User:Hmanck/Historical development of the Cantonese language

In here, we will explore the development of the phonemes of the Cantonese language, and retrace it to Middle Chinese, and ultimately Old Chinese.

Sound Changes
Speakers of the younger generations are beginning to experience a phenomenon called "lazy-sounds", where:
 * Initial ŋ is dropped
 * Final ŋ is sometimes merged with final n
 * Initial n pronounced as "l", mixing "l" and "n" initials
 * Delabialization of velar consonants: kʷ and kʷʰ are pronounced as k and kʰ.
 * Syllabic ŋ is undergoing a two-step process to merge fully with syllabic m.

Additionally, before a high vowel, the alveolar "s" is sometimes pronounced as ɕ. This sound is actually an older sound that was merged with s in around the 19th Century. The glide ɥ is actually allophonic with j.

Vowels and Permissible Rimes
(Taken from Cantonese phonology)

Like other Chinese languages and Sino-Tibetan languages, a Cantonese syllable is composed of an initial (typically a consonant or a glottal stop), and a rime. A rime is therefore made up of a vowel nucleus and an optional coda consonant. In Cantonese, m and η are syllabic nasals, but cannot appear with another consonant in the same syllable as a Standard Cantonese word. In total, there are 53 permissible rimes, which include long and short consonants and the codas: m, n, η, p, t and k.

The chart below lists the permissible rimes in Cantonese:
 * Syllabic nasals:
 * ¹Finals, , , and only appear in colloquial pronunciations of characters. They are absent from some analyses and romanization systems.



Based on the chart above, the following pairs of vowels are usually considered to be allophones:
 * -, - ,  - ,  - , and  -.

Although that satisfies the minimal pair requirement, some linguists find it difficult to explain why the coda affects the vowel length. They recognize the following two allophone groups instead:
 * - and  -  -.

In that way, the phoneme set consists of seven long vowels and three short vowels that are in contrast with three of the long vowels, as presented in the following chart:
 * Syllabic nasals:

Historical Chinese phonology regards the stop codas "-p", "-t" and "-k" to be the allophones of the nasal codas "-m", "-n" and "-η" in entering tone position. I will therefore call them "labial", "coronal" and "velar" codas to describe them as related series. Based on pronunciations from older speakers and various dialects such as the Taishan dialect, I believe that the œːŋ and œːk rimes originated from -ioŋ and -iok respectively.

Tones and Allotones
There are 6 phonemic tones, but are usually transcribed as 9. The first six are divided into a yin/yang set (yam/yeung), which corresponds to the original level, rising and departing tones. These are: The three "other tones" developed from a distinction between closed syllables (entering tones). As with the other tones, there was a tone split based on the voicing of the initial consonant. Furthermore, the darker tone category was split on the vowel length. In total there are three "entering tones".

Allotony
Cantonese tones undergo allotony, which may perhaps reflect an older value of the tone. For example, words that are pronounced as level-yang, may have a tone value of 35 (resembling rising-yin) in many informal contexts, such as nicknames. Note: level-yang has a value of 35 in Standard Mandarin and Minnan.

In modern kinship terms, the first syllable is duplicated and carries a low tone (11), while the repeated syllable carries a high or rising tone. (5 or 35).

Early Modern Cantonese
Not many records have been kept to document historical Cantonese, much of our information would come to early British dictionaries from the time that Early Modern Cantonese was spoken, and comparison phonology to derive reconstructions. Analysis can also be taken from dialects that deviate slightly from Standard Cantonese today, such as the Sai Kwan (Xiguan) dialect of Guangzhou. (http://zh.wikipedia.org/wiki/%E8%A5%BF%E9%97%9C%E5%8F%A3%E9%9F%B3)

Early Modern Cantonese should reflect the Standard Cantonese spoken in Guangzhou between perhaps the late 18th Century to the early 20th Century, when the prestige standard was shifted to Hong Kong following the Communist takeover of Mainland China.

Vowels
I believe the biggest change during the Early Modern Cantonese era would be the i rime. Strangely, Modern Cantonese "i" rime exists only in sibilants (fricatives and affricates) and in front of the glide j, and with other initials only in non-native words or onomatopoeia/nursery words. 咪 is generally pronounced as "mɐi3" today, but is pronounced as "mi1" in informal and nursery cases. In fact, not all sibilants and fricatives retain the historical "i" rime, which had evolved into "ei" or "ɐi" by the early 20th Century. (Yau Ma Tei was once written Yaumati then) Cantonese initials today that correspond to Mandarin retroflex affricates (tʂ) and (tʂʰ) and Mandarin plain affricates (ts) and (tsʰ) are the only initials that retain the historical "i" rime. Note the retroflexes derived from Middle Chinese, which were introduced into Cantonese as alveolar-palatals, before they merged into plain coronal affricates in the modern era.

The same could be said for retroflex and plain fricatives, (ʂ) and (s). In Mandarin today, they are pronounced with a reduced vowel such as ɿ or ʅ (si, shi). The only issue however, would be the fricative sibilants in the departing-tone classes and some rising-tone classes - in which case they are pronounced sɐi (逝), sei (四, 死). This issue could be viewed as confusion: the Xiguan dialect retains "si" for all of these pronunciations. Similarly, the Middle Chinese retroflex ʂ was pronounced as an alveolar-palatal, before it was merged in the modern era.

The fact that these initials have an "i" in Modern Cantonese and that the "i" rime in other words have evolved to diphthongs suggests to me that the "i" present today was not in fact an "i". Based on the nature of the Mandarin vowel, it might have been that the Cantonese "i"-like vowel may have been somewhat like ɨ. Meanwhile, the actual "i" vowel in EM Cantonese might have elongated first from /i:/ to a vowel with a glide: /ij/, and a split occur, with the vowel nucleus slightly backed or centralised to /eɪ/ or /ɐɪ/. Here is a chart approximating what had possibly happened:

Consonants
As mentioned, the biggest change I would see in the consonant inventory from Modern Cantonese would be the presence of alveolar-palatal sibilants, which corresponded to the Mandarin retroflexes. So I would edit the consonant inventory like this:

Late Middle Chinese to Early Modern Cantonese
It is very difficult to retrace earlier stages of Cantonese, but there is a consensus that much of it had developed from Late Middle Chinese spoken during the Song Dynasty. The Song court had moved southwards to Hangzhou, and the prestige variant was shifted south, as opposed to the Chang'an variant that had been in standard in the Northern Song and the Tang era.

But what distinguishes LMC from Cantonese?
 * Development of labial allophones for labialised laryngeals
 * Devoicing of initials/Tone-split
 * Labial initials cannot have labial coda rule
 * Lenition of most non-palatalised velars (DIV I, II and maybe IV) to fricatives, before merging with laryngeal fricatives
 * Depalatalisation of velars
 * Development of syllabic nasals from reduced vowels
 * Reduction of vowel clusters
 * Merger of MC Alveolar-Palatal sibilants with Plain Sibilants
 * Vowel shift

Velar Lenition
Both Hakka and Cantonese today show massive quantities of aspirated velar initials (kʰ, 溪）being pronounced as /h/ or its historical allophones, /f/ today. My theory is that there was a lenition of many unpalatalised aspirated velar syllables that had triggered this phenomenon. But, given the fact that it did not affect any syllables that were pronounced with a g (群), which had devoiced and had aspirated in some cases, it should be thought that this had occurred prior to the devoicing and tone-split.

It is uncertain whether the velar lenition was a one-step process or a two-step process. I have always argued that Cantonese and Middle Chinese had /h/ in place of the /x/ phoneme. Thus the lenition, in the two-step process, would have generated a /x/ phoneme on top of the existing /h/ phoneme. The /x/ phoneme would have merged later with the /h/ phoneme. Evidence from Vietnamese supports this position. A one-step process would have assumed that the velar consonant had lenited directly to a /h/ phoneme (which I had argued to be the norm in Southern Middle Chinese), but this would have been very unstable, due to the fact that /kʰ/ and /h/ do not have the same place of articulation.

As this certainly happened before the tone split and devoicing, it had perhaps occurred in the middle of the Tang period, as the tone split was seen in AD 880, attested by the Japanese monk Annen. (See later.) I will call this Early Cantonese, as this is an attestation of a series of diverging splits from the speech in Southern China that will become Modern Cantonese. If all this is true, the consonant inventory of Early Cantonese should have looked like this:

Labialised laryngeals become labial fricatives
After the velar lenition that brought many aspirated velar initials to /x/, there was certainly a period where the phonemes /x/ and /h/ merged. A condition exists afterwards, whereby a laryngeal, when followed by a w (labiovelar glide), becomes a labial fricative. We see this change in Modern Cantonese and in Hakka. For example, 花 is pronounced fa in Cantonese, while hua in Mandarin (with a laryngeal > velar). Similarly, 華 is pronounced wa in Cantonese, while hua in Mandarin. Also 苦 is pronounced foo in Cantonese, while ku in Mandarin (kh > h > f in Cantonese).

The sound change should best be analysed like this: the labial fricatives were in fact allophones of the laryngeals in rounded environments. These allophones were most likely bilabial allophones: β and ɸ, the latter which became f when labiodentalisation of the labial plosives took place. Furthermore, this conjecture supports the fact that Middle Chinese had laryngeals instead of velar fricatives; glottal fricatives tend to undergo this transition more often than their velar counterparts. Notice in Korean, that ɸ is sometimes an allophone for its /h/ sound.

This sound change may have occurred prior to the initial devoicing and tone-split, but was not impacted by it. In Hakka, there is one phoneme /f/ for these labialised laryngeals, while /f/ and /w/ are these reflexes in Modern Cantonese. Compare 華 in Hakka, which is read as fa, and wa which is read in Cantonese. The two are not merged, like they were in Hakka, suggesting that tone split did not affect them.

Devoicing of initials/tone-split
The devoicing of voiced initials began with the tone-split that has created new tones in one, more or all of the tonal categories. Southern languages generally have a more complete tone split, with two or more tones for each Middle Chinese tonal category - Ping, Shang, Qu and Ru (level, rising, departing and entering).

In order to study the tones in Middle Chinese, we have to look at want the tones were described as. A Japanese monk named Annen wrote a piece in AD 880 on Middle Chinese tones: (http://www.chinese-forums.com/files/8726pdf.pdf) The level tone is described as "level" and "low". The rising tone is described as "high" and "level or straight contour". The departing tone was drawn out, and the entering tone syllables were abrupt and had a high and low pitch. There seemed to be confusion of the "heavy allotone" for the rising tone and the departing tone. Modern Cantonese and Mandarin have merged much of the heavy rising tone (yang-shang) with the departing tone. In modern Cantonese, sonorant and liquid initials are the usual words to have heavy rising tone.

I believe the "heavy allotones" were pronounced with breathy voice; I believe this is somewhat true of the yang-ping (heavy level tone) in Modern Cantonese. The voiced plosives of Middle Chinese were devoiced and aspirated: bing -> pʰing. The lowered tone gives the aspiration a more breathy quality.

Based on evidence from Tibeto-Burmese languages, it can be assumed that Old Chinese had true voiced stops and Middle Chinese might have retained them to an extent. However, during the tone split, I believe slack voice arose during this tone split phenomenon.

Labial initials and codas
Between Tang Chinese and Song Chinese, there seems to be a rule to disallow labial initials and codas to be part of the same syllable. 法, supposedly pronounced pjap, was most likely pronounced as fat (as it is in Cantonese; Mandarin "fa"). Korean writes this as "byeop." Similarly, 范, was supposedly pronounced as byam. But was pronounced as van in the Song era (Cantonese and Mandarin "fan"). Compared Korean byeom and Vietnamese pham.

Consonant mergers
By the Song era, many changes in the consonant inventory are apparent:
 * The labial allophones of the laryngeals merged with similar sounds: /β/ with /w/; /ɸ/ with /f/
 * Retroflex stops become retroflex affricates, possibly beginning with the aspirate: /ɖ/, /ʈ/ and /ʈʰ/ with /ɖʐ/, /ʈʂ/ and /ʈʂʰ/ (LMC development, common to most Chinese languages)
 * Original retroflex affricates merge with palatal affricates (indicated by the 36 initials)
 * /ʐ/ and (original affricate) /ɖʐ/ merge into /ʐ/

Rise of the syllabic nasal
The syllabic nasal ŋ, most likely came from the syllable ŋu. This can be reconstructed due to the fact that Mandarin has "wu" for 五、午 and 誤, all of which are pronounced as the syllabic nasal ŋ in Cantonese. The number 5 had been reconstructed as ngaʔ in Old Chinese, which would have yielded ngɔ in Middle Chinese, which had become ngu in Late Middle Chinese. In the course of development of the Cantonese, the vowel nucleus of ngu was reduced and syncoped (note that ng is a velar consonant and u is a back-round vowel), leaving the vocalisation of the consonant element as a vestige of the original vowel.

In the more modern era, the syllabic nasal that is pronounced with a level tone became syllabic /m/; rising and departing tone syllabic ng retained its velar element. However, in the recent era, as part of the "lazy sounds", ng and m have completely merged and the phoneme is pronounced as syllabic /m/.

Simplification of vowel clusters

 * The rime *-ia became *-e before a sibilant and *-ie before a null initial/y-glide. 謝 (tse, Cantonese; xie, Mandarin; dʑia, Shanghainese Wu; ɕia, Hokkien Min) 夜 (ye, Cantonese and Mandarin; ya, Hokkien Min) (也 seems to be the only exception, I have yet to find a reason for this.)
 * The rime *-uan became *-uon (Middle Cantonese?) when preceded by a consonant other than a glottal stop:
 * Simplified to *-un (Modern Cantonese) after labials and velars. 盤 (pun, Cantonese; pan, Mandarin); 歡 (fun, Cantonese; huan, Mandarin --fun is treated as a labial due to the labialisation assimilation with laryngeals-- ); 官 (gun, Cantonese; guan, Mandarin)
 * Simplified to *-yn after coronals and palatal glide: 亂 (luen, Cantonese; luan, Mandarin), 川 (chuen, Cantonese; chuan, Mandarin), 原 (yuen, Cantonese; yuan, Mandarin)


 * The vowel nucleus iɛ, before a coda consonant, becomes reduced to i. Meanwhile the vowel nucleus i, before a coda consonant, is chain-shifted to ɐ:
 * -iɛp -> *-ip, (協; hip, Cantonese; xie, Mandarin) ;*-iɛm -> *-im (鹽; yim, Cantonese; yan in Mandarin)
 * -iɛn -> *-in (歇; hit, Cantonese; qie, Mandarin); -iɛt -> -it (別; bit, Cantonese; bie in Mandarin)

OC Voiceless Sonorants
脫, meaning remove or strip, is pronounced tuet3 (from MC thwat); however there is a colloquialism in Modern Cantonese meaning "remove" and it is pronounced lɐt55 (甩). It is rather unusual for words with sonorant initials to have higher tones (yin tones), as they usually have yang tones. This suggest that this word does not exactly come from Middle Chinese, but possibly as a loan word or from an older stratum of Old Chinese. According to current Old Chinese reconstructions, there is a series of voiceless sonorants: lh, mh, nh, ngh, rh, wh. The phonologists Sagart and Baxter have analysed that:
 * lh > th
 * lhy > sy
 * mh > h or hw
 * nh > th
 * nhy > sy
 * nhr > thr
 * ngh > h

The perception that lɐt1 (甩) is a reflex of an Old Chinese pronunciation that gave rise to (脫) tuet3 cannot be overlooked. Through analysis, it has shown that the initial in OC for 脫 is lh, or a voiceless (aspirated) lateral. Voiceless laterals usually correspond to the alveolar voiceless aspirate plosive, th, unless when there is palatalisation. However, it could be that lɐt55 (甩) emerged as a dialectal variation that was never affected by the MC pronunciation. The voiceless aspiration of the lateral is certainly lost, but a vestige of it lies in the pronunciation of lɐt55 (甩) with a high tone (yin-ru).

Iambic Syllables
Iambic syllables or sesquisyllable that have been reconstructed from Old Chinese have retained a disyllabic form, especially in the vernacular names of some animals and body parts: kak33-lɐk55-tɐi35; gloss: armpit, compare with literary (胳, kak33) 馬騮 (ma23-lɐu55); gloss: monkey, compare with literary (猴, hɐu1)

The gloss for armpit has two variations, the literary (which is less-common than the spoken vernacular) stems directly from Middle Chinese. The vernacular seems to have stemmed from Old Chinese. The dichotomy between the literary and the vernacular is common in Southern languages like Wu and Min. The vernacular variation features three syllables, of which the first two bare resemblance to the reconstructed OC pronunciation for 胳 as k-lak. Notice how the phonetic 各 is present and is seen in words such as 落 and 格, which features laterals and velars, respectively. The reconstructed form sees this phonetic element as a fused syllable of a velar prefix with a lateral root. In some cases, the velar disappeared; in others, the lateral vanished. However, in this variant of armpit, Cantonese had retained both the velar and the lateral, separating it effectively into two syllables. Unlike MC syllables, where MC is a truly monosyllabic language, these Cantonese syllables reflect OC syllables, which may or may not represent an entire word. The tɐi35 corresponds with the normal pronunciation of 底, meaning under, and should be unrelated to the OC pronunciation of 胳, which is being constructed as k-lak.

The vernacular for 馬騮 was originally written as 馬留, the second word is typically pronounced with a yang-tone because of a sonorant, is pronounced as yin-ping. Again, we point to either a voiceless sonorant present, or the presence of iambic syllables. According to the Chinese Wikipedia article of monkey (http://zh.wikipedia.org/wiki/%E7%8C%B4), there is a variant in Standard Chinese for monkey called 猱 (nau11, or Mandarin nao2 in yang-ping). Chinese dictionaries describe it as: "animal name, simian species".《漢語大辭典》：「獸名，猿類，身體便捷，善攀援」With the regularity of yang tones assigned to sonorant initial words, we understand that this is a Middle Chinese-stemmed word. Furthermore, in the word 猱, we see the phonetic 矛, which is pronounced (mau1, mao2). Thus we point to a labial nasal initial for this word. It is interesting, however, how m-l- typically corresponds with m- but not n-. I treat this as an irregularity, and a dialectal feature that would drop m- and mutate l to n- to compensate for the lost nasal. If this is true, 猱 and 馬騮 are indeed related to each other, and thus would reflect an OC iambic syllable. Notice how the Zhuang language features the word "maxlaeuz" for monkey. This is undoubtedly related to the Cantonese pronunciation and, perhaps, the OC reading. It is unsure, whether or not if this was a Zhuang borrowing, or a borrowing from OC into Zhuang. While 猱 had fell out of usage, modern Chinese varieties have reflexes of OC iambic clusters for animals, including "ant" (螞蟻), that reflect an earlier m- prefix or initial.