It was related to me that Tang Dynasty poetry is meant to be pronounced in the style of the Cantonese dialect. Is this correct, incorrect, or a subject of contention?

  • 1
    Yes, Cantonese is closer to ancient Chinese than Mandarin. Tang poems do not correctly rhythm in Mandarin, but they do in Cantonese.
    – Tang Ho
    Commented Apr 21, 2017 at 19:55
  • 1
    Speakers of many different southern dialects claim that . A further example is 閩南語. It is essential that you examine not only the rhymes, but also the tonal patterns. These dialects typically have entering tones. When focusing on the rhymes, I was told, that 閩南語 has separate vernacular and official readings for most characters, so evaluating the rhyme might not be that straightforward!
    – Ludi
    Commented Apr 22, 2017 at 11:19

3 Answers 3


well, this answer is "limited" only, in relation to poetry in 唐 dynasty & cantonese.

1 cantonese (and several southern dialects/languages) is highly correlated to middle chinese, which was used in 唐 dynasty.

2 cantonese still has 入聲 (entering tone), while mandarin lost it. e.g.

屋 uk1 sound file

沃 yuk1 sound file

燭 juk1 sound file

in mandarin, these three are pronounced as:

屋 wū (陰平, dark level)

沃 wò (去, departing)

燭 zhú (陽平, dark level)

3 韻書 (rime dictionary), e.g. 切韻, 廣韻; one can find character based on cantonese, or other southern dialects / languages, but not mandarin.

4 poetry in 唐 dynasty had rules and patterns. compare to a language without entering tone, characters not in accord with rime dictionary; cantonese is a better choice.

last, there was a book 文鏡秘府論 by japanese monk 遍照金剛 (空海), which was written at about 809. it talks about the rules / patterns of poetry & chinese phonology in details.

the internet archive has a copy:


or, 文鏡秘府論匯校彙考

have a read, please.

  • 1
    I don't want to wage a war that I have done for countless times, especially not on stackexchange. But Please consider the fact that Guangzhou Cantonese is the only known extant language in China to have lost nearly all its medials (介音) No other dialect in China does that. How could you say that a dialect that has basically completely done away with its medial system is a better choice than Mandarin when Mandarin preserves medials really very well? Commented Apr 22, 2017 at 17:24
  • 1
    I'm not trying to be provocative, but cherry picking does not work. Each dialect has its own innovations and conservations and it is wrong to only look at one particular thing and say "that's the sole thing that matters". For one, the disappearance of medials nearly completely wrecked the rhyming system of Cantonese -- Middle Chinese */-uɑn/ and */-ɑn/ in Cantonese don't nearly rhyme -- they are un~yun and on; whereas, in Mandarin these do not only rhyme (uan and an), they keep the actual /an/ sound. Similarly MC */-iɑŋ/ and */-ɑŋ/ became C oeng and ong but M iang ang. Commented Apr 22, 2017 at 17:30
  • 2
    Ok, let me respond point by point. Point 1 is wrong: Mandarin is equally highly correlated to Middle Chinese, and the only reason people think Cantonese could be closer is that people use (the distance to) Mandarin as their standard for the archaicness/recentness of a dialect; this is not an objective standard. This comparison does not work, especially when one party is exactly Mandarin. Commented Apr 22, 2017 at 20:57
  • 1
    Point 3 is straight factually incorrect that people can use 韻書 (rime dictionaries) with Cantonese but not Mandarin. It is just factually wrong and unlike point 1, there even is no dispute about this. Rime dictionaries do not apply as much to Cantonese as it does not to Mandarin. Commented Apr 22, 2017 at 20:59
  • 1
    i opened another question, provided an answer, in lieu of comment here, have a look: chinese.stackexchange.com/questions/23502/… Commented Apr 23, 2017 at 4:05

This is incorrect. Cantonese is a modern language just like Mandarin, English and every single other modern language is.

Exactly speaking, Tang Dynasty poetry were pronounced back then in Tang Dynasty Chinese, which is called Middle Chinese.

Middle Chinese is different from any single existing Chinese language, and all contemporary Chinese languages have undergone drastic changes since then. Generally speaking there is no difference which dialect you use to read the poetry. Beijing Mandarin is as good as Guangzhou Cantonese.

Sometimes people think Cantonese is more conservative and closer to Middle Chinese than Mandarin. Well, this is not really true; it's a myth. Parts of the reason why people think so are, apart from cherry picking, misconceptions, smear attacks, and vague ethno-culturo-nationalist thoughts, and many other things:

1) Most people speak Standard Mandarin, the de facto national language, so they take whatever that is in SM as granted. And when they see features in Cantonese not present in SM but present in Classical Chinese, they notice, but they don't when it's the other way around.

  • One example is that Mandarin still uses the Classical existential word 在 ("(be) at"), but for the same function Cantonese has been using 喺 hai2 which lately developed from the Cantonese copula 系 hai6 with a grammatical tone sandhi. Here Mandarin is the "orthodox" or "more Classical/archaic" one, while Cantonese is the "newer" and counter-Classical one. But I have never seen ordinary people criticizing Cantonese or praising Mandarin for this.

2)The way historical Chinese linguistics has been describing how "old" a dialect is is by...checking how many sound changes the dialect has not shared with the contemporary lingua franca! Since Mandarin directly descended from practically all the lingua francas China has ever had, of course by this measure Mandarin is the "youngest" language of all Chinese. But...

(Essentially, when people in historical Chinese linguistics talk about how "old" a dialect is, they are actually talking about how much the dialect deviates from the lingua franca lineage. This is not a neutral measure, obviously.)

People have asked me to give examples where Cantonese is much less close to Middle Chinese than Mandarin in sounds -- well, there are a ton. I'll give two major example and a couple minor ones. The point of most of them -- Cantonese's vowel system is much more messed up compared to Middle Chinese than Mandarin's.

One major thing Cantonese uniquely has gone through is that it almost completely disposed of the rich medial system in Middle Chinese.

A quick recap, Chinese syllables are composed of maximally five to six parts, the onset, the medial, the nucleus, the coda, and a tone. The last four collectively are known as the rhyme 韵, and the last three as the final. One such maximal syllable is sian1, which has all five parts. This is the typical syllable structure for basically all Chinese languages today, including Mandarin. Some MC theories and also most of the Min language branch allow a glide before consonantal codas, so their maximal syllable is something like siaung1, but this is now not seen outside of the Min language branch.

The Guangzhou Cantonese syllable has no medial. Its maximal syllable structure is something like siu2 /siːw35/ or gwong2 /kʷɔŋ35/. This is because Cantonese did do away with basically all its medials, and the only extant independent sound segments that correspond to MC medials are the w thing seen after only two onsets -- k and g, and the initial consonants j (y) and w. Modern acoustic experiments and phonological analyses agree that these are now either part of a single consonant (kw gw) or its own consonant, and so in Cantonese the medial structure does not exist.

Middle Chinese, on the other hand, has an extremely rich medial system. Contemporary mainstream reconstructions have three basic medials, *r, *i, *u, and these are combined in any manner you wish (a combination, not permutation), so in MC there were 8 distinct medials. The words with medials are numerous as well. For just *i, more than fifty percent of Early MC syllables have it.

In fact, the MC rhymes were arranged based on their medials in contemporary dictionaries. People divided up rhymes into four tones, into many "final groups" according to the finals, and into four "levels/Deng" 等 and two "open-closed mouth" classes 开合(口) according to their medials. The third Deng 三等 had the *i medial, the second Deng 二等 had the *r medial without *i, and the first Deng 一等 and the fourth Deng 四等 had neither. The first Deng was comprised of syllables with nuclei */ɑ/, */ə/, */o/, */u/, while the fourth Deng was comprised of those with the nucleus */e/; it merged into the third Deng in later MC (i.e. these syllables acquired a medial *i). The "closed-mouth" 合口 were syllables with the *u medial, and the "open-mouth" 开口 were those without *u.

*r has disappeared in basically all dialects. In Northern Mandarin dialects it was idiosyncratically converted to i after dorsal sounds (k, g, ng, h, glottal stop, etc), preserving some traces.

One thing that has further happened to Guangzhou Cantonese, however, is that, except when it began the syllable, *i disappeared completely. For some finals, *i simply disappeared with no traces left, So GZC no longer distinguishes words like 九 (GZC gau2 MC */kiəu2/ BJM jiou3) and 狗 (GZC gau2 MC */kəu2/ BJM gou3), or words like 菊 (GZC guk1 MC */kiuk/ BJM ju2) and 谷 (GZC guk1 MC */kuk/ BJM gu3). For others, which is most finals, the *i medial merged with the nucleus, resulting in different vowel qualities. For example:

Middle Chinese */-iɛn/ and */-yɛn/ -> GZC in /-iːn/ and yun /-yːn/, and BJM ian /-iɛn/ and yuan /-yɛn/ if non-retroflex, and an /-an/ and uan /-uan/ after retroflex;

Similar things happened to the *u medial: it was preserved only after k, g, h, and word-initially, and even then hu- merged with f, leaving only ku- gu- and w-. As said above, in Modern GZC ku- and gu- have become single consonants kw and gw. They no longer contain a medial segment.

As a result, for many finals MC first and third Dengs and/or open and closed classes no longer rhyme. So a large class of words do not rhyme in Cantonese while they did in Middle Chinese and they do in Mandarin. For example:

Middle Chinese */-iɑŋ/ and */-ɑŋ/ -> GZC oeng /-œːŋ/ and ong /-ɔːŋ/ and BJM iang /-iɑŋ/ and ang /-ɑŋ/

The vowel quality for a vaster class words have also deviated much more from Middle Chinese than they are in Mandarin.

Here there is another major thing in Cantonese: many finals split after dorsal/labial onsets (g, h, ng, p, f, m, ...) and coronal onsets (t, d, s, l, n...), making them even no longer rhyme.

For example

Middle Chinese */-ɑn/ and */-uɑn/ ->

GZC {aan /-aːn/ after coronals, on /-ɔːn/ otherwise} and {yun /-yːn/ after coronals, un /-uːn/ otherwise}

(/y/ is the vowel of French u and German long ü; it is also /i/, English ee with lips rounded)

BJM an /-an/ and uan /-uan/;

Middle Chinese */-ən/, */-in/, */-uən/, */-yin/ ->

GZC an /-ɐn/, an /-ɐn/, eon /-ɵn/, eon /-ɵn/

BJM en /-ən/, {in /-in/ if non-retroflex and en /-ən/ after retroflex)

uen /-uən/, {yun /-yn~yin/ if non-retroflex and uen /-uən/ after retroflex);

Middle Chinese */-uo/ ->

GZC u /-uː/ after k g f and word initially, disappears after ng, and ou /-ou/ otherwise

BJM u /-u/;

For the minor ones:

  • Cantonese has lost the Middle Chinese retroflex consonants, which are preserved in Beijing Mandarin.

    • For what would become sets of coronal affricates in Later Chinese, Early Middle Chinese makes a four-way distinction. It distinguished between retroflex stops, retroflex affricates/fricatives, postalveolar affricates/fricatives, and alveolar affricates/fricatives. All four sets collapsed into one in Guangzhou Cantonese, while Beijing Mandarin still preserves a two way distinction, largely the first three groups with the last.
  • Cantonese, like most Southern Chinese dialects, have shifted its MC */ɑ/ completely to */ɔ/. This is only seen in Mandarin for MC syllables with nucleus */ɑ/ and no coda. About the largest set of words in MC had */ɑ/, so this is pretty significant.

  • The second and third largest vowel class in MC were */e/ and */ə/. Both Cantonese and Mandarin have messed around with */e/, but MC */ə/ stays /ə/ in Mandarin while it has become a more open /ɐ/ in Cantonese, which sounds like British English (traditional Received Pronunciation) sound uh in cut, instead of the first vowel in the English word about.

    • Since MC */i/, the fourth most common vowel, merged with */ə/ in both Cantonese and Mandarin when there is a coda, and in particular Cantonese does not retain the medial (MC */i/ > Mandarin /iə/ when there is coda), its effects are also not insignificant. i.e. MC */i/ > Cantonese /ɐ/ whenever there is a coda.

      • This can be compared with Mandarin, where /iə/ is [i] unless there is no coda (so MC */i/ > Mandarin [i] anyway)
  • Cantonese diphthongized its original high vowels /iː/ /uː/ /yː/ in the 19th century, so they are now /ei/ /ou/ /œy/ (stay, know, stay with rounded lips).

    • Further, older Cantonese */uiː/ has now become /(w)ɐi/ (North American English (w)ife). */uiː/ and */iː/ differ only by a medial and have the same final, so they used to rhyme like in MC, but now they no longer do.
  • The MC onset */kʰ/ (English k) have become /h/ in Cantonese irregularly in large numbers.

  • The earlier Cantonese sequence */hw-/ became f in Modern Cantonese. Coupled with the change above you get pronunciations that do not appear to be regular but actually basically are: 科 MC */kʰuɑ1/ Cantonese fo1 /fɔː1/, 阔 MC */kʰuɑt1/ Cantonese fut3 /fuːt3/,

In short, as said at the beginning, every dialect has its own numerous innovations (=deviations) and accordingly preservations. The innovations of each dialect happen in different aspects and different directions, so every dialect preserves its own parts of Middle Chinese while having lost its own parts of the Middle Chinese traits. No dialect preserves "objectively the closest" sound system and correspondence to Middle Chinese, because in this reality it does not make sense to speak of this single, unitary "closeness" dimension relative to Middle Chinese.

Now for the rhyming myth...Let's try out a famous poem by 李绅, titled 悯农:





In contemporary Beijing Mandarin (obsolete literary readings are not featured):

chu2 he2 ri4 dang1 wu3
han4 di1 he2 xia4 tu3
shui2 zhi1 pan2 zhong1 can1
li4 li4 jie1 xin1 ku3

In conservative (i.e. preserving more older features, and still in use) Guangzhou Cantonese:

co4 wo4 jat6 dong1 ng5
hon6 dik1 wo4 haa6 tou2
seoi4 zi1 pun4 zung1 caan1
nap1 nap1 gaai1 san1 fu2

So here you have BJM rhyming in -u3, and GZC rhyming in...ng5/ou2/u2? Clearly Guangzhou Cantonese does not rhyme here.

Another example, the arguably single best known Tang poem, by 李白, titled 静夜思:





In Mandarin:

chuang2 qian2 ming2 yue4 guang1
yi2 shi4 di4 shang4 shuang1
ju3 tou2 wang4 ming2 yue4
di1 tou2 si1 gu4 xiang1

While in Cantonese:

cong4 cin4 ming4 jyut6 gwong1
ji4 si6 dei6 soeng6 soeng1
geoi2 tau4 mong6 ming4 jyut6
dai1 tau4 si1 gu3 hoeng1

Here again, in Cantonese this does not rhyme because it converted MC *iɑŋ to oeng, but MC *uɑŋ to (w)ong. In Mandarin this does rhyme; the MC *ɑŋ final is simply preserved. In MC 霜 had the *iɑŋ rhyme, which in Mandarin was converted to uang after retroflex onsets. Cantonese actually went through this sound change as well; 霜's pronunciation of soeng1 is actually irregular, and would have been song1 if it were regular.

  • 1
    @Rethliopuks. Your example of 喺 and 在 is not a good one. In Cantonese, 喺 /hai2/ is the colloquial counterpart of the literal Cantonese 在 / zoi6 /. In Cantonese, speaking and writing are separated-- we don't speak the same as we write, and we read Tang poems which are literal material with literal reading.
    – Tang Ho
    Commented Apr 22, 2017 at 8:36
  • 3
    It is 谁知盘中飧(sūn) zdic.net/z/28/js/98E7.htm not 谁知盘中餐
    – Tang Ho
    Commented Apr 22, 2017 at 11:24
  • 1
    飧 [syu1] (juyping); [shu1/] (pinyin)
    – Tang Ho
    Commented Apr 22, 2017 at 11:31
  • 1
    One instance does not a rule make. One needs to take into consideration the vast body of Tang poetry to determine which modern dialect is closer.
    – monalisa
    Commented Apr 22, 2017 at 16:41
  • 2
    @Rethliopuks This answer has been extraordinarily instructive, and is a resource I'm sure I will be returning to often. I find the statement "Generally speaking there is no difference which dialect you use to read the poetry. Beijing Mandarin is as good as Guangzhou Cantonese" to be particularly attractive because subjectivity is a functional component of all fine arts, and capacity for reinterpretation has been proposed as a quality related to longevity of a text in terms of appreciation and value.
    – DukeZhou
    Commented Apr 22, 2017 at 23:25

水巷孑蠻 and Rethliopuks gave quite a lot of details. But this does not address quite right to the question.

Read the origin question: "It was related to me that Tang Dynasty poetry is meant to be pronounced in the style of the Cantonese dialect. Is this correct, incorrect, or a subject of contention?"

The question is not about the actual pronunciation in Tang Dynasty. The keyword is Tang Dynasty poetry and the style of the Cantonese dialect.

Some common senses

We speak not like the people in a thousand year ago. A language evolves everyday by its speakers. The discussion of actual pronunciation is not right. There were wide variety of Chinese languages in Tang Dynasty as in nowadays.

Unlike Western languages, Chinese language writes no phonetic value. The rime books in Sui, Tang and Song just categorise types of rimes, it do not dictate its phonetic value. The preface of rime book 切韻 says that "評論古今是非、南北通塞、編成《切韻》一書。" They tried to distinguish the rimes in the past and now, the languages of north and south China as fine as possible. You can expected some categories indistinguishable in some Chinese languages in Sui, Tang, and Song. The language in 長安 is not the same as in 洛陽. Every village had its set of phonetic values.

The phonetic values in current publications about Middle Chinese are just a form of educated guess. They are derived from the rime of poetry, the Buddhist name in the translation of Sanskrit, the related documents in Korean and Japanese, and most importantly from the current Chinese languages.

Among numerous rimes, they are grouped under four categories, namely 平 (flat), 上 (upward), 去 (leaving), 入 (entering). They are four tones in the Chinese languages. There are large number of words in 平 tone and some less in 上, 去, and 入. In poetry, 上, 去, 入 are grouped in 仄 (non-flat) in contrast to 平. The use of 平仄 is the basic construction of poetry from Sui and afterwards.

入 tone is special that it must end in -p, -t, -k. On the contrary, No -p, -t, or -k are in 平, 上, 去 tones. Some degenerated in later languages ends with -?. Mandarin loses the 入 tone completely due to the loss of -p, -t, and -k.

Tang Dynasty poetry

Tang poetry is musical. The core values of Tang Dynasty poetry is the strict use of 平仄 (the flat tone vs other tones, namely, non-flat tone)). The long and short, the high and low, produce some musical effect. The basic types of poetry are 絕詩 (four-line poetry) vs 律詩 (eight-line poetry), and 七言 (poetry seven-word in a line ) vs 五言 (poetry with five words in a line). There are some longer types like 樂府. Each have some rules on 平仄 to make poetry sounds good.

押韻 - Rhyme usually occurs in the last word of a line.

For example, 五言絕詩 is a poetry of four lines and each line with five words. Its rules is not easy to explain in a few words. Here is some basic rules.

  1. It is not necessary to rhyme first line.
  2. Rhyme the last word of the second and fourth line. The rime must be in 平 tone.
  3. The last word of the third line must be in 仄 tone.

Some common patterns are in the following:

Pattern 1



Pattern 2



Pattern 3



Pattern 4



The one in () is suggestion only.

Let us examine an early Tang poetry《相思》by 王維



We extract this the four tone first.

平去平平入 平平入上平

去平平上入 上入去平平

Then translate to 平仄

平仄平平仄 平平仄仄平

仄平平仄仄 仄仄仄平平

It deviates from the common patterns but it follows the rules above.

  1. 枝 and 思 is the in same rime of 平 tone.
  2. 擷 is 入 tone and it is 仄 tone.

As you see above, the use of 平仄 is essential to Tang poetry. Let's turn to examine Cantonese

The style of the Cantonese dialect

Cantonese is not a dialect. It is a language, one of the Chinese languages.

Cantonese is a tonic language with 9 tones (or 10 tones). But it can be categorized like those old rime books.

平 - 陰平 (high flat), 陽平 (low flat)

上 - 陰上 (high upward), 陽上 (low upward)

去 - 陰去 (high leaving), 陽去 (low leaving)

入 - 陰入 (high upward), 中入 (middle entering), 陽入 (low entering)

We don't expect Cantonese sounds like those speeches in Tang. If one use this as augment to talk about Tang poetry, he got it wrong. When we come to Tang poetry, we concern only that "how well Cantonese follows the rule of 平仄 and rhyme of Tang poetry?" Cantonese largely obey these rules. There are tons of examples.

In compare to other Chinese languages, the 介音 (middle vowels) in Cantonese are degenerated. But when it comes to poetry, it does not affect rhyme much.

Most works of poetry like《相思》by 王維 rhyme well and follow the 平仄 rules well in Cantonese pronunciation. But we can't expect there is no exception as languages evolve. The poetry of 憫農 is an outliner and one should not use it to make the conclusion that Cantonese does not fit at large. It is misleading.

Not all Chinese languages evolves equally but Cantonese evolves in the way fit well in Tang poetry. On the contrary, Mandarin evolves in the way that the loss of -m, -p, -t, -k, hurts the alignment to many families of rimes, and violates the 平仄 of Tang poetry. Mandarin does not fit well in Tang poetry.

  • your answer is clear & concise, i must learn from now on :) may i ask one thing: what is the rationale of 10 tones "Cantonese is a tonic language with 9 tones (or 10 tones)."? Commented Apr 23, 2017 at 11:40
  • Some counts 超平 (namely, extra-high flat tone) as one tone. It is a bit higher pitch than 陰平。"千" (thousand) is 陰平 tone. When one say a man "出老千" (cheating on gambling), the pitch of "千" is a bit higher than normal.
    – OmniBus
    Commented Apr 23, 2017 at 14:30
  • It is incorrect to say that the medials' merging with the main vowel and their disappearance "do not affect rhyme much". These precisely do affect it a lot, which is a huge reason I don't recommend Duke Zhou to use Cantonese. Commented Apr 28, 2017 at 5:10
  • Also, 悯农 was not illustrated as an outlier. By citing 悯农 I was pointing out how the very common and usual MC *-uo final gets messed up in Cantonese, systematically and categorically -- as I said in my answer a bit above the example, that sound was split in three ways that absolutely do not rhyme depending on the initial consonant. Commented Apr 28, 2017 at 5:36
  • 1
    It is better to define the 入 (entering) tone for those not familiar with it. 入 tone is the final stop. Losing this stop make Mandarin loses this tone and also a great number of rimes. If you read my introduction carefully, I've already pointed out some languages degenerating this stop from -p, -t, -k to -?. And, Mandarin loses it totally. That seriously the affect the quality of reading Tang poetry. The -p, -t, -k gives Tang poetry a definite short and forceful sounds that are not those degenerated ones can compare. Cantonese fits best here in Tang poetry.
    – OmniBus
    Commented May 2, 2017 at 13:51

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.