Jump to content
Learn Chinese in China

Mandarin tone sandhi


Recommended Posts

Site Sponsors:
Pleco for iPhone / Android iPhone & Android Chinese dictionary: camera & hand- writing input, flashcards, audio.
Study Chinese in Kunming 1-1 classes, qualified teachers and unique teaching methods in the Spring City.
Learn Chinese Characters Learn 2289 Chinese Characters in 90 Days with a Unique Flash Card System.
Hacking Chinese Tips and strategies for how to learn Chinese more efficiently
Popup Chinese Translator Understand Chinese inside any Windows application, website or PDF.
Chinese Grammar Wiki All Chinese grammar, organised by level, all in one place.

I hear 胡同 as hu2tong4, though I'm not sure if dictionaries agree.

2000年 is a good example of 'trochaic' rhythm. At the lowest level of approximation, Chinese likes rhythmic chunks of 2 or 3 syllables each, which is probably why 2000年 is read the way it is.

'Stress' in Chinese is an elusive concept. There are many disagreements out there and you'll find probably no two books or researchers say the same thing. Syllable duration I think is the most obvious correlate of 'stress' in Chinese; generally, given a 'rhythmic chunk' the last syllable is the longest, the first one is second longest, and the ones in between are shorter. The most fascinating theory about stress in Chinese I read is Prof. Shen's (see 北大中文论坛), though it's so complicated that you'd go nuts if you tried to remember all that stuff when you speak. :shock:

As to those who believe that 'tones are not important' (especially teachers), I'd just like to say that not all learners are the same. Some of us have to speak in public, and you want to be as clear and accurate as possible, to the point that you don't draw too much distracting attention on you 'being foreign' (obviously having a Chinese face is an advantage, but not all there is to it). The point is that obsession with pronunciation should not stop you from achieving fluency, improving your vocabulary etc. So many things to learn, so little time.

Link to comment
Share on other sites

yep...way too much stress on the tones.

Its confused the hell out of me when I started.

Actually I am sure its better to spend time either learning more vocab or (even better) listening to the language being spoken. I have found that by far a more effective way to learn Mandarin.


Link to comment
Share on other sites

In fact, I had textbooks and dictionaries for a long time but didn't venture to start learning because of the tones. I thought - I will satrt learning incorrect pronunciation and then it would be hard to relearn.

Now, I've got a lot of audio materials but I am still obsessed with not being able to pronounce certain words or sentences with the right tones. Tones are pain but I want to overcome this. I am marking characters with tone marks, so I can read them later ( I forget the tones sometimes).

Link to comment
Share on other sites

Bless your heart. Tones are important, but it's not the no.1 in your study list. If you pay more attention to do these:

Listen to appropriate listening materials (not TV, not movies)

Increase your vocab (slowly)

Free practice under certain guidance

Use Mandarin when you have a chance,

Tones will take care of themselves. For practical reason, don't bother to read those scholar's acedemic stuff. (True reason could be: If they don't write it, they won't get promoted, or they can't further up to Phd, etc.)

Edited by roddy
Link to comment
Share on other sites

  • 5 weeks later...
It's sad see so many people are strugling with tones. Foreign students are overwhelmed by tones, and that is NOT necessary. The simple truth is if you speak fast enough, people will only hear fluent Manarin and understand you very well. Tones disappeared on the surface of earth.
Thanks shiaosan for your encouraging explantions. I have been learning Manadrin for just over a week. Studying on my own (with a CD) I am worried that I will not be understood by anyone: although I don't think I will have to say this very often -"Ma1 ma qi2 ma3 ma3 man2 ma1 ma2 ma3". On tones I found that:

(a) If 3rd tone follows 3rd tone then first goes to 2nd tone.....ni3, hao3 > "Ni2 hao3";

(B) 3rd tones in series.... wo3, hen3, hao3 > "Wo3 hen2 hao3" unless next 2nd tone is in the next part like "Qing3 gao4su wo3 ni3 de dian4 hua4 hao4 ma3";

© When bu4 is followed by another 4th tone then bu4 > bu2 like "Wo3 bu2 shi4 Zhong guo2 ren2";

(d) For repeated words, the first word remains in its original voice and the next word(s) is soft, without tones, like "Xie4 xie"!

Is the Mandarin language really divided into just 4 or 5 tones? I expect that in real life this is just a straight-forward way of trying to explain to non-native speakers the key divisions of tone. What I am trying to do in my study is listen to words spoken before looking at the text (pinyin with tone marks or tone numbers) so I can remember how the sound sounds rather than looks :-)

Anyway, once again shiaosan, thanks for your execllent posts on this subject.


Link to comment
Share on other sites

It has been my experience that chinese people cannot understand you without the correct tones. I once said' date=' wo3 men dei3 qu4 shang4 ke4. And my friend quickly blurted out, "it is dei3" not "dei4, third tone!". All the while I thought I used the third tone! So I think tones have a great deal of importance.


If this is important, what happens in the context of popular music, where there are no tones?


(also, "dei" 得 apparently only exists with the 3rd tone, and no others, so...)

Link to comment
Share on other sites

As for songs, I hear people have lyric sheets

you are serious? this is what people use? hmm i wasnt expecting the language to be inadequate like that. Unfortunately there doesn't seem to be any fix for this, as Chinese must have evolved without musical influence/or a differet variety. music is a special case, though.

Link to comment
Share on other sites

also, about the thread, I am starting to believe that the full third tone is not fully pronounced in regular speech (especially faster speech)

if you listen to a proper pronounciation of a third tone (in learning audio, etc..) it seems almost two-syllabic, or drawn out, due to the dipping nature of the third tone. ie some random syllable, "qing3"... the first part (qi) is pronounced with a stonger/forced sound like a falling tone, then (just as the sound trails off) there is an extra push of air with the "ng" to make the tone rise up again --> qi /ng/...wo /oh/...lia /ng/ etc... I think there would be a noticable hang-ups in the flow of speech if someone tried to pronounce these 3rd tones fully while talking quickly. It seems they often turn them into tone 2's.

Link to comment
Share on other sites

My feeling is that if a foreigner is learning Chinese, and he/she cares about tones, there is no point in putting it off, as it only promotes hard to break bad habits. Since we did not grow up hearing and making or caring about tones, it takes varying degrees of effort to make yourself hear the tone differences and reliably produce the one you intend. I say varying because, it certainly helps if you play/played a musical instrument and therefore have some experience critically listening and correcting sounds you produce. So unless you are actively working on making the right tone, that is listening hard in a way your brain is not used to, I can’t see that an adult brain is ever going to start processing the tone information to allow one to produce, even down the line after lots of exposure to native speakers. Adult brains just aren’t the same as kid brains.

So, unless there is a vast body of hard evidence for adult learners of Chinese, I am extremely skeptical that "tones will take care of themselves" for nonnative speakers.

As for when to neutralize, consider the text Spoken Standard Chinese vol 2 by Huang and Stimson. They are explicit about neutral tones. Not just in the vocabulary lists, which for example listed zhi1dao and bu4zhidao4, zhong1guo, mei3guo, rather than zhi1dao4, etc.

But also in the pinyin transcriptions of the dialogs. Here for example is dialog 1 from lesson 21.

A: wo yi4shuo1 qi kao3shi4 lai jiu zhao1ji2, wo yi4zhao1i2 jiu shui4buzhao2 jiao4. etc.

I just listened to associated audio of this and the pinyin and indicated tones or lack thereof is pretty accurate although there is always room to quibble.

I am not sure, but I think what they did was have native speakers, who were also trained linguists, transcribe tapes of other native speakers “acting” the dialogs, and then had them figure out where the neutral tones were in natural native speech. Keep in mind these texts were produced before desktop computers of any kind. Might have had a Wang, but probably an IBM Selectric.

anyway, my 2 cents, recognizing that what works for me, almost certainly doesn’t work for everyone

Link to comment
Share on other sites

With tones, I've found, when in doubt, leave it out.
I'd check it out in a dictionary :mrgreen:
"As for songs' date=' I hear people have lyric sheets" [/i']

you are serious? this is what people use? hmm i wasnt expecting the language to be inadequate like that.

With some styles of music in English, I still need the lyric sheets :wink:
I am extremely skeptical that "tones will take care of themselves" for nonnative speakers.
Very well said, kudra! 8)
Link to comment
Share on other sites

  • 3 weeks later...

I don't speak any Chinese language/dialect, but work on trying to get machines to recognize tones. A (non-native) Mandarin speaker and I did a quantitative study last year on how important it was to recognize tones in Mandarin. (To find it, try googling "importance tone mandarin".) It was based on language statistics from news broadcasts, not conversational speech, but hopefully that doesn't matter too much.

We found that recognizing tones in Mandarin is as important as recognizing vowels.

If you consider isolated words (i.e. you can't use contextual information), someone who cannot recognize tones in Mandarin loses 2% of all available information. To compare, someone who can't recognize vowels (i.e. all vowels sound the same) also loses about 2% of all information, while someone who can't recognize consonants loses about 8%.

Of course, with context, less information will be lost, so you can survive quite a bit without tone using only guessing. For example, y__ m_st h_v_ pl_y_d g_m_s wh_r_ s_m__n_ l__v_s _ll th_ v_w_ls __t _nd y__ h_v_ t_ g__ss wh_t th_ w_rds _r_. I'm not sure if vowels are written in Hebrew - I think not. Point is, one can survive without vowels (in written language), and survive, but it's a lot easier when you can use them. It's possible that tones are similar. Maybe.

Link to comment
Share on other sites

If you consider isolated words [ ... ] someone who can't recognize vowels (i.e. all vowels sound the same) also loses about 2% of all information
That sounds quite incredible. I take it that is meant for the English language (as for Mandarin, you'd be lost without vowels), but even there it seems very, very little to me.

Add to that: I'm still not very far with my Chinese, but I think I would much prefer listening to Chinese without tones (in fact, I have the suspicion that this is what I do!) then listening to English, or even German, without vowels.

Link to comment
Share on other sites

I think I agree with Gougou and that you need to explain your terms a little bit.

I skimmed the study you mentioned, and I thought it expressly said that one could not talk about the percentages in absolute terms, but only in relative terms. In other words, I thought you could say that vowels and tone were equally important in Chinese, but not that vowels or tones carried 2% of the information in a given sentence.

In lay terms, it would seem to me that someone unable to recognize consonants in English or Mandarin does not lose "8%" of all available information, but rather 100%. No information can be conveyed in these languages in the average sentence without distinguishing consonants. It does strike me as reasonable, however, that consonants in Mandarin are 4 times more important than vowels.

Your major finding that vowels and tones carry about the same functional load in Chinese strikes me as interesting and probably correct; however, I think you need to be careful to interpret this in lay terms. For instance, many Chinese speakers are used to hearing tone mistakes and can guess meanings even when these exist. I think there is much less familiarity and therefore tolerance for random substitution of vowels.

Also, I think this sort of thing is highly language specific. Some languages rely relatively little on vowels, such as Hebrew and Arabic; whereas others rely tremendously on them, such as Hawaiian.

Link to comment
Share on other sites

Yes, I should definitely clarify my remarks.

Gougou : when we speak of "not recognizing vowels", we don't mean the situation where all vowels disappear, but where all vowels are replaced by a single vowel placeholder (e.g. in English with unrecognized vowels, "amen" and "omen" would be seen as the same word, but not the same word as "men")

Altair is correct - functional load is a relative measure and our study only claims that tones are as important as vowels. I had used them absolutely in my post thinking that that would make it more understandable in lay terms, but I was wrong. Sorry. While it is true that if the following assumptions hold (and the only ones I'm really worried about are 1 and 2)

(1) you know where words begin and end

(2) you only consider isolated words

(3) you use information theory to take into account word frequency

(4) you use word frequencies based on a large collection of news broadcast speech

then the amount of information lost when you can't recognize vowels/tones/consonants is 0.02/0.02/0.08 respectively (see example later), it is not clear how those numbers correspond to a measure of difficulty in the full speech recognition process (especially word segmentation). So they are best interpreted relatively. And hopefully someone else will come along later and propose a better measure than we used.

Btw, even if you can't recognize consonants, there's still lots of information available to help you recognize words - word length, the vowels, the tones. A lot of words will sound the same. If you choose one of those words with probability proportional to the (logarithm of) how many times the word normally occurs, then you will be correct about 92% of the time. That seems high, but it's because we are not considering all words equally - the more frequent ones are more important. (And remember that we assume that we know word boundaries.)

Different languages should give different importance to consonants/vowels/various aspects of language. We only did calculations for Mandarin, German, Dutch and English (and to some extent, Cantonese). We can do the calculations for any other language with a lot of data - and doing so for Hawaiian would be very interesting - I should go hunt for some data... thanks for the idea!

Link to comment
Share on other sites

  • 4 years later...

An interesting phenomenom is the fact that in whispered and sung Chinese tonal information is lost, yet the Chinese can understand what's being said. In cases where a word is ambiguous, the host sentence will usually provide sufficient context to fill the gaps.

For this reason one would have thought that, since tones are not needed in singing or whispering, messing up a couple of tones here and there in normal speech wouldn’t be much of an issue. However, it is, and here's why: the listener is expecting to hear one thing, but he/she is hearing another. If I told you to look at the red car while I point my finger at a blue car, wouldn't you be confused? Sure you would! It would be less confusing if I kept my hands in my pockets. A similar conflict occurs in speech, where the context is telling one thing but the tones are pointing in a different direction. This situation will have the listener baffled for a split second, and by the time they've resolved the conflict in their head (i.e. figured out what you actually meant), they will already have missed out the next couple of words you said afterwards, making it extremely hard to follow the conversation.

Generally it is better to say the tones correctly, particularly at the beginning, even if it sounds a little unnatural and exaggerated. Once you've internalized the correct tones through frequent listening and use they should begin to sound more natural. But you do need to try. As regards the 3rd tone sandhi... I can't see how it's ever going to sound natural, even from native speakers, because if they were to rattle off an unfamiliar sequence of 3rd tones (one they've never said before), it'd be impossible for them not to improvise! Furthermore, two native speaker may well use different tones!

The tones and tone sandhi are indeed the works of the devil. I enjoy learning Chinese but I still think tonal languages are way inferior to non-tonal languages because it's easier to be misunderstood. I am Spanish, and after spending two months in England I could already be understood quite nicely. It was evident I was a foreigner, but I could be understood. After 5 months in China I still have trouble making myself understood even when I say very simple sentences. There's just too much stuff to think about, and tones aren't helping either.

Here's a very contrived mofo I came up with. I am sure there'll be nastier ones:


mǐlǎoshǔ zhǐ xiǎng mǎi hǎo shuǐguǒ. :clap

Link to comment
Share on other sites

Join the conversation

You can post now and select your username and password later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Click here to reply. Select text to quote.

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.


  • Create New...