Whither the Velar?
« previous post | next post »
A few days ago, Cyndy Ning sent me this Website for learning pinyin pronunciation. It has both female and male voices which you can activate by clicking on nánshēng 男声 and nüsheng 女声 just above the initials D, E, and F at the top of the table. I also found similar tables here and here.
This is a neat tool, BUT, in playing around with it, I discovered that nearly all of the 4th tone -ANG syllables in the system come out sounding like -AN. A similar phenomenon holds true for all other 4th tone syllables ending in -NG; that includes -ENG, -IANG, -ING, -IONG, and -ONG, -UANG. This is especially the case with the male voice, where I have to strain very hard to hear even a semblance of a [ŋ] at the end, and sometimes I can't hear it at all. Mind you, this is only on the 4th tone! I can hear the final [ŋ] well enough on all of the other tones spoken by the male voice, and I can even hear it fairly well for 4th tone syllables when listening to the female voice.
I asked many Chinese-speaking friends their opinion about the phantom velar at the end of 4th tone syllables spoken by the male voice. About half of them said it didn't bother them and they thought the male voice was adequate as a model. Some colleagues, however, were more critical. For example, here is what Liwei Jiao had to say:
When I was in China, I was Guo2jia1ji2 Pu3tong1hua4 Shui3ping2 Ce4shi4yuan2 (Tester of Proficiency of Standard Chinese, National Level). According to my training and experience, "bang4, hang4, lang4, qiang4, tang4, wang4, xiang4" in your link could be judged as "wrong", while "ang4, jiang4, kang4, nang4, pang4, rang4, zang4" could be judged as "defective". Basically your judgement is correct.
In truth, there seems to be a range of realizations of 4th tone final -NG according to idiolects, sociolects, and topolects. For example, the final velar of 4th tone syllables ending in -NG is clearer on this site with a female voice.
I wonder, though, whether there may not be some physiological-cum-phonological factors at work that lead to the vanishing velar at the end of 4th tone syllables. In particular, the tonal contour is such that it finishes at the very bottom of the speaker's vocal register, and this may make it difficult to clearly and fully enunciate the velar at the end. Bear in mind also that the ending point of the female voice is at a much higher pitch than that of the male voice. All of these factors may be significant in the phonological and acoustic realization of the spoken syllable.
I should also note, along with several of my colleagues, that — particularly with the male voice — the vowel quality changes as the final [ŋ] is reduced.
Moreover, it seems that velars have a tendency to disappear in many other environments beside Mandarin 4th tone final -NG. Here at Language Log, we have recently been discussing G-dropping. And last summer there was an intense debate over whether there really was a velar hidden in the middle of "Uyghur" (here and here). And that segued into an analysis of the missing medial velar in some renderings of yoghurt.
As a sort of addendum to the list of English "pronouncers" for Uyghur that were mentioned in those posts from last summer, I may add that, in the University of Washington announcement for an intensive Uyghur summer program, the suggested pronunciation is ooey-GHUR.
Incidentally, some people seem to think that the medial consonant of "Uyghur" is a voiced uvular fricative [ʁ], while others consider it to be a voiced velar fricative [ɣ]. You can listen to the two sounds on this page (by clicking on the IPA symbol).
You can hear the word "Uyghur" pronounced by a native speaker here. Most educated Uyghurs I know say that their ethnonym should be pronounced with a voiced velar fricative in the middle, but some of my professional phonetician colleagues say that it sounds like a voiced uvular fricative to their ear. And then there are those who swear that there is neither a velar nor an uvular to be heard. How odd that a sound so muscular and substantial as a -GH- has a propensity to vanish into thin air!
A tip of the hat to Jiahong Yuan and John Wells.
Bee said,
March 10, 2010 @ 2:54 pm
Not sure whether someone else has already thought of this, but I believe that it is a characteristic of some Southern Chinese dialects/accents that they have very little distinction between -n and -ng, or have a hard time distinguishing between them. They also have trouble with the z/zh distinction.
xyzzyva said,
March 10, 2010 @ 4:26 pm
My Chinese study barely progressed beyond a single semester at Confucius Institute, but I could swear I've seen somewhere an analysis of the -ng finals as vowel nasalization instead of a velar consonant. Is this an area of instability in Mandarin?
Áine ní Dhonnchadha said,
March 10, 2010 @ 4:59 pm
Xyzzyva, as a former student at Beijing University, I found rather that the area of instability was -n finals, which were often nasalised, as opposed to -ng finals, which were more clearly spoken.
As for Uyghur… I say it the way I learned, which was in Beijing. I had never seen the word before I went there, so I don't know how to say it other than in Uyghur or Mandarin ("Weiwenr").
There are a lot of Uyghurs in Beijing, by the way; we used to go daily to Weiwenrcunr (which we called "Uyghurville" in English) for food, etc.
NW said,
March 10, 2010 @ 5:14 pm
The velar quality is acoustically realized as a 'pinch' of formants F2 and F3 moving together. But as I understand it, both tone and speaker sex are located on the much lower F1, so I wouldn't expect them to interact. However, this is not my area, so I await a real phonetician to answer.
Pavel said,
March 10, 2010 @ 5:26 pm
Bee said,
[…] characteristic of some Southern Chinese dialects/accents that they have very little distinction between -n and -ng […]
This only affects the [-in]/[-iŋ] rimes, and it's [-iŋ] that surfaces instead of [-in] (or [ ĩ ] for that matter), so it wouldn't explain that pattern VM is describing.
bulbul said,
March 10, 2010 @ 5:31 pm
xyzzyva,
Duanmu in "The Phonology of Standard Chinese", p. 68:
He gives, among others, the following examples:
Now I'm no phonetician or phonologist (in fact, my attempts to learn Mandarin have proven that I'm deaf as a stump and couldn't tell the second tone from the fourth if my life depended on it), but I'd say that dropping a nasal whereby the preceding vowel becomes nasalised is quite different from a velar nasal turning into a dental one, which I think is what Victor describes.
Jesse Tseng said,
March 10, 2010 @ 6:27 pm
@bulbul
I think that is exactly the phenomenon described in the post. Some of the 4th tone -NG syllables in the recordings are pronounced with very little closure. They don't sound anything like the -N syllables, just to the left. (So I am a bit puzzled by the "come out sounding like -AN" part, and I also didn't notice the effect in -ING syllables.)
Syz said,
March 10, 2010 @ 7:01 pm
Two of my Sinoglot cobloggers and I have discussed -ng / -n endings at length recently. None of us (all in China's north: two in Dongbei and myself in Beijing) hear the linked-to pronunciation as problematic, so the quote from Liwei Jiao about it being "wrong", even with quotation marks, is fascinating. It seems he's speaking from his experience teaching pǔtōnghuà, which is an instructed language and (arguably) not anyone's native language.
Picking up from @duanmu and @jesse tseng's discussion, I think the n/ng similarity introduces this question:
…at least in some variations of Mandarin. My hypothesis is yes, and we just need the right experiment to show that.
bulbul said,
March 10, 2010 @ 7:28 pm
Jesse,
could be. In fact, that passage I quoted goes on to say:
And sure enough, the vowel in ANG4 sounds a bit more back than the one in AN4.
Still, I believe what Duanmu describes is a process very similar, if not identical, to what happened to nasals in certain positions in French, Portuguese or Polish and that involves the complete loss of the consonant and the nasalization of the vowel. Whereas here …
Hell, bù zhī dào. Someone with better ears fire up praat and let us know.
Richard Littauer said,
March 10, 2010 @ 7:34 pm
From a study I've done on external sandhi in final velar stops, I found that there was a direct inverse proportion in the value of f2 to f0 in the preceding vowel. Where f2 went up by around ten points, f0 would fall by one. These seemed to only occur predictably in cases where there wasn't a burst. I am not an expert phonetician, however, but it still might be worth noting.
Dan Bloom said,
March 10, 2010 @ 11:35 pm
Victor, since you know some things about Mandarin, and I know very little, despite living in Taiwan for 13 years and Japan for 5 years — I can only read Hebrew and French, brain is full — but I love living in Asia and being surrounded by Mandarin script, in fact, some days it appears like Hebrew letters to my eyes as I walk along the streets and then I blink and remember, oh, this is Taiwan, not Jerusalem, where did I take the wrong turn….SMILE….but question on this: many Westerners incorrectly believe that the Chinese word for crisis is the same character as for opporunity, when in fact they are two different double character terms: crisis is WEI JI — "danger [ji]" and opportunity is JI WEI – [ji] meeting…and both JI characters are NOT the same. So the canard oft repeated that "in China crisis is the same word as opportunity" is pure western romantic blind BS. But tell me more, as I am still trying to figure this one out. Reason i ask is that the movie The Hurt Locker in Japan is being called HA TOE LOKKA in hiragana of approximate English words, the Japanese sometimes do this rather than translate the title into real nihongo…but in Taiwan the movie is being called "wei ji dao su" which means Crisis Countown"…..and certainly does not mean countdown to opportunity. anyways, whatever you know, dish!
Does the Chinese character for CRISIS also mean OPPORTUNITY as many Western pundits and oped writers and reporters often say? NO. This is an Asian myth: there are two different words and characters for Crisis and Opportunity, and they are NOT the same. The West needs to be disabused of this falsehood. HOWever, there IS an old saying in Chinese to the effect that "Every crisis also has in its the seeds of an opportunity" — perhaps this is where the confusion comes from….
Katie said,
March 10, 2010 @ 11:56 pm
@NW–I do have quite a bit of phonetics training but am puzzled by the "low F0 makes it difficult to enunciate a velar consonant" hypothesis. Perhaps it's because it makes it difficult to maintain voicing at all, but I'm not sure why this should affect a velar more than an alveolar.
@Syz–I'm suspecting the hypothesis about the distinction being in the vowels might be correct. Surely someone has done this research? Anecdotally, I'm surprised by how often I mishear /n/ and /ng/–after all, as a native English speaker, you'd think I wouldn't have much problem. Lately I have been wondering if it's because I'm listening for the wrong thing. Some speakers, at least, show a very strong vowel distinction for /i/ before a nasal, and even my introductory Mandarin class taught us about the distinction for /a/.
A friend from Shanghai recently told me that southerners don't distinguish between /n/ and /ng/, and northerners laugh at them for it. Interestingly, her description of the difference between the two was not one of a consonant at all but of (what I interpreted as) one of nasalization on the vowel. Anybody know of any nasal airflow studies in Chinese?
Randy Alexander said,
March 11, 2010 @ 1:05 am
I'd like to shed some light on the difference between those of us who have said that the model in question is an adequate model, and Liwei Jiao's comment that Victor quoted.
The model honestly reflects what is accepted by your average Chinese speaker, and is an honest model in terms of preparing them for what they will hear all over greater China.
Liwei Jiao's perspective is that of a tester of the "national level" of the Standard Mandarin test. There are three levels for this test. Passing the highest level, what he calls the national level, is required of TV and radio announcers who broadcast nationally, and province-wide. They have to get 97% or higher to pass.
Putonghua is a prescribed model that in its pure form can generally only be heard in the speech of these announcers.
The looser model given in the pinyin website reflects (in my experience, at least) what one hears being taught in elementary school classrooms, for instance. Teachers of kindergarten, elementary school, middle school, and trade schools must get from 80% to 87% on the test.
Randy Alexander said,
March 11, 2010 @ 1:13 am
Another important datapoint: teachers who teach foreigners (对外汉语) must get between 87% and 92%.
Randy Alexander said,
March 11, 2010 @ 1:27 am
@Dan Bloom: try googling:
"language log" crisis opportunity
for a list of essays with links to other essays. This has been discussed at great length.
戴 蓋瑞 said,
March 11, 2010 @ 3:38 am
I've had Mandarin by four different instructors in the last nine months, a woman in her thirties from mainland Xian and a woman in her forties, a man in his fifties, and a woman in her sixties all currently living in Taiwan but whose backgrounds are unknown to me.
The differences in vowel qualities and final nasals is a phonetic trainwreck, slowly burning on its way down the tracks. We regularly hear the same incorrect final during comprehension exercises, and the fact that we do so as a group seems to indicate that it's a production issue.
While it seems to be irritating now, I think it will desensitize us to the variance we hear out in the rest of the world.
Another area that is hard to pin down are the front fricatives and affricates. Bopomo (Pīnyīn ), for example, is pronounced what sounds like [ʃ] by two of our teachers but [ɕ] by another, and out in public it ranges from [ʃ ~ ɕ ~ s].
There have been times that we have been corrected by two people simultaneously, each producing what sounds like a different sibilant to us.
I've been checking JSTOR and ordered _The Phonology of Standard Chinese_ by San Duanmu to try to make sense of this, but would welcome references to other literature about production variance in Mandarin.
Lugubert said,
March 11, 2010 @ 7:39 am
A Hangzhou (or Hangzou ;-)) woman made a very clear distinction between pan [pæn] and pang [pɑ˜n]. When I asked her if there was a similar difference between, say, in and ing, she didn’t understand the question, but replied something like ”You can’t compare them: in and ing are different.” After many years in Sweden and a couple of language courses, she was still at most into initials and finals; phonemes was an alien concept.
NW said,
March 11, 2010 @ 11:43 am
Erratum: Subtract 1 from all the formant numbers in my earlier comment.
James Dew said,
March 11, 2010 @ 3:12 pm
Victor, I'll copy here what I sent to you several days ago:
Yes, this is an interesting site, and fun. Listening to the -an/-ang sets I am reminded of an early teacher's (maybe Y. R. Chao) admonition to listen carefully to the VOWEL. The a in an is pronounced further forward (more similar to the vowel in the name "Anne") than the a in ang (like fAther or even a bit further back). The difference is very clear in the male voice, sometimes somewhat less clear in the female. I think you focused on tone 4 because the final consonant is (I think) shorter than in the other tones. (The whole syllable is shorter too.) To my ear the ng is especially short
in the first set, with no initial consonant – almost inaudible in the female voice – but what there is of it is ng, I think, not n. Some speakers feel that the vowel difference is as important as the consonant difference.
If you listen carefully to the -en/-eng sets you can hear a similar
distinction there, perhaps even more notable than in the a's. In a close phonetic transcription the more forward e's and a's are
transcribed with different IPA symbols than those of the further back e's and a's. See, for example, Chao & Yang's Concise Dictionary 國語字典, p. xii.
The phonemic explanation is that the two different e sounds can be united into a single phoneme because of complementary distribution, the one occurring only in front of n and the other only in front of ng, and similarly for the two different a sounds. Or turned around the other way, the phonemes e and a assimilate forward toward the contiguous front nasal -n and backward toward the back nasal -ng.
I've focused on another feature that distinguishes the syllables, but to come back to the -n/-ng question, I really didn't sense a problem with the articulations that I listened to on that site. Perhaps the problem is in the listener's ear rather than the speaker's articulation.
Jim
Caio said,
March 12, 2010 @ 10:29 am
I live in Jiangsu province, near Nanjing, in a city that speaks what could more or less be called Mandarin. Those who speak with a thick local accent (taxi drivers, street hawkers) usually pronounce the 'ng' as 'n' but lengthen the syllable (sorry for my lack of IPA I am lazy), and more middle class people will sometimes do this, or sometimes pronounce the ng very quickly, so I can barely hear it. However, people speaking something more standard (my boss) will always pronounce the ng clearly.
In the city just to the south, though, where they speak something more like what wikipedia calls 'Wu', the 'ng' sound is very distinct and what in Suzhou would be 'nong' usually comes out as 'ng' (like 'ng hao').
Don't know about other parts of China, don't travel much.
minus273 said,
March 13, 2010 @ 10:21 am
Food for thought.
http://www.ling.sinica.edu.tw/eip/FILES/journal/2007.3.9.12623232.4768408.pdf
From Pulleyblank the Great, with a quotation from ever greater Y.R. Chao.
Julie said,
March 13, 2010 @ 2:48 pm
I'm taking a first year Mandarin course right now, and on one of the computer programs we're using (called Professional Interactive Chinese), I have a lot of trouble sometimes distinguishing the velar codas… there's something odd about them… they often sound alveolar.
FS said,
March 15, 2010 @ 1:40 pm
Re: Katie – Anybody know of any nasal airflow studies in Chinese?
Zhang, Jie. 2000. Non-contrastive features and categorical patterning in Chinese diminutive suffixation: MAX[F] or IDENT[F]? Phonology 17 (3). 427-478.
Section 2 of above mentioned paper reports an airflow study comparing the amount of nasalization in the vowel preceding /n/ vs. /ng/ codas for two male native speakers of Beijing Mandarin. The results indicate that " that the /ng/-coda induces a significantly longer nasal flow duration on the vowel nucleus than the /n/-coda" (436).
The fourth tone in Standard Chinese tends to be significantly shorter than the other tones. This could explain the loss of distinctiveness especially in the fourth tone examples, when the lengthening cue (velar nasal airflow) and the shortening cue (fourth tone) conflict.
Concerning the Hangzhou and Shanghai examples: The absence of alveolar nasal codas is indeed widespread especially among southern Chinese dialects such as the Wu dialects (compare Chao, Yuen Ren. 1967. Contrastive Aspects of the Wu Dialects. Language 43 (1). 92-101.).
On an anecdotal note, my first Chinese teacher, originally from Hangzhou and thereby native speaker of a (velar nasal-less) Wu dialect, payed particular attention to my distinguishing velar and alveolar nasal codas when I did my first steps into acquiring Standard Chinese, as did her teacher with her when she first learned Standard Chinese. Maybe the observed substitution of velar for alveolar codas on the website mentioned above can be traced back to an over-generalizing speaker with a southern background?