Dissimilation, stress, sandhi, and other tonal variations in Mandarin

« previous post | next post »

A few months ago on the Penn campus I heard a Chinese guy and a girl having a conversation in Mandarin, and I was surprised when he twice said, "Wo3 ming2bai4 le."  The rest of his speech was standard, but then he came out with this strange transformation of "Wo3 ming2bai le".  Of course, I shouldn't have been surprised, because I've heard the exact same thing before.  Nonetheless, it still sounded odd to me, since from first-year Mandarin on I've had it drilled into me that this sentence should be pronounced "Wo3 ming2bai le" and that any other pronunciation of ming2bai was wrong.  This was reinforced by the canonical pronunciation ming2bai given in dictionaries and other authoritative sources.

Yet you can hear the 4th tone for yourself in the recording here, even though the phonetic annotation is given as míngbai (neutral tone on the second syllable).

How do we account for this?  Is it some sort of emphasis, with stress on the second syllable, which normally should be neutral?

It is universally recognized that Sinitic languages are tonal, but if people play fast and free with the tones this way, how can intelligibility be maintained?

On the other hand, are tonal languages like Chinese subject to ad hoc change of tone or other phonological aspects of speech the same way as non-tonal languages?

One native Pekingese speaker demonstrated for me that she pronounces the second syllable of 明白 three different ways, depending upon the situation:

1. míngbai — when making a normal, declarative sentence

2. míngbái — when pointedly asking someone else if he understands

3. míngbài — when affirming emphatically that she understands

I've also heard some people drop the neutral tone on the second syllable so low that it sounds like a third tone (what some teachers of Mandarin call a "half-third tone".

We have previously discussed on Language Log the phenomenon of intonation and stress superseding tone in, among other places, this post:  "When intonation overrides tone" (6/4/2013) (see also the cited links and comments thereto).

Or perhaps the change from míngbai to míngbài is simply a kind of dissimilation, since the two syllables of míngbai 明白, if pronounced separately, would be míng bái, which would result in a kind of tone sandhi, viz., míngbài.

San Duanmu explains:

Yes, this is called 轻声变去声 (T0 to T4). An example is pu2tao2 'grape', which becomes pu2tao0, and then pu2tao4. The rare thing about that speaker is that he uses all the three versions at the same time. T0 to T4 usually happens when the preceding tone is T2, and the change is probably phonetic at first, then reinterpreted as H spreading: LH-0 –> LH-L –> LH-HL

According to Jonathan Smith,

I know what you're referring to and think the realization of the neutral tone in this particular tonal context is easily reinterpretable as the full fourth tone — it's likely that the speaker here would still report he had produced the neutral tone, but perhaps given enough momentum, this sort of thing can lead to real category change? I've actually often wondered if the Beijing tendency to pronounce 室 in the third rather than the fourth tone got started with tonal neutralization in certain contexts (particularly after fourth tone), with realizations of those neutral syllables gradually taking on third-tone-ish contours.

In the case of míngbai (underlying míngbái) –> míngbài, it is the second syllable that changes, but in the case of 买马 ("buy a horse"), it is the first syllable that changes:  underlying mǎi mǎ becomes mái mǎ (this [two third tones in succession] is the most common form of tone sandhi, and usually the only one that is taught).  Yet there are plenty of other categories of tone sandhi in Mandarin that generally are not taught, but that naturally (physiologically) are used in daily speech.  For example, my name is Méi Wéihéng 梅维恒, but only a neophyte would attempt to pronounce it that way, and it would sound very awkward if he did.  Instead, it becomes Méi Wēihéng.

At Harvard, Professsor Rulan Chao Pian, daughter of Y. R. Chao, prepared a lengthy pronunciation guide that was used for the first several weeks of the introduction to Mandarin course.  She developed it on the basis of notes made by her father, who had a particular interest in such phonological niceties.  I used that guide in a summer school Mandarin course I taught at Harvard one year, and I remember that it included many different types of tone sandhi, each of which is real, but few of which are ever taught systematically, if at all.

So, it's certainly not just míngbai –> míngbài, that we have to contend with, but a whole host of phonological changes in spoken Mandarin that are different from what is to be found in textbooks and dictionaries.  Here are just a few examples:

wǒ kànjiàn le 我看见了 ("I saw") instead of wǒ kànjian le

wǒ dè 我的! ("mine!") instead of wǒ de

shénmè 什么?! ("what?!") instead of shénme

Gloria Bien discusses a particularly complicated case that has also puzzled me for a long time:

The tone change I've mused about, and maybe even wrote about before, is between yīhuǐr , yìhuǐr, and yíhuìr.  The "yi" is pronounced quickly, so sometimes I hear it as first tone, though the "rule" says "yi" here should be fourth tone.  The characters chosen for it are 一会儿.  Wenlin says 会 should be pronounced 3rd tone when it means "moment."  But I often hear yíhuìr, with an actual 2nd tone on the "yi" and what sounds to me like an emphatic fourth.  But maybe it sounds emphatic because I'm expecting a third tone.  I have never heard it without the "er" sound.  I have heard "yíhuìr" (2nd, 4th) even from Chinese language teachers.

For the record, the individual tones of the underlying morphosyllables that constitute this expression are:

1 ("one")

4 ("can; meet; conference; society; opportunity", etc. — about two dozen other meanings)

2 ("child; son; youth; noun suffix") — as a sign of retroflexion (as in the present case), of course, it has no tone

Now, if I've heard this expression once, I've heard it ten thousand times, and it was always pronounced thus (giving a complete sentence):

děng yīhuěr

As written in characters, that would be 等一会儿 ("wait a moment"), which, according to dictionaries, would supposedly be děng yīhuì'er, but if I heard someone say that, it would grate on my ears.  My wife and all of her friends always said "děng yīhuěr", and I can guarantee that the final third tone was emphatic and clear.  The mainland dictionaries I've looked at, including the great Hanyu da cidian [HDC], do not even list a third tone pronunciation for 会 (though HDC, 5.782-3, besides huì, also gives kuài and kuò for specialized meanings).

So where does the third tone reading, which is the only one I've known for the expression yīhuǐ'er, come from?  The impressive new Běijīng huà cídiǎn 北京话词典 (Dictionary of Pekingese), by Gāo Àijūn 高艾军 and Fù Mín 傅民, does give the pronunciation yīhuǐr, as well as yīhuǐzi, with a different nominal suffix (see p. 984).  In both cases, Gao and Fu explain the unusual pronunciation for 会 (huǐ instead of the normal huì) by saying that it is "an old reading").  I'm beginning to wonder whether 会 (trad. 會) is the right character for this morpheme after all.

The huǐ part of yīhuǐr (yīhuěr) and yīhuǐzi may have been a morpheme in the spoken language for which there was no known character.  But under the compulsion to find a character for every syllable in the language, someone may have arbitrarily assigned this syllable to 会 (trad. 會).  (See the discussion of běnzì 本字 ["original character"] here and in other Language Log posts.)   The fact that some people are now pronouncing yīhuǐr as yīhuìr would be an example of how script can influence speech.

[Thanks to Jonathan Smith, San Duanmu, Gloria Bien, Fangyi Cheng, Liwei Jiao, Stephan Stiller, Randy Alexander, Ziwei He, Wei Shao]


  1. Michael Watts said,

    August 26, 2014 @ 11:26 pm

    It is universally recognized that Sinitic languages are tonal, but if people play fast and free with the tones this way, how can intelligibility be maintained?

    Presumably the same processes are at work that allow Chinese people to understand song lyrics, or pinyin written without tones. The evidence would suggest it's not very difficult.

  2. Bob Ladd said,

    August 27, 2014 @ 2:26 am

    Michael Watts is correct that "it's not very difficult" to understand when tones get overridden or omitted in writing. I'd go further and say that this whole topic is affected by the assumption that there's something special about tone in this respect. This assumption is not helpful. After all, it's not very difficult for whispered speech to be understood in languages with voicing distinctions, or for competent readers of Hebrew and Arabic to get by without vowel marks. Language is redundant, and we all use top-down cues all the time. It's only glottocentric assumptions on the part of speakers of non-tonal languages that make this seem like a weird and mysterious problem.

  3. Michael Watts said,

    August 27, 2014 @ 4:12 am

    This bit of text beginning the comment is meant to distract the spam filter, which seems to have swallowed my original comment.

    t mght b wrth pntng t tht nglsh cn ls b rd wtht vwls. Y gv p lt, bt y cn d t.
    n th thr hnd, lss cmmn wrds cn prsnt dffclts, sch s f wntd t sy "Vctr Mr lks t ffct snstnlzd wrtng stl, bt h my rlz tht hs rhtrc s ftn vr th tp".

  4. Victor Mair said,

    August 27, 2014 @ 6:14 am

    It seems as though some folks do not understand the function of rhetoric in writing.

  5. leoboiko said,

    August 27, 2014 @ 6:19 am

    My glottocentric assumptions made it seem weird and mysterious the way that English speakers can communicate while reducing half of the vowels to neutral schwas (or omitting them altogether). I recall my astonishment at this BBC ESL guide – Latin letters ‘a’, ‘e’, ‘o’, ‘u’ all read as the same sound /ə/! (I had been pronouncing them Latin-like for years…) Stress-timed languages are weirder and mysteriouser than tonal languages.

  6. Richard W said,

    August 27, 2014 @ 6:48 am

    There are numerous audio examples of 明白 on the following page. Although not all of the "Play Audio" buttons worked for me, many of them did.

  7. Victor Mair said,

    August 27, 2014 @ 6:58 am

    @Richard W

    This is a wonderful resource. Thanks so much for sharing it. Starting with the main entry and then listening to quite a few of the other recordings, it sounded to me as though many of the -bai syllables were 4th tone, and even some of the ones that were probably intended to be neutral came out in an ambiguous fashion or even a little bit on the 4th tone side.

  8. Richard W said,

    August 27, 2014 @ 7:04 am

    Yes, I find the Chinesepod Glossary very interesting and useful at times.
    Here's their page for 一会儿:

  9. Bathrobe said,

    August 27, 2014 @ 8:15 am

    Regarding yīhuǐr, I'm curious whether there is any connection between this phenomenon and Beiing bàngōngshǐ (办公室 'office'). I always use bàngōngshǐ, but many 'correct' speakers use the correct form, bàngōngshì, which frankly grates on my ears. Is it a coincidence that this follows the same tonal pattern as yīhuǐr?

  10. Akito said,

    August 27, 2014 @ 10:15 am

    Don't textbook-style yíhuìr and jiàoshì—away from Beijing indigenous yìhuǐr and jiàoshǐ—represent the process of standardization? I guess that's what pǔtōnghuà is all about, after all.

    As a beginner learner of Mandarin for the last 48 years or so, I seem to hear shěnme much more frequently than shénme. I agree that language is redundant for the most part, but wonder if these small variations don't actually carry subtle nuances in style or register.

  11. GAC said,

    August 27, 2014 @ 12:58 pm

    My wife (from Henan), produces something that sounds like 4th tone to me, but she perceives it as a neutral tone, so that would seem to be a data point in favor of what Duanmu told you, that the apparent 4th tone can apply after the tone is neutralized.

  12. richard said,

    August 27, 2014 @ 2:16 pm

    Aha! I thought I heard this from a desk clerk in Shanghai just a month ago (specifically the wǒ dè example listed above), but since my Chinese would need a lot of work just to be classified as "rudimentary," I asked my wife (a native of Hunan) if the tone changed in a case like this. She suggested the clerk just made a mistake, that it shouldn't change like that.

  13. Victor Mair said,

    August 27, 2014 @ 2:39 pm

    There you go, Richard. In nearly all of the cases I cited above, the speakers would probably tell you that they pronounced the items according to the canonical tones prescribed in dictionaries, but the empirical evidence demonstrates clearly that they did not.

  14. Bathrobe said,

    August 27, 2014 @ 7:46 pm

    There is also wǎng and wàng for 往 (since standardised).

  15. Bathrobe said,

    August 27, 2014 @ 7:47 pm

    Also, isn't possible to pronounce 我的 as wǒ dì (perhaps in jocular style)?

  16. Victor Mair said,

    August 27, 2014 @ 8:11 pm

    I still often say wǒ dì, in emulation of Chinese friends from whom I learned it decades ago.

  17. Dave Cragin said,

    August 27, 2014 @ 9:39 pm

    Is the lack of self-awareness that someone is not pronouncing canonical tones prescribed in dictionaries similar to the many examples when spoken English is different than written English?

    i.e., most Americans say “boddle”, not “bottle”, “baddle”, not “battle,” “gonna” not “going to” etc without actually thinking about how they say the words. Water & library are 2 other common examples.

    If Chinese asked an American “how do you say “bottle”?”, I’m guessing most Americans would give the dictionary pronunciation without realizing they rarely say this.

    It was interesting to see the post about the Chinesepod pronunciation guide. I subscribed to Chinesepod for years, but I primarily downloaded lessons and I was never aware of the guide.

  18. Victor Mair said,

    August 28, 2014 @ 5:09 pm

    From W. South Coblin:

    It looks like your various interlocutors have pretty well covered the issue. But, for what is is worth, I put your question to the two current members of my little linguistics seminar today. Both are Mainlanders, one from Shijiazhuang and the other from Weihai in Shandong. To me, they both sound like northerners speaking entirely standard PTH. In their common opinion, míngbái sounds like Taiwan Guoyu, míngbài is what just about everybody, including them both, actually says on the Mainland now, and míngbai is hardly said by anybody, being just a dictionary form today. They also suggest that the bài of míngbài has evolved from the neutral tone form of the syllable. They do not consider it an emphatic form, just the way it is said in modal speech. That’s straight from the mǎ’s mouth. I make no claims for the truth or accuracy of any of it. Like Confucius, I’m just passing it on. They are both professional linguists and dialectologists whose current interest is, like mine, is comparative reconstruction.

  19. Richard W said,

    August 28, 2014 @ 9:09 pm

    In regard to "míngbái sounds like Taiwan Guoyu":

    A couple of the Taiwan dictionaries I checked give míngbai as the pronunciation (not that this can be relied on). However, 兩岸詞典 does give míngbái as the Taiwan pronunciation (though it gives the usual míngbai as the 大陸 pronunciation).

  20. Richard W said,

    August 30, 2014 @ 5:57 am

    Regarding "pu2tao2 'grape', which becomes pu2tao0, and then pu2tao4", I believe you can hear pu2tao4 in the following recordings:

    Regarding Professsor Rulan Chao Pian, you can hear her pronunciation of 明白 in the following audio snippet (recorded perhaps 50 years ago). I believe she is saying "míngbai" here (not míngbái or míngbài):

  21. Ned Danison said,

    August 30, 2014 @ 8:06 am

    This seems like a good place to carry on my campaign to make native speakers of Chinese admit that when they say 然后 ránhòu it sounds like 那后 náhòu.

    (Just between you, me and the lamppost, I still remember the day I realized that what they were saying corresponded to the characters 然后. But hey, I was self-taught.)

  22. Bathrobe said,

    August 30, 2014 @ 11:02 am

    "make native speakers of Chinese admit that when they say 然后 ránhòu it sounds like 那后 náhòu"

    I've never noticed this in Beijing.

  23. Ned Danison said,

    September 3, 2014 @ 7:02 am

    Bathrobe, I got this 那后 impression while living in Kaohsiung, Taiwan in the 90s.

  24. Richard W said,

    September 3, 2014 @ 4:18 pm

    @Ned Danison
    I got that impression, too, living a little north of there in Taichung, also in the 90s. And on the following page, too, it sounds like náhòu to me.

  25. Richard W said,

    September 3, 2014 @ 4:49 pm

    Oops, I take back the second part of my comment (about the ChinesePod audio). The relevant page is actually
    and it sounds like ránhòu there.

  26. Yuanfei said,

    September 3, 2014 @ 8:05 pm

    What a subtle post. I am getting to wonder how frequent the term 这会子 can be interchanged with 这回子. They all appear in the Dream of the Red Chamber to mean "at this moment," but the tones of the two characters are not the same.

  27. Ned Danison said,

    September 4, 2014 @ 8:06 am

    Richard W, I'm starting to get the feeling “那后” is a local phenomenon and occurs only in fast connected speech. I hereby suspend my campaign.

RSS feed for comments on this post