Jump to content
Learn Chinese in China

A merging of the alveolar and retroflex initials in the Sinitic language family


Recommended Posts

  • New Members

Hello, I'm new here. I am not studying Chinese at a school as of currently, I'm therefore more or less a layman. My field of study is applied mathematics and that takes up all of my current time. I do have a need to learn Mandarin however, which is how I got into this.

I have an interest in phonetics and while I was in Guangzhou with my partner (her city of residence) I noticed that her family, originally from Hunan, and her friends, mostly from Guangdong, always merged the retroflex initials /ʈʂ/ [拼音: zh], /ʈʂʰ/ [拼音: ch], and /ʂ/ [拼音: sh], with their respective alveolar equivalents, i.e. /ts/ [拼音: z], /tsʰ/ [拼音: c], and /s/ [拼音: s]. I noticed the same with a couple of Shanghainese speakers. Both Cantonese and Shanghainese share the lack of /ʈʂ/, /ʈʂʰ/, and /ʂ/, which could be the explanation of their alveolarized Mandarin. I actually don't know about Hunanese however, maybe someone here could fill me in on Hunanese phonetics. But my questions:

  • What's the etymology of said alveolarization? (when were the retroflex initials dropped)
  • How come Mandarin have retroflex initials?
  • Which dialects have it, which dialects doesn't?

Hit me with all you got! :)

Link to comment
Share on other sites

Site Sponsors:
Pleco for iPhone / Android iPhone & Android Chinese dictionary: camera & hand- writing input, flashcards, audio.
Study Chinese in Kunming 1-1 classes, qualified teachers and unique teaching methods in the Spring City.
Learn Chinese Characters Learn 2289 Chinese Characters in 90 Days with a Unique Flash Card System.
Hacking Chinese Tips and strategies for how to learn Chinese more efficiently
Popup Chinese Translator Understand Chinese inside any Windows application, website or PDF.
Chinese Grammar Wiki All Chinese grammar, organised by level, all in one place.

Interesting questions.

According to what is apparently the currently accepted linguistic community consensus, Old Chinese didn't have the retroflex initials, then Middle Chinese (both Early and Late) developed the retroflex initials.

So that explains "How come Mandarin have retroflex initials?".

I didn't bother looking into Shanghainese, but Cantonese (which is usually more conservative than Mandarin when it comes to preserving Middle Chinese-ness) for whatever reason merged the Middle Chinese retroflex sibilants with the alveolar sibilants. My surface wiki skim has not provided anything as to precisely when.

As for your dialects question, you're going to have to be a bit more precise with your wording. Do you mean which Chinese varieties have retroflexes and which don't?

Link to comment
Share on other sites

  • New Members

Yes, I mean which Chinese dialects has voiceless retroflex affricates and fricatives. All of Wu (as far as I know, Shanghainese), Yue (well, only Cantonese as far as I know), and Xiang (again, I can't be certain that all Xiang dialects merge), come from Middle Chinese but still (as I've heard) dropped the retroflex sibilants of Middle Chinese? Highly intriguing that the retroflex initials emerged, isn't merging more common than the adding of consonants?

Link to comment
Share on other sites

Disclaimer: am also an amateur with respect to phonetics and phonology, especially the diachronic kind.

I think it is fairly well-established that most Middle Chinese retroflex stops and sibilants emerged from the Old Chinese dental versions [stops and sibilants] + /r/. Additionally, Early Middle Chinese had separate palatal stops and sibilants; these came from the Old Chinese dentals + /j/ (quite analogous to "tune" and "sure" in British English).

Naturally there are some exceptions: the Wikipedia article gives as being velar in Old Chinese but palatal in Early Middle Chinese [compare which is velar in Old Chinese, in Middle Chinese, and in Mandarin and Cantonese]. Hence its emergence follows fairly standard patterns (compare retroflex consonants in Norwegian and Swedish). Also, a mysterious distinction made in the Qièyùn 切韵 (the thing about some of the grade-III syllables having a column 3 vs 4 distinction) implies [according to some] that you can have both /r/ and /j/ in the same Old Chinese syllable.

Between the late Sui dynasty when the Qièyùn 切韵 was compiled and early Song when the Guǎngyùn 广韵 was done, the retroflex and palatal groups merged; apparently the retroflex dominated. Hence, with the tenuis affricates, 莊 and 章 are subsumed under 照; with aspirates, 初 and 昌 are subsumed under 穿; with the voiced, 崇 and 禪 are subsumed under 牀[/床 in modern orthographies].

The voiceless sibilants 生 and 書 are merged under 審, while their voiced counterparts undergo a bit of reshuffle: the rare sound 俟 goes under 牀 and then 船 becomes 禪 [yes, it resurfaces under a different classification], but it turns out 禪 and 船 from the Early Middle Chinese tables had probably been swapped and that 牀 disappears in slightly later editions, having merged into 禪. Complicated.

The palatal nasal 日, on the other hand, stays separate from retroflex nasal 娘; it seems to have become retroflex and to have lost its nasalisation in any case.

Additionally (as there are no palatal 'stops' in Early Middle Chinese) the retroflex non-nasal stops 知, 徹, 澄 remain separate.

So, come the second millennium CE, palatals have been wiped from the phonemes of the language, and retroflex consonants dominate amongst the 36 initials of the Song dynasty rime tables. However the rime tables are meant to be a representation of the various pronunciations of the Northern and Southern Dynasty period from the point of view of conservative phonologists in the Song dynasty.

From Late Middle Chinese to modern standard Mandarin (not in chronological order):

1) all of the voiced non-nasal / non-liquid stops, affricates and fricatives are devoiced and tonogenesis occurs and so they merge into their voiceless counterparts. For the stops and affricates in Mandarin, whether the consonant ends up tenuis or aspirated depends on the tone: aspirated in 平, otherwise tenuis.

2) the retroflex non-nasal stops merge into the retroflex affricates. So 知 comes to start with the same consonant as 照, while 徹 merges into 穿. They end up as zh-, ch- in Mandarin Pinyin.

3) the retroflex nasal 娘 on the other hand merges with the dental nasal 泥 in the overwhelming majority of cases; they end up as n-.

4) the retroflex rhotic approximant 日 develops along its own track into Mandarin Pinyin r- or occasionally into the syllable "er" after metathesis (which is meant to have happened way before 入声 endings were lost: see this StackExchange answer).

Hence northern Mandarin, including standard Mandarin, mostly preserved the Late Middle Chinese retroflex consonants (only the retroflex nasal merged out of the retroflex group). Central and western Mandarin merged them with the dental / alveolar group... I've yet to find out when this happened. On the other hand, Yuen-Ren Chao in his study of the 鐘祥方言 of central Hubei (so a variety of southwestern Mandarin) in 1939 said all the historically dental sibilants have retroflex values!

In Cantonese, the tonogenesis-inducing devoicing and the merger of the retroflex stops into the affricates occurred. The formerly voiced consonants become aspirated (as in Mandarin) in 平聲 but also in colloquial 上聲 [though I'm not sure what colloquial Cantonese characters come from retroflex consonants... I have found the example of 眝, which is zhù in Mandarin Pinyin with tenuis initial but cyu5 in Cantonese Jyutping with aspirated initial].

It is well known that phonemes recorded by 19th-century missionaries record a separate "ch-" and "ts-", "j-" and "dz-", "sh-" and "s-", and the gradual loss of this distinction is recorded even up to 1950. The Wikipedia page for Cantonese phonology gives a few sources, right up to the Meyer-Wempe dictionary of 1947. Hints of the distinction still exist in the romanisation of such places as 尖沙咀 [sha for the second syllable] and people such as 蔣介石 [shek for the last syllable]. Hence we have a rather well-documented loss of the phoneme.

Note that the correspondence with Mandarin is not one-to-one (色 being the most prominent example: standard Mandarin is alveolar, although preserving the retroflex consonant in its literary pronunication; whilst the Cantonese of Meyer-Wempe gives it as palatal; naturally of course, modern Cantonese is mainly alveolar. 阻 is another with ramifications for Cantonese aspect particles, where Mandarin has lost the retroflex quality of Late Middle Chinese from this character).

Nowadays in most speakers' Cantonese, palatalisation of the alveolar phonemes is allophonic before a subset of vowels, which to my ear differ from region to region (and from overseas communities too).

Whether the merged retroflex/palatal phoneme of Late Middle Chinese was ever actually realised as a retroflex amongst proto-Yue speakers, or if it was actually retroflex merging into palatal in the southern area of the Song empire ... that's a question as yet unanswered. Nonetheless, with the alveolo-palatal/alveolar merger in Cantonese in the 20th century, even Chao documents in the 1970s that the emerging phoneme was pronounced "neither sharply dental nor sharply palatal"; now of course the allophonic variation between straight alveolar and absolutely alveolo-palatal is pretty clear, in Hong Kong media at least.

Retroflex 娘 also merged into dental 泥 - it seems that this is a distinction that was lost very early on from Middle Chinese, as it seems to be near-universal. The palatal nasal-turned-approximant 日 loses its nasality but has come out to be palatal - again, whether this is after going through a retroflex intermediate like Mandarin or whether it has retained its palatal nature through the centuries... see above.

The Wu group has also undergone a merger of the retroflex stops into the affricates, though the voiced consonants retain their breathy voice as well as doing the whole pitch-depressor thing. But interestingly, one of the biggest differences between new and old Suzhou dialect is the loss of the retroflex-dental distinction. Chao completed his survey of the Wu dialects by the end of 1920s, so that again gives us a time frame. The actual nature of this distinction seems a bit curious though: it is written that the 章/昌/船 group (as in, the palatals from Early Middle Chinese) had become retroflex (or retained their retroflex nature); it is not specifically written that the 莊/初/崇 group (the retroflexes from Early Middle Chinese) had become dental by the 1930s, although that seems to be what I'm concluding from it. Hmmmmm...

Also, Wu is famous for keeping its palatal nasal as an alveolo-palatal nasal (in colloquial readings, of course; in literary readings it imported the Mandarin version, and it is now alveolar, although it is debatable whether it was alveolar on import or not)! Nonetheless, the retroflex nasal was subsumed under the alveolar one even before Chao did his study on the old Suzhou dialect.

The Min group is famous for its keeping the retroflex stops separate from its affricates, although they've still all become alveolar in the end. I've not yet looked into when yet.

  • Like 2
Link to comment
Share on other sites

More like most did, and lost their distinction some time after Proto-Mandarin. One thing I wonder though is why Japanese and Korean don't seem to have any differentiation either, and apparently borrow vocabulary from Middle Chinese. I haven't looked into this.

Also, I'm not sure about how much you can say about Min when comparing it to Middle Chinese.

Link to comment
Share on other sites

I'm confused, most did? I don't see how the information in Michaelyus's post points to that conclusion... Perhaps I'm missing something.

Wouldn't it make sense that Japanese and Korean wouldn't have a distinction if they didn't have retroflexes in their phonological inventory?

Link to comment
Share on other sites

  • New Members

Michaelyus, thank you for the highly informative answer (amateur may be, but so was Fermat). What should I read (beside the Wikipedia articles) next for further information on the matter?

陳德聰: Northern Mandarin brought them with, Central and Western dropped it. But regarding the rest of the Middle Chinese descendants I am also confused.

Michaelyus, you do say:

Whether the merged retroflex/palatal phoneme of Late Middle Chinese was ever actually realised as a retroflex amongst proto-Yue speakers, or if it was actually retroflex merging into palatal in the southern area of the Song empire ... that's a question as yet unanswered.

So it is unknown whether the merged retroflex/palatal (with retroflex dominating) was retroflex in proto-Yue or got palatal in the Southern Song before Yue emerged? Or stated more simplified, it is unknown whether retroflex/palatal -> palatal, or retroflex/palatal -> retroflex (and then dropped), in proto-Yue? I.e, did proto-Yue ever get to implement the retroflex?

Here's an attempt at a shorter notation.

Old Chinese -> [i.a. adds retroflex stops/sibilants emerged from dental stops/sibilants] -> Middle Chinese -> [i.a. merges the retroflex and palatal group, retroflex dominates] -> Late Middle Chinese -> [Michaelyus; 1, 2, 3, 4, and division] -| {Northern Mandarin [keeps retroflex], Central and Southern Mandarin [merges retroflex with alveolar and dental]}

But this explains Mandarin only. So is "Did proto-Yue ever get to implement the retroflex?" unanswered? I suppose someone has tried to make a map of the phonetic developments from Middle Chinese to its descendants, or maybe this is not even conceivable?

Link to comment
Share on other sites

Taking the Changsha dialect as an example of Xiang Chinese / Hunanese, the loss of the retroflex consonants is in process, with a marked difference between the old and new Changsha dialects: again, the retroflex stops and affricates of old Changsha have become their alveolar counterparts. But the nature of the retroflex consonants in old Changsha is not one-to-one with those of Late Middle Chinese: see this part of the Chinese Wikipedia article on 长沙话, where it shows that the retroflex consonants of old Changsha correspond to a subset of the Late Middle Chinese retroflexes (specifically, 三等开口 or grade III open, as I mentioned in my post above); the other grades caused the retroflexes to become alveolar or alveolo-palatal. [i don't know what's happened with the sibilants] Time frame? Further on in the article it talks about the 湘音检字 of about 1930 mentioning that 舌 and 昌 were starting to be pronounced with alveolar initials in 俗读 (I hesitate to use 'casual' reading, but I can't use 'colloquial', can I?).

Like Wu, it looks the palatal nasal was and is preserved as an alveolo-palatal in Xiang, but it looks somewhat inconsistent, dependent on the individual villages. Additionally, Xiang did what western Romance did with Latin earlier and turned some of the dental nasals into palatals, hence merging them with the palatal nasal. Furthermore, Xiang borrowed a literary superstratum from Mandarin, which is variously realised (again depending on area) as the voiced retroflex fricative /ʐ/, or as the alveolo-palatal version /ʑ/, or as the alveolar version /z/, or as the unvoiced alveolo-palatal fricative /ɕ/, or (finally) as the voiced alveolar lateral fricative /l/. See the Chinese Wikipedia on 湘语 for that.

Link to comment
Share on other sites

Personally, I'm mainly concerned with Cantonese. The information in that wiki is about the merging of alveolars and alveo-palatals, and doesn't give any insight into retroflexes in Cantonese. So I'm still confused as to how that gives a time frame at all for retroflexes, or even speaks to Cantonese having retroflexes ever at all... Could you clear that up for me?

Link to comment
Share on other sites

Regarding Cantonese, AFAIK, there was/is definitely a time when Middle Chinese (alveolo-)palatals and retroflexes (章昌禪書船, 莊初崇生俟) corresponded with Cantonese alveolo-palatals, and MC alveolars/dentals (精清從心邪) corresponded with Cantonese alveolars/dentals. Since MC palatals and retroflexes eventually merged into retroflex, one could guess that at one point Cantonese (or whatever became Cantonese) had retroflexes in place of the modern alveolo-palatals among conservative speakers. Of course, most speakers now merge them all still into alveolars/dentals.

Link to comment
Share on other sites

Wouldn't it make sense that Japanese and Korean wouldn't have a distinction if they didn't have retroflexes in their phonological inventory?

Kind of. Languages can adopt sounds from foreign languages, especially when importing lots of vocabulary. Also, they don't have to differentiate them using the same sounds. I think Vietnamese does this to some extent.

Link to comment
Share on other sites

Apparently, as found in One Country, Two Systems, Three Languages: Changing Language Use in Hong Kong edited by Wright, Karlgren's Études sur la phonologie chinoise (1915-1926) had a retroflex (or apico-prepalatal) sibilant but alveolo-palatal affricates in his transcription of Cantonese, both phonemically distinctive from the alveolars except before /i/ and /ɨ/ (apparently Cantonese as recorded by Karlgren once had this 'apical' vowel?!). This seems difficult to corroborate though: Jones and Woo's (1912) A Cantonese Phonetic Reader does not make this distinction.

The 1997 edition of Linguistic Models focusing on Studies in Chinese Phonology (edited by Wang & Smith) has an article by Pulleyblank concentrating on Cantonese vowel history. He states that the merger of Early Middle Chinese's palatal and retroflex series brought them "into complementary distribution, since the palatals were already found only in front of high front vowels", but that "the single coronodorsal series... has behaved as retroflex, i.e. [-front]".

Link to comment
Share on other sites

  • 1 year later...

According to several Romanization schemes and Wikipedia, Cantonese had palatals and the alveolar sibilants distinct.

Late Middle Chinese had a shibilant series and an alveolar sibilant series. The shibilant series came from the merging of the earlier retroflex shibilants and the palatal shibilants. Mandarin merged the two series as retroflex shibilants, while Cantonese merged them as (alveolo-)palatal shibilants. By 1940, the palatals were depalatalized in Cantonese, so the shibilants were lost in Cantonese.


Here though, pronouncing the Mandarin zh ch sh series as z c s is simply considered "wrong" by teachers.

Link to comment
Share on other sites

  • 3 weeks later...

Let's discuss the LMC 床 and 禪 initials, because there seems to be considerable confusion.


Historically speaking, the following groups represent the mergers that were believed to have taken place during the transition from EMC to LMC:





For the purpose of this discussion, we'll refer to Bernard Karlgren, 王力, and 李榮 collectively as the 'old camp', while referring to 邵榮芬, 鄭張尚芳, 潘悟雲, Edwin Pulleyblank, and William H. Baxter collectively as the 'new camp'.


The old camp interpreted the voiced alveolo-palatal initial 船 of EMC as an affricate /ʥ/, while the new camp interpreted it as a fricative /ʑ/. The voiced alveolo-palatal initial 禪 of EMC (often labeled as 常 to avoid confusion with the LMC initial of the same name), on the other hand, was interpreted as a fricative /ʑ/ by the old camp, and as an affricate /ʥ/ by the new camp. In other words, each camp maintained the obverse position of the other. While the old camp's stance was based on the fact that the initial 禪 was common to both EMC and LMC–and therefore intuitively presumed to have been carried over intact–the new camp's stance was founded on the belief that the authors of the original rime tables had confused the two initials due to a phonemic merger at the time of compilation, which they justified with newly uncovered evidence that hadn't been available to the old camp at the time of their initial research. Pulleyblank notes that this merger of 床 and 禪 in LMC–with a lingering distinction kept only in orthography for the sake of etymological integrity–created a single voiced-retroflex-sibilant phoneme with two allophones: a fricative before high-front vowels (Division III, 止攝, 深攝, etc), and an affricate elsewhere. Pulleyblank further notes that it was only much later–when the retroflex stops merged with the retroflex sibilants–that this affricate/fricative distinction (between 床 and 禪 respectively) fully 'stabilised', as they had been in free variation in a number of regions beforehand.

  • Like 1
Link to comment
Share on other sites

As far as I know the change is something that happens in the South. I first encountered it in Guangxi and then later in Guangdong. The explanation I was given when I asked one of the locals was that it had to do with them having a hard time with the tongue position necessary to make sounds like shi2 so you wind up with si2 for 10 instead. Which causes a fair amount of confusion for foreigners that aren't as adept at identifying the difference between the tones.


Of course there might well be other reasons, but I've seen the same thing going on with Vietnamese native speakers having a hard time making that sh sound and I suspect that the reason is similar.

Link to comment
Share on other sites

While the old camp's stance was based on the fact that the initial 禪 was common to both EMC and LMC–and therefore intuitively presumed to have been carried over intact–the new camp's stance was founded on the belief that the authors of the original rime tables had confused the two initials due to a phonemic merger at the time of compilation, which they justified with newly uncovered evidence that hadn't been available to the old camp at the time of their initial research.


Do we know what kind of evidence this was? Phonological? Based on a greater variety of 方言? Or was it palaeographic / historical?

Link to comment
Share on other sites

Join the conversation

You can post now and select your username and password later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Click here to reply. Select text to quote.

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


  • Create New...