As an inveterate lover of all things bereft of past, future, and use, I was thrilled to discover the “New Phonetic Character for writing Shanghai dialect” invented by Protestant missionaries in the 1850s; a few books were published in it, heroically catalogued by David Helliwell in this 2012 Serica post, and it was then consigned to the dustbin of history, so utterly forgotten that an inscription in it was posted at douban and no one could identify it. I learned about it from Victor Mair’s Log post, where in the first comment Jichang Lulu identifies the writing shown; I also highly recommend P’i-kou’s comment on the Helliwell post, going into fascinating detail about what the phonetic representation tells us about the history of Shanghainese.
Fascinating. I’d long ago heard that Shanghai (or Wu region) tones were very different, or even that those languages were nontonal. I can now see how merging of rising and falling tones in citation forms might create that impression, but even more so the sandhi effects that alter or obliterate all but the tone of the initial syllable in a phrase. Writing New Phonetic Characters highlights these differences.
There used to be a website that explained in quite some detail that (specifically) Shanghainese has pitch accent rather than tone: 1) all syllables with voiced initials have low pitch; 2) otherwise, the stressed syllable of a word (and the words are longer than Pinyin implies for Mandarin) has middle or high tone, and this distinction is phonemic; 3) the pitches of all other syllables are fully predictable; 4) the traditional analyses in terms of five tones (usual for Wu) and ludicrously complex tone sandhi should be blamed on Chinese tradition, especially the exclusive recognition of characters (syllables) instead of words. Alas, I can’t find it anymore.
Pity, it sounds really interesting.
The key sentence from LL is here:
Decoded, it comes out like this:
Middle Chinese (MC) had four tones, three for syllables not ending in a stop and one for syllables ending in a stop. The tones then split based on whether the initial consonant was voiceless or voiced. Mandarin tone 1 is derived from the MC first or ping tone with a voiceless initial. Mandarin tone 2 is also derived from the MC first tone, but with a voiced initial. Mandarin tone 3 is derived from the MC second or shang tone, and Mandarin tone 4 from the MC third or qu tone. (Some syllables changed tones in the process.) All syllables ending in a stop, the fourth or ru tone, lost the stop, as you can see in the word ru itself. They were randomly distributed (due to dialect mixture in Peking) among the surviving tones.
In Shanghainese, the story is different. Voiced initial consonants remained voiced, and all final consonants became glottal stop rather than disappearing. There is no longer any tonal distinction in syllables from the first, second, and third tones with voiced initials. However, there is still a distinction between MC first tone voiceless-initial syllables and and their second and third tone counterparts, and this is the only place where tone is unpredictable — so there are basically two phonemic tones. Because of the tone sandhi, only the first syllable in a phonological word (which can be quite long) shows this tone distinction, so in effect the accent has become a pitch accent entirely. The loss of redundancy has been made up for by the proliferation of new vowel distinctions, which also compensates for the loss of diphthongs and triphthongs.
Nice summary, John. About Mandarin, I feel like adding that the rumours of rù tone’s demise are slightly exaggerated. Only the unvoiced-initial rù undergoes unpredictable changes. Sonorants became 4th tone (merging with qù), as in the word rù itself. Other voiced initials became 2nd tone, merging with voiced-initial píng.
Non-sonorant voiced-initial shǎng consistently became 4th tone. This includes the normal pronunciation of 上 shàng, but that character developed a special “spelling pronunciation” shǎng that refers only to the traditional tone category, reflecting the predominance of 3rd tone reflexes.
Lest there be any perplexity, all non-sonorants in Mandarin (and most Chinese languages) were devoiced, so the conditions for the tone splits are no longer predictable. There’s an interesting distributional effect because voiced stops became voiceless aspirates if they were originally píng tone, but they became tenuis otherwise. Thus, while 2nd tone derives exclusively from voiced initials, the reflexes of píng and rù remain distinct because the former are aspirated. The random distribution of voiceless-initial rù would introduce a wildcard in any event.
Speaking of things bereft of past, future, and use, I really like the idea I had recently of writing Chinese in the Tibetan alphabet using Middle Chinese phonology. Middle Chinese because, by and large, modern dialects all descend from it, and therefore it’s by and large neutral between them all. The script could be read phonetically the same way one reads Irish or Tibetan: by knowing a few rules, the confusing mass of letters voila becomes a phonetic representation of the spoken language. Tibetan script because medieval Tibetan and medieval Chinese happen to have very similar phoneme inventories, so the same alphabet suits both pretty well. Pakpa script would work, too, naturally.
I like it! I suggest you publish a book in it, providing work for future bibliographers.
Or failing that, at least a detailed explication on a blog.
Thanks for the additional details.
“Phonetic tones” in my message above is a grievous blunder for “phonemic tones”. Pkease fix.
Aren’t there some Mongol-period Chinese inscriptions in phakpa script? I think Chris Beckwith told me this a long time ago.
Pkease fix.
Sone.
Chinese in ‘Phags-pa script.
Sure, there are some. It would be weird if there were none, since Pakpa was originally intended to replace Chinese characters. Not sure roughly how many surviving documents there are. Not a lot.
For whatever reason, linguists don’t seem to put too much stock in those documents when reconstructing historical Chinese phonology. I guess because Middle Chinese is so well-documented, they don’t find that the Pakpa documents add much that isn’t already known. Yuan-era phonology probably tends to be less an area of interest than other periods, such as Tang and earlier.
The standard reference on ‘Phags-pa Chinese is Coblin’s book. The sound system is closer to Ming Guanhua than to MC.
The Shanghainese citation (monosyllable) tones are as explained by John Cowan. There are five at the phonetic level. If the initial is ‘voiced’ or there is a final glottal stop (< MC -p, -k or -t), tone is predictable; if not, one live tonal contrast remains. This five-tone analysis isn’t “common for Wu”; other forms of Wu have different tone inventories.
The Wikipedia page on Shanghainese also explains the tone sandhi. This site describes a more ‘accentual’ way of looking at it, but at the cost of ignoring some distinctions. I’m not sure I’ve seen an explicit description of a pitch accent system for Shanghai (as opposed to a statement that such a description is possible). The accentual system I am aware of is for (New) Chongming. Chongming is the (vast) island just north of Shanghai. It’s famous, among other things, for supplying Shanghai with its rice wine， taxi drivers (崇明的哥), and nearest pitch-accent system. A Chongming dictionary I haven’t seen claims 28k entries.
“Ludicrous” complexity is surely in the eyes of the complexity theorist, but at least in a Chinese context Shanghai tone sandhi is rather simple. A mnemonic is: tone spread plus downdrift. A polysyllabic word takes the tone of the initial syllable and ‘stretches’ it along the whole word. So the melody of the full word is just the tone of the initial syllable played at a slower speed; except that, if the melody ends in a high register, the last few syllables will gradually fall in tone. Wicky has examples.
The five-tone system is not just a hangover from the Chinese tradition: both initial voicing and final stops do affect the tone of the syllable (in Wu and elsewhere). At different levels of analysis, depending on the language, you can obviously reduce the number of phonemic tones, and on occasion come up with e.g. a pitch-accent description. In the case of Shanghainese, that shouldn’t be done before a more careful analysis of what you mean by a ‘voiced’ initial. Shanghainese ‘voiced’ stops are, at most, slack voiced, and, especially for younger speakers, can be just unvoiced. In my experience, the crucial role in distinguishing e.g. b- from p- is played by tone, not by voicing. So even the standard assertion that Shanghai has only one live tonal contrast might have to be revised.
In fact, ‘Phags-pa Chinese, Ming Guanhua and Zhongyuan Yinyun (ancestor of MSM) belong to three different dialects already distinct by ca 1000AD.
1) Reflex of Late MC *ʈi, *tʃi and *tʂi (*-i from Early MC 脂 *-i, 支 *-ie, 之 *-iɨ).
‘Phags-pa: *ʈi, *tʃi > tʃi; *tʂi > tʂz.
Ming Guanhua: *ʈi, *tʃi > tʃi; *tʂi > (*tʂz >) tsz.
Zhongyuan Yinyun: *ʈi > tʃi; *tʃi, *tʂi > tʂz.
2) Reflex of Late MC rhymes with -k (*æ from Early MC *rV; *ɔk merged from Early MC *-ak and *-rawk).
‘Phags-pa=Zhongyuan Yinyun: *-æk > -ai; *-ək > -əi; *-ɔk > -au (Pp) -ɔu (ZY)
Ming Guanhua: *-æk, *-ək > -əʔ; *-ɔk > -ɔʔ.
Except Min. Min is not descended from Middle Chinese; it preserves (at least) an Old Chinese consonant distinction that MC had already lost.
Anyway, the neutrality wouldn’t extend to vocabulary or grammar, if that’s what you mean.
Ah, that may be an updated version of what I was looking for! The transcription is the same.
Indeed. I may have confused it with a Min example.
If the contrast between the slack-voiced (“voiced”) and the IIRC stiff-voiced (“voiceless”) plosives is disappearing, you obviously end up with a phonemic tone distinction, as happened in pretty much all of Chinese except Wu at the end of Middle Chinese.
By z, do you mean the modern [ɯ]?