« Of wolves and wloves. | Main | Why do we spell anyway? »




[這個好] phoenix,

you might be interested in the diploma thesis ( http://wolf.ontonym.net/thesis/ ) i wrote a couple of years ago. it is too long and a horrible read (in german), so brace urself. anyway i tried to show that

(i) the syllable finals in the modern chinese language (guoyu/putonghua) exhibit a surprising number of symmetries, which can be depicted graphically in a number of ways;

(ii) it does make sense to analyse this language in a way that the nucleus in compound syllables is always either ‘a’ or ‘œ’ (or a/⊕ or 陰/陽 or ○/● or whatever symbols you prefer);

(iii) in phonology, any definite existential claim and in fact any epistemological statement that goes beyond ‘model A makes object X look like described/depicted below’ is highly dubious.

i was deeply influenced by Y. R. Chao’s famous article on the non-uniqueness of phonemic descriptions.* chao's perspective makes it very hard to go on believing in statements like ‘i proved language L *has*/*possesses* 4 vowels’; you’d feel much more comfortable when you rephrase that as ‘it is interesting to see how nicely everything falls in place when we put all of [eəɛɤɔ] under the single heading of /œ/’.

other than that, i think both pinyin and zhuyinfuhao have their distinctive merits and demerits. i don’t think the makers of pinyin intended the ‘o’ in ‘long, xiong’ to indicate these syllables have a phoneme ‘o’ that is equal to the same letter in, say, ‘gou, guo’. they just wanted to keep westerners from reading ‘lung’ as [lʌŋ]. in other places, those who were not afraid of westerners reading ‘xing’ as [ksiŋ] won over. thus, pinyin, though being an ‘artificial’ orthography, does have its share of historical baggage and illogical points (moreso since it wasn’t created out of thin air, nor meant to be theoretically supreme, but just working for 20th century china).

btw, not only do the chinese syllable finals lend themselves to displays in multiple variations of highly symmetric arrangements, the initials – consonants – do so in a similar way. i never published anything on that subject, but it is an extremely interesting line of inquiry.

FTPlus: i was intrigued by how well the famous vowel triangle, that was first published in the 1730s, lends itself to describe vowel patterns in chinese. of the many papers i read on the subject, those who came up with the most awkward and unconvincing solutions were generally those who did not have a working knowledge of the language, or were deeply committed to a very dogmatic hue of generative phonology, or both.

i think when someone contemplates latin, german and french vowels and then comes up with a very simple diagram that permits to depict the centuries-old chinese descriptions in terms of kaikou-hu, qichi-hu, hekou-hu, cuokou-hu and hongyin vs. xiyin (see ch5, pp57, 60 of my thesis) as simple straight lines, that one has shown that beyond my epistemological concern uttered above, there may indeed be some reality in some descriptions. – sorry for the longish post everyone.


Yuen-Ren Chao (1934) The non-uniqueness of phonemic solutions of
phonetic systems. Bulletin of the
Institute of History and Philology, Academia Sinica Vol. IV,
Part 4, 363-397. Reprinted in E. P. Hamp, M. Joos, F. W. Householder
and R. Austerlitz, eds. Readings in
Linguistics I


ah, and then something. please don’t analyze nü as nyay. it severly hurts, and it is also highly unlikely because of what i dubbed the dimsum rule (see http://wolf.ontonym.net/thesis/06-chapter6/index/pamhc-part1-p077-ch06.html ). it basically states that if we analyze ‘ü’ as some sort of compounded ‘i,u’ and assume that there can at most be one ‘i’ and one ‘u’ in any given syllable, we have both explained why there are no syllables like *guou, *qiai in chinese (although there is the marginal case of yai in taiwan guoyu, which has vanished in mainland putonghua), *and* why there is no such thing as *nüei, *nüou in mandarin – you simply run out of material when you try to form such a syllable, as the ü already needs all the ‘i’s and ‘u’s that you can spend.

secondly, pls don’t say things like ‘x IS an allophone of h’, coz that hurts, too. it hurts for epistemological reasons, but also because anyone who has so far come up with a statement to this effect has failed to convince me: why is it that if ‘x’ is an allophone of ‘h’, it is not that ‘h’ is an allophone of ‘x’? and why is ‘x’ not an allophone of ‘s’? or ‘sh’? that would make sense, too!

the least one can do, i think, is to state that ‘x’, ‘h’, ‘s’ are in a conspicuous complementary distribution, so that an entity of whatever status that comprises all three of them is an option. but that entity is not ‘really’ ‘h’, or ‘x’, or ‘s’, and then transmogrifies into ‘x’ under certain conditions. (in fact i prefer to think that f-h-S form one triplet of (partly
complementary, partly oppositional) fricatives, where S, the sibilants,
form themselves another similar triplet s-x-sh: you have fu—hu, fa—ha, fuo—he and su—shu, sa—sha, se—she with xi, xü serving as the third point for both triplets. [observe how neat it looks when analyzed with a single mid vowel here: fa—fœ, ha—hœ; fœ—hœ, sœ—shœ]).

let’s not forget that chinese, unlike indoeuropean languages, does not have – with the notable exception of erhua/rhotization – *any* process that systematically affects the sounds of a syllable. in german you say ‘Koch’, ‘Köche’, where two forms of the same word display two variants [xç] of what may be the same underlying thing. but in chinese there is no process at all to systematically link any two syllables. there are things like 行 read as both hang and xing, but that’s isolated cases without any morphosemantic rules. 
ok, sorry to lecture that much.


Your right that these variants can't be proven. It's true though that the complete j q x lines exactly fill the gaps of g k h. But I see your point on how they could somehow also combine with other phonemes now. But isn't this just a matter if interpretation. And historically g,k,h and j,q,x are far more likely to have the same origin (and they do) than the other ones.

But I guess you are right to say that, synchronically speaking, it's wrong to stat that j, q, x are allophones of g,k,h.

I am aware of how much [nü] = /nyay/ hurt. It hurt me too ;) But I find it hard to live peacefully with a phoneme /ü/ when it's so marginally attested. It's really only seen without an onset, with n as onset and j,q,x as onset. I'm not saying that this is impossible It could very well be that the other letters merged ü and i or maybe ü and u. But it leaves an odd gap in the system, so I felt free to find the ü a suspicious 'phoneme', and I still do. But right now I'm of the opinion it is probably better to analyse the ü as a thinly distributed phoneme rather than some bizarre allophone of /-yay/.

Thank you for showing interest in my blog, and thanks for confirming my 2 vowel theory, which I found so surprising that I had trouble believing it.


although historical development is interesting, its relevance for synchronic phonology, i’m afraid, has to be proven in each case. as for mandarin, my guess is that when you compare it with other chinese dialects past and present, there are overarching commonalities as well as distinct phenomena. for example, consonantal clusters as in ‘strengths’ are ‘unthinkable’ in mandarin, but some dialects do have them (and some people believe archaic forms of chinese had them too and trace words like ‘kulong’ back to ‘klung’).

going by historical systems has at least two pitfalls: one, how do we know what those phonological systems of yesteryear looked like? it is difficult enough to get any two phonologists agree on any given detail in any given *modern* language! if the situation is not *much* worse for older variants of mandarin, it is definitely solely the result of there being relatively few widely read westerners who publish on that subject, and the immense difficulties that arise in the study of historical phonological texts (a subject with a tradition of around two thousand years, not without reason also known as 苦學. believe me astrology must count as science in comparison to the interpretation of some old phonological literature). not only do we hardly ‘know’ what system-behind-the-surface present-day mandarin ‘has’, much less can we tell even what *actual* speech sounds were used hundreds of years ago. how can we build a sound argumentation on this uncharted and shifting ground?

next, if we insist on building our synchronic theory on diachronic reasoning, are we not already partial? such an approach clearly favors a conservative approach! your statement about the historical origins of ‘jqx’ in northern chinese are, as far as i can recall, considered correct these days. but if we insist that because ‘g>j’ occurred some time ago, it must be true now that ‘j’ is ‘but an allophone of what is really /g/’, we have already put our name under a contract that essentially states that ‘change is not’. the medium becomes the message, and our theory, instead of guiding and enlightening us, becomes the prompter of our words.

for it may well be that it is **precisely this innovation** (and i believe, the entire two-vowel-theory and the multiple triplets i outlined above, together with many other features of the mandarin sound system that point in this direction) that characterizes mandarin: that it has become a language with few ‘atomar sounds’/‘phonotypes’ (s. below) and a scarcity of syllables, but also a highly innerconnected system of sounds, a system that closely ties syllable onsets (‘consonants’) and syllable finals (‘vowels’) in multiply symmetrical ways that are probably hard to find in other languages (and maybe even rare among chinese languages? another potential line of investigation).

don’t get me wrong: it is interesting that ‘g>j’ etc. occurred, and i do think it makes the interpretation of a phoneme /g/ with phonotypes [g, j] easier to swallow. but this shift happened within the context of an entire net of changes that resulted in today’s sound system. to say that ‘j’ is ‘not really on a par with’ ‘g’ entails the peril of missing sight for this system in its entirety, a system that is as a *holon* characterized by a fair number of consonant/vowel co-occurrence constraints.

one very important aspect that should not go unmentioned when talking about mandarin phonology is the uncertainties that surround any attempts to describe the segmental setup of chinese syllables. the main reason for this is simple (see above): where in indoeuropean languages you have processes like ‘work’ -> ‘worker’ and  ‘groß’ -> ‘größer’ that help you single out segments by re-syllabification and morphophonemic soundchange, chinese has no such processes at all (except erhua).

even with the most written-about language of this planet – english – there is no universal agreement to be seen on whether a word like ‘low’ has one long vowel, or a diphthong, or maybe two vowels. it is just not clear. maybe our methods are not advanced enough, or perhaps this ambiguity is built into human language. we don’t know that.

under these circumstances it is hard to firmly believe that mandarin ‘gou’ ‘really has two vowels’. perhaps it is merely the ‘postlabialized’ or ‘lax’ variant of ‘ge’?

or, to take a better example: if we say that ‘de’ unmistakably has a single vowel, where ‘duo’ has two, then what about ‘fo’, ‘bo’? observe that the authors of pinyin opted to write both with a single ‘o’, something avoided in most other places in that orthography.

my guess is that syllables like ‘bo, po, mo, fo’ cover *two* squares instead of one in the phonological table (why are those square most of the time? suggested reading: ann harleman stewart, graphic representation of models in linguistic theory, isbn 0-253-32624-9). an entity covering more than a single position in a grid that is used to define an orthography has more than one systematically correct spelling: which is why all of ‘fe’, ‘fo’, ‘fuo’, even ‘fue’ make sense.

agreement on this point: that it is quite unclear whether ‘fo’ has one or two vowels (phonotypically speaking – i.e. like we’d do in a broad phonetic transcription), means we can neither go on building our phonological analysis on the ground that ‘fo’ has one vowel, neither can we go on building it under the assumption that ‘fo’ has two vowels. we can’t even go on pretending ‘fo’ has *either* one or two vowels. all these avenues are no more once the ambiguity of ‘fo’ is accepted; we must somehow carry that phonotypical ambiguity into the phonology. again, alas, theory threatens to dictate the results of our endeavors. but i am confident the pits i am prone to fall in when going this way are less confining than the other ones.

to show you this is not just hairsplitting business, one more consideration: phonologists have written many pamphlets on whether a given linguistic situation involves a single or two consonants, and have gone to lengths to show e.g. that german ‘pf’ is really one sound, but ‘tr’ is really two (just invented that example). now, few writers consider the possibility that pinyin ‘j, z, zh, q, c, ch’ should count as two sounds each. that hypothesis is likely to be dismissed as ‘improbable for a sinitic language’. ok. but let me cast just so much doubt on this stance by pointing out that these six sounds have *both* a twofold lenis/fortis opposition as d/t have, *and* a threefold palatal/apical/retroflex opposition as x/s/sh have. striking, eh? makes it more likely mandarin does ‘have’ consonantal cluster onsets: ‘dX, dS, dSH, tX, tS, tSH’, right?

ps. ever thought about systematic vs accidental gaps to solve that ‘scarcity of ü’ problem?


now, few writers consider the possibility that pinyin ‘j, z, zh, q, c,
ch’ should count as two sounds each. that hypothesis is likely to be
dismissed as ‘improbable for a sinitic language’. ok. but let me cast
just so much doubt on this stance by pointing out that these six sounds
have *both* a twofold lenis/fortis opposition as d/t have, *and* a
threefold palatal/apical/retroflex opposition as x/s/sh have. striking,
eh? makes it more likely mandarin does ‘have’ consonantal cluster
onsets: ‘dX, dS, dSH, tX, tS, tSH’, right?

That is interesting indeed!

I appreciate your keen interest and elaborate explanation a lot. I have only had one class of phonology for my first year of Indo-European studies, so you are definitely helping me gain some more expertise in the subject ;).

As for the systematic vs. accidental gaps.

Systematic gaps would mean that there is something inherent to the ü which would prohibit any other consonants to mix with  /ü/, while an  accidental gap just means that the language never got "round" to making  words to fill those gaps right?

I can imagine the gaps of the nü- words to be accidental. While it seems that the labials is a systematic gap, since labial + labial could collide, and would then be avoided.

But the dental stops and the palatal/apical/retroflex fricatives and affricates not having any ü glide combinations is odd.

There's not much in those consonants which would prohibit them from combining with those any more than the would combine with the i and u glide. But they are so systematically avoided, it feels strange to interpret them as purely accidental.

Glen Gordon

[ciò è buono]

I agree with LoveEncounterFlow, this theory gives me great stabbing pains. I've developped a persistent twitch as a result of the brain trauma.  ;)

In all sincerity, I never fully understood why someone would want to break down a language like this to a level completely beyond the human experience unless perhaps they were programming a computer for speech analysis. It doesn't seem good for human insight. Normal human beings just don't think about a language this way and I guess I'm too concerned with how human beings and language connect (such as the connections between language and culture) to ponder on these trivialities.

Once, someone (I won't name names) tried to convince me that since it was determined that Sanskrit has only one underlying vowel /a/, then it could be said that Indo-European itself can be analysed in this way. The person in question seemed quite frustratingly adamant on abusing this analysis by imposing it onto the diachrony of Proto-IE, all in order to reduce *e and *o to a single vowel for some reason. Why do we need to reduce everything to a single vowel or consonant? What's wrong with some complexity? Irritating. I had issues with it because of my own theories on Pre-IE that assume that older stages of this protolanguage are... um... NATURAL AND HUMAN!!! :P It may very well be brilliantly mathematical in design and it may be very enticing to do this for autistic fun, but one has to be sternly warned that it does not help to understand the historical linguistics of a given language, whether it be Sanskrit, Indo-European or Chinese.

Whatever the state of modern Mandarin phonemics (assuming you even have a standard dialect of Mandarin in mind when you're analysing the language because, for example, in the busy streets of Beijing, a vowel+nasal sequence often becomes a nasal vowel) this just wasn't the case in the past. For example, Mandarin liu "six" (六) is known to come from *tʀuk based on comparison with other ST languages (Cantonese lok, Classical Tibetan drug), so while you might analyse liu as CYV, it in no way helps you understand where Mandarin came from. In some cases, your analyses will clash horrendously with their known origins. For example, most Chinese speakers will recognize that Mandarin  "woman" (女) is related to Cantonese nöy, yet your analysis of *nyay is completely divorced from that higher plane of reality where a labial component is clearly present crosslinguistically.


For example, most Chinese speakers will recognize that Mandarin "woman" (女) is related to Cantonese nöy, yet your analysis of *nyay is completely divorced from that higher plane of reality where a labial component is clearly present crosslinguistically.

Let me stress I agree that this is an inelegant and indeed, unprovable theory, which also seems to make less sense. I just tried to find the most economic positioning of letters in a table, not regardless, but to a lesser extend regarding the phonetic possibilities. I by now completely disagree with /nyay/ :P

I myself am a 'PIE has one vowel' man :P, it's a typical Leiden university thing. We love that theory, and try to explain away the *o through analogy and such.

I personally think taking out the *o like that isn't very feasible. If the system of the indo-europeans really was so 'beatiful', there would be absolutely no reason for the indo-europeans to start throwing about their vowels through analogy.

If the Indo-Europeans did that, you'd expect it to happen in other languages as well where roots heavily depend on the vocalisation. And I dare to say that exactly the vocalisation is the only part of, say, Arabic, which is hardly influenced by change. So why the hell would Indo-European do it? :D

Besides that I think present *e/zero and perfect *o/zero system is hard to explain if you say that it used to be one vowel.

Assuming that Indo-European was not a language which invented itself on its own but came from another proto-language, it seems highly unlikely that the e and o weren't phonemic yet in the earliest stage of Indo-European.

Wow, I managed to write a reply on Indo-European on my blog entry on Chinese, haha.

Just to get back though, it's pretty silly throughout my whole class of phonology, I never really asked myself 'why' we were doing this. I mean, sometimes it'll be useful to create an orthography for some unwritten African language, but it seems hard to find any other goal for analysing a language's phonology.

Verify your Comment

Previewing your Comment

This is only a preview. Your comment has not yet been posted.

Your comment could not be posted. Error type:
Your comment has been posted. Post another comment

The letters and numbers you entered did not match the image. Please try again.

As a final step before posting your comment, enter the letters and numbers you see in the image below. This prevents automated programs from posting comments.

Having trouble reading this image? View an alternate.


Post a comment

Your Information

(Name is required. Email address will not be displayed with the comment.)