 12 is a short rise at the bottom of the voice. These values are from Xu Baohua and Tang Zhenzhu’s standard 1988 description of urban Shanghainese:

Tone	Pitch	Kind of syllable	Example
1	53, falling	high register	家 ka ‘family’
5	34, rising	high register	好 hau ‘good’
6	23, low rising	low register	人 nyin ‘person’
7	55, short	high register, ends in a glottal stop	一 ih ‘one’
8	12, short	low register, ends in a glottal stop	学 ghoh ‘to study’

Other studies notate the same tones a little differently, writing tone 5 as 334 or tone 6 as 113, for instance. Those are differences of fine detail; the five-tone system itself is agreed.

The odd numbering is historical. Middle Chinese had eight tone categories: four tone classes, each split by whether the syllable began with a voiced or a voiceless consonant. Shanghainese still had eight when Joseph Edkins described it in 1853. The urban language has since merged them into five, which keep their old numbers, and older speakers are described with six.

The consonant predicts the register

The division into high and low register is not arbitrary. Low-register tones occur only after voiced consonants, and high-register tones only after voiceless ones. That means the spelling does much of the work. In the romanisation Verbavia uses, a syllable beginning with b, d, g, v, z or gh, or with m, n or l, is low; one beginning with a voiceless consonant is high; and one ending in -h is short and checked. Compared against the tone data in Xu and Tang’s description, the spelling predicts the tone class of about 97% of characters.

The exceptions are marked. An apostrophe before m, n, l and a few similar sounds shows a high register where the sound would normally be low, and it separates real words: 妈 'ma ‘mum’ and 买 ma ‘to buy’, 猫 'mau ‘cat’ and 毛 mau ‘hair’.

Tone sandhi: the first syllable decides

In running speech, a Shanghainese word does not string its syllables’ tones together. The tone of the first syllable spreads across the whole word, and the later syllables lose their own tones. This is called left-dominant tone sandhi. In the words of the phoneticians Chen Yiya and Carlos Gussenhoven, the tone of the first syllable determines the pitch contour of the whole unit.

First syllable	Two-syllable word	Three-syllable word
Tone 1 (53)	55 + 21	55 + 33 + 21
Tone 5 (34)	33 + 44	33 + 55 + 21
Tone 6 (23)	22 + 44	22 + 55 + 21
Tone 7 (55, short)	33 + 44	33 + 55 + 21
Tone 8 (12, short)	11 + 23	11 + 22 + 23

Read the table by rows and something useful appears. Whatever tone the second or third syllable would have on its own, a word beginning with tone 1 comes out high and then falling, and a word beginning with tone 6 comes out low and then rising. Tones 5 and 7 lead to the same melody. Detailed phonetic studies do find small effects from the second syllable, such as whether it is checked, but the pattern is set by the first.

Take 中国 tson kueh ‘China’. On their own, the two syllables are a tone 1 fall and a short high tone 7. Inside the word they become 55 + 21: the first syllable’s tone takes over and the second simply falls away. Mandarin 中国 Zhōngguó keeps both of its tones. Or take the Shanghai slang word 十三点 zeh sae tie ‘foolish’. Said one at a time, its syllables are low checked, falling and rising. As a word it is a single low rise: 11 + 22 + 23.

When a phrase keeps two melodies

Not every string of syllables is one word. Tightly bound words, such as compounds and adjective-noun combinations, take left-dominant sandhi. Looser combinations, especially a verb followed by its object, split into two units, and the second unit keeps its own tone. Descriptions of Shanghainese call these the broad and narrow patterns.

The classic illustration is 炒面 tshau mie. As a noun, fried noodles, it is one word with one melody: 33 + 44. As a verb phrase, to fry noodles, it is two units: the verb is levelled to 44, and noodles keeps its own rising tone, 113. The syllables are the same; the structure, and so the melody, is not.

Checked syllables

Tones 7 and 8 occur only on syllables closed by a glottal stop. They are noticeably shorter than the others, and they are what remains of the Middle Chinese entering tone, which Mandarin lost. In sandhi, a word beginning with tone 7 follows the same melody as one beginning with tone 5, while tone 8 has its own low pattern.

How this differs from Mandarin

Mandarin changes tones in context too, but through a few specific rules, and its best-known rule works in the opposite direction. When two third tones meet, the first one changes: 你好 nǐ hǎo is pronounced ní hǎo. The last syllable keeps its tone and the one before it adapts. Mandarin also has special rules for 一 yī and 不 bù, and a neutral tone on unstressed syllables. Outside those, a Mandarin syllable largely keeps its tone inside a word, which is why Mandarin learners drill tones syllable by syllable and in pairs.

In Shanghainese the unit that matters is the word. Knowing that a character has tone 6 tells you surprisingly little about how it sounds in the middle of a compound.

How to learn it

Learn words and short phrases as whole melodies, from recordings, rather than memorising each character’s tone on its own.
Pay attention to the first syllable. Its tone, which the spelling usually predicts, decides the shape of the whole word.
Notice structure. A noun and a verb phrase built from the same syllables can sound different.
Do not carry Mandarin tones across. A shared character usually has a different tone in Shanghainese, and inside a word it may have no tone of its own at all.