Mandarin tongue twisters are designed to be difficult, but not randomly so. Each one is built around a specific pair of sounds that some speakers find hard to distinguish, packing them as tightly as the language allows. Counting the sounds in pinyin shows what the twister is for, and who will find it hard.
The inventory being drilled
Standard Mandarin, as defined by the 1958 scheme that established pinyin, has exactly 21 initial consonants: b p m f, d t n l, g k h, j q x, zh ch sh r, z c s.
Many summaries list 23, adding y and w. The scheme itself does not. In its own tables y and w appear only as spelling conventions for syllables that begin with a vowel sound, not as initials. It is a small correction, but it matters when counting, because a syllable written yī has no consonant to drill.
Besides the initials, each syllable has a final (the vowel and ending) and one of four tones (plus a neutral one). Twisters can work on any of these, but they usually concentrate on pairs of initials or finals.
Counting one: the s and sh drill
The best-known twister, 十和四, sets the numbers four and ten against each other:
四是四，十是十，十四是十四，四十是四十……
sì shì sì, shí shì shí, shísì shì shísì, sìshí shì sìshí…
Across the thirty syllables of its full text, 22 begin with s or sh — eight with s and fourteen with sh. Nearly three syllables in every four force a choice between the two. The tones move constantly too: twenty of the twenty-nine steps between adjacent syllables change tone, because 四 is fourth tone and 十 is second.
The contrast is subtler than the spelling suggests. s and sh differ in where the tip of the tongue sits — against the teeth for s, curled back behind the ridge for sh. And the vowel written i in both is not the same vowel: the i after s and the i after sh are two different sounds, so 四 and 十 actually differ in two places at once.
Counting another: the n and l drill, and how it was built
The twister 牛郎恋刘娘 ("the cowherd loves Liu Niang") drills n against l.
牛郎恋刘娘，刘娘念牛郎……
niú láng liàn liú niáng, liú niáng niàn niú láng…
Of its first twenty-four syllables, every single one begins with n or l — fourteen with n, ten with l. There is nothing else to hide behind.
The tones are the interesting part. Only eight of the twenty-three steps change tone, because the twister uses second and fourth tones almost exclusively. With the tones held that still, the consonant is the only thing left changing.
Who finds which one hard
Each twister is hard for speakers whose own variety of Chinese merges the pair it drills. For those whose variety keeps the two apart, it's trivial.
| Pair | Merged in |
|---|---|
| zh ch sh vs z c s | Southwestern Mandarin, most Lower Yangtze Mandarin (though not Nanjing), Taiwan Mandarin, Cantonese; Shanghainese has no retroflex sounds at all |
| n vs l | Changsha and Chengdu merge towards n; Nanjing merges towards l |
| -n vs -ng | Lower Yangtze Mandarin, including Nanjing; Taiwan Mandarin, in some finals |
For a speaker from Chengdu, 四 and 是 are the same syllable, and so are 十 and 石. The opening of 十和四 — sì shì sì, shí shì shí — collapses to two alternating syllables.
The twister that is not hard for the speakers you would expect
Southwestern Mandarin, spoken around Chengdu, does merge -n with -ng — but only after certain vowels. In its readings, 金 and 京 are the same, and so are 根 and 耕: the in/ing and en/eng pairs are gone. Yet 板 and 綁 stay distinct, and so do 船 and 床. The an/ang and uan/uang pairs survive.
That matters because two of the best-known -n/-ng twisters are built on exactly the pairs that survive. 船和床 sets chuán against chuáng; 扁担长，板凳宽 sets bǎn against bǎng. For a Chengdu speaker, neither is especially hard. The twisters that would catch them are ones built on in against ing, or en against eng — a different set entirely.
What they do not drill
Mandarin has tone rules that change a syllable's sound depending on its neighbour. The best-known rule changes a third tone to a second tone when another third tone follows it.
Counting the four twisters above, there is not one third-tone-followed-by-third-tone sequence in any of them. The same was true across the wider set of eight examined for this article. The only tone changes they contain are the separate rules for 一 and 不.
They are not on the test
Tongue twisters are closely associated with the national Putonghua proficiency test, and are widely used to prepare for it. But the test itself contains no tongue-twister section.
Its syllabus sets five parts: a hundred single syllables, a hundred syllables of multi-syllable words, a section on standard and dialect usage that provinces may drop, a four-hundred-syllable passage read aloud, and three minutes of unscripted speech. The third-tone sandhi that twisters skip is examined deliberately in the second part, which must include at least three words with two consecutive third tones.
Twisters appear instead in the training material around the test, and in the pronunciation textbooks used on broadcasting and announcing courses, one of which carries them as an appendix of drills.
How the counting was done, and its limits
The inventory of initials is from the 1958 scheme itself, read from the education ministry's scan of the original. The phonetic descriptions of the sounds and of third-tone sandhi are from a published description of standard Beijing Chinese in the Journal of the International Phonetic Association. The dialect mergers are from linguistic surveys of each variety and from dictionary readings of Sichuanese. The test structure is from the national syllabus. Every syllable of each twister was checked against dictionary pronunciations before counting, and the counts were then made here, from that verified pinyin, on 21 September 2026.
The counts use each character's dictionary tone rather than the tone actually spoken after the rules for 一 and 不 apply; counting spoken tones would change the tone figures slightly, though not the consonant ones. And twister texts vary between sources — several lines have common variants — so the counts describe the versions quoted here rather than every version in circulation.
