Tone Sandhi

Discussions on the Cantonese language.
James Campbell

Tone Sandhi

Post by James Campbell » Thu Jul 04, 2002 5:46 am

Just wondering if anybody has any information on tone sandhi patterns in Cantonese, or any of its dialects. I haven't come across any information but it makes me wonder if it does exist anywhere.

One instance does come to mind, and since I don't know very much Cantonese, I would like to know how you use it. There are usually two readings in the first tone. Do you find there's a difference between Hong Kong and Guangzhou speakers? Which city makes more use of both readings? Finally, my main question is, what environments causes you to pronouce it level or falling? I'm thinking about the word xian1-sheng1 (Mandarin spelling) followed by 1st tone surnames like Zhang1.

Does anybody know of any other rural dialect where tone sandhi takes place or where you pronounce a word with two different tones depending on the situation?

Thanks for any help with this.


Re: Tone Sandhi

Post by ichi » Thu Jul 04, 2002 7:56 am

Actually this is a great question, because I'm wondering the same thing.

Tone sandhi definitely occurs in Mandarin. If I understand correctly, this is modification of tones in a string of tones, right?

Anyways, IIRC at least one of the cases of tone sandhi in Mandarin should be 3rd followed by 4th; in this case the 3rd tone is clipped to just the falling part. I think sometimes this occurs for 3 followed by one as well ... like lao3 shi1 (老 師).

Tone sandhi in mandarin is documented; I believe a textbook such as _Integrated Chinese_ should contain this information.

My guess is that it occurs in Cantonese, but it's not something that I've ever paid concious attention to.

I hope a linguist will come by and clean up the mess I'm making :)

Side note -- there are definitely words mean different things with different tones (but otherwise sound the same), and there are some words that change *sound* (not just tone) depending on context.
Mark Williamson

Re: Tone Sandhi

Post by Mark Williamson » Thu Jul 04, 2002 4:06 pm

Generally non-Mandarin dialects have more tone sandhi than does Mandarin

And generally Mandarin tone sandhi is not nearly as confusing as that of other dialects, which usually have appterns even though those patterns are often very complex. Dylan Sung's sent me a pattern chart for Hakka tone sandhi. Unfortunately for Mandarin sandhi people who are learning the language see things like Tone so-and-so with tone so-and-so changes to tone so-and-so however often the tone is not the same as one of the original tones, or at least not exactly although generally it doesn't make a difference.

The Hakka BASE tones are:

#1. 陰平 /44/
#2. 陽平 /11/
#3. 上聲 /31/
#4. 去聲 /53/
#5. 陰入 /1/
#6. 陽入 /5/

陰平+陰平 /44/ + /44/
陰平+陽平 /35/ + /11/
陰平+上聲 /35/ + /31/
陰平+去聲 /35/ + /53/
陰平+陰入 /35/ + /1/
陰平+陽入 /44/ + /5/

去聲+陰平 /53/ + /44/
去聲+陽平 /55/ + /11/
去聲+上聲 /55/ + /31/
去聲+去聲 /55/ + /53/
去聲+陰入 /55/ + /1/
去聲+陽入 /53/ + /5/

輕聲 (qing sheng, neutral tone) is always /*3/ apparently, however:
陰平+輕聲 /35/ + /*3/
去聲+輕聲 /55/ + /*3/

So, apparently in Hakka all other combinations are read as their base tones. Please correct me if I'm wrong here, but that's what I interpreted from Dylan Sung's email.

About Cantonese, I'm not quite sure.

However about some words changing sound depending on context- yes, that happens however generally from what I from what I can tell, there's no method to THAT madness (except perhaps in Fuzhou dialect where there are regular sound changes with a pattern to them)

I do know that this happens in Japanese:

かきくけこ ka ki ku ke ko - がぎぐげご ga gi gu ge go
さしすせそ sa shi su se so - ざじずぜぞ za ji zu ze zo
(nihon roomaji-- sa si su se so - za zi zu ze zo)
たちつてと ta chi tsu te to - だぢづでど da ji zu de do
(nihon roomaji-- ta ti tu te to - da di du de do)
はひふへほ ha hi hu he ho - ばびぶべぼ ba bi bu be bo OR ぱぴぷぺぽ pa pi pu pe po

SO when we have two kanji (hanzi) together, for some words we will use the Chinese pronunciation (most), and for some the Japanese pronunciation (some).

音読み (onyomi) CHINESE READING - 人間 - にんげん ningen HOWEVER separately they are read as にん nin and けん ken if you use that reading. Generally computers won't convert NINKEN to those Kanji, you have to type NINGEN because that is the way it sounds.

訓読み (kunyomi) JAPANESE READING - 守り神 - まもりがみ mamorigami, if you use these readings to read them separately you read them as まもり かみ MAMORI KAMI, when they occur together you read them as MAMORIGAMI.

More interesting to me is 人人 hitobito also written as 人々 hitobito because the Kanji is repeated, as when HITO occurs again after HITO (Japanese reading), the second HITO changes to BITO. If I am correct, the same thing occurs in ALTAIC languages, of which JAPANESE and KOREAN have been said by some to be a part of (you can see in non-Chinese vocabulary connections, generally Korean and Japanese are closest to Tungusitic of which Manchu is the largest language)
Sum Won

Re: Tone Sandhi

Post by Sum Won » Fri Jul 05, 2002 3:20 am

I'm more into ethnology, and have read some stuff on the origins of the Japanese. Though they still say their origins are pretty murky and mysterious, they also make note that many Koreans came over to Japan, not necessarily as the original ancestors of the Japanese, but during Korea's times of crises. So, I'm not sure if we can really mark the Japanese langauge under the Altaic-Tungusian category.
Posts: 134
Joined: Thu Feb 03, 2005 3:53 pm

Re: Tone Sandhi

Post by Mark » Fri Jul 05, 2002 7:01 am

Generally, that's where linguists put it. If I were you, I'd put some research where my mouth is.

After generally looking thru records of OLD JAPANESE (pre-Chinese contact, I forgot how they know that stuff but they do) and comparing with PROTO-TUNGUSITIC and PROTO-ALTAIC, as well as other ALTAIC PROTOFORMS, I generally find some sort of relation although sometimes it is only with the initial consonant.

I'm not sure about the IPA capabilities of everybody here, unfortunately.


ALTAIC PROTOFORM (reconstructed):
TURKIC PROTOFORM [no cognates for Turkic]
MONGOLIAN PROTOFORM (reconstructed):



KOREAN PROTOFORM (no Korean cognates)
*ku/ni/ (/ni/ is regarded as a suffix here)


*ky/r OR *ky/rw@/r
*k@t@ OR *k@t@-pa


That one's fairly confusing, however I think you should note that SLASHES as well as other marks indicate things that can/could've been used as suffixes/prefixes but eventually got added as part of the word.

Generally if you look further your evidence will prove more to you than this, however as if they are related it's a fairly distant relationship you will hardly come up with conclusive evidence. Look at similarities in phonological systems, as well as grammatical systems (generally Old Japanese fits the profile better than does Modern Japanese because Modern Japanese has loads of Chinese influence)
Sum Won

Re: Tone Sandhi

Post by Sum Won » Fri Jul 05, 2002 11:39 pm

I wouldn't put any research into my mouth, because that's not where research is supposed to go. Research itself is a process of looking for information. Sometimes, this information is unaccessible to people. Topic threads such as this one, was created to answering questions, and sometimes even asking new ones.

Keep that in mind...
Posts: 134
Joined: Thu Feb 03, 2005 3:53 pm

Re: Tone Sandhi

Post by Mark » Fri Jul 05, 2002 11:55 pm

LoL! I think you knew what I meant ;)

And thanks to a certain Sergei What's-his-name, many large etymological databases are searchable online, including DOC (dialects of china), the online version of HanYu FangYan CiHui.
Sum Won

Re: Tone Sandhi

Post by Sum Won » Sat Jul 06, 2002 1:03 am

Thank you for the link, and as the purpose of this thread was set up --the dissemination of information-- you should try reading this:
James Campbell

Re: Tone Sandhi

Post by James Campbell » Mon Jul 08, 2002 5:31 am

Sorry I haven't been able to view these postings or reply for a few days. But I'm back.

In response to ichi: well, let's not mess with your mess, since it doesn't really help with the matter. We're not wondering about Mandarin or what tone sandhi really is--I think we know that much already (hey! and words mean different things with different tones? You don't say.... and does your mom still burp you?). Oh, yeah, and I've got a side note too: I have a degree in linguistics and have been doing linguistic research for 10 years, so I don't know if that qualifies me as the so-called linguist that you're looking for.

Thanks to Mark for Hakka tone sandhi, however that doesn't help me understand anything about Cantonese tone sandhi. And as I said in a recent email, the consonant changes that occur in Fuzhou and in Japanese are completely different phenomena, unless you have a way to prove it to me. I wouldn't call what happens in Japanese morphophonology like what happens in Fuzhou and Korean. And as in your response to that email: yes, it looks like it makes it easier to pronounce--that's exactly what morphophonology is about.

The Altaic and Tungus protoforms, though interesting as they sound, neither help to answer my original question. (I think there's a sci.lang group for this). I appreciate your work at providing all that info. I'm familiar with it and have been using Starostin's website for years, though I find that a lot of the Chinese dialectal data is not as accurate as I wish it could be.

However you wrote that there is now an online version of Hanyu Fangyan Cihui, so I immediately went to go check for this update, but I could not find it. Do you mean the original dialect data that has been available there for years on individual characters (ie. Zihui)? If there is a Cihui database, please let me know.

And thank you Sum Won for your link. It is an interesting read (I'm not done with it yet). But I will keep looking for Cantonese tone sandhi.

Also, can anybody explain how to pronounce the 1st tones in Cantonese for phrases like Zhang Xian-sheng (am I repeating myself here). I thought there were several Cantonese speakers here who could help. What I would like to know is which words are spoken level and which ones are spoken falling.

Or let's make it more fun. How about this sentence:
張先生今天吃真多鮮冬瓜。 (Ok, I know you don't really use these words in Cantonese--but maybe there's a similar sentence you can up with and describe the use of level and falling first tone in Cantonese?) Thanks.

Thanks if you can help

Posts: 134
Joined: Thu Feb 03, 2005 3:53 pm

Re: Tone Sandhi

Post by Mark » Mon Jul 08, 2002 10:25 pm

Generally I was quite surprised at what Ichi said as well, however I don't think that he/she was aware of the level you expected a responce on, perhaps he/she didn't understand what you said so answered it in the wrong way.

Sorry about the Hakka tone sandhi-- I guess it wasn't really a good idea to post that here as it wasn't a responce to your question, but I already posted it (and for what reason? lol)

As for morphophonology in Fuzhou, Japanese, and Korean-- generally this wasn't in responce to your post but in responce to Ichi's, so I responded to him on his own level, indicating one thing-- that MORPHOPHONOLOGY (not including tone sandhi, this is something that happens even in Mandarin to some extent) occurs on a very grand scale in both those languages, but not that the systems of the two were related in any way.

Just curious about Fuzhou, but does this have to do with where a character is in a combination?

And what I said in the e-mail about making pronunciation easier-- I was saying that this is what makes morphophonology in Japanese, Fuzhou, and Korean (and any other language for that matter) similar. Although beyond that it isn't all that similar.

Fuzhou morphophonology DOES seem to affect some Chinese loanwords in Japanese, although I can't think of any examples at the moment.

As for the protoforms-- this was just for Sum Won to show why I thought Japanese was an Altaic language.

And about HanYu FangYan CiHui-- yes, that was a typo, I did mean ZiHui (derned keyboard ;-<)

Generally, this information isn't very very accurate, but it's there... (considering for some of these dialects, unfortunately, that's the only information available to me locally, perhaps I'll request an interlibrary loan sometime in the future) Haven't had time to check it out yet, but somebody told me about

For the rest of your post-- I can't help you, but I can ask some people I know (though as they don't know much about linguistics, but rather are native Cantonese speakers, I might not get as much info as you'd like: they aren't exactly that aware of the nuances of their language)

Re: Tone Sandhi

Post by ichi » Mon Jul 08, 2002 10:43 pm

James --

I had no way of knowing you were a linguist. Also, I now realize that I misread your question (in asking for dialects of Cantonese, as opposed to Chinese.). Instead of jumping on my back, maybe you could let it slide? Anyways.
Posts: 43
Joined: Thu Feb 03, 2005 3:53 pm

Re: Tone Sandhi

Post by Helmut » Mon Jul 08, 2002 11:19 pm

Tone Sandhi:

Quoting from "Cantonese - A Comprehensive Grammar" (Matthews/Yip):
". . . In Cantonese, tone change is restricted to one main process which is not regular in this sense, but occurs due to a number of morphological and semantic factors. For these reasons, the Cantonese case is generally referred to as tone change rather than tone sandhi. The functions of this tone change are highly complex."
This is followed by a four page quick overview of the most important cases. (Don't ask me to type that in for you, please.)

There is all sort of such tone changes in Cantonese. The most frequent is in my experience the change from tones 4 (low falling) or 6 (low flat) to tone 2 (high rising), but there are all sorts of other types. Most important: there are no easy rules to apply. Learners of Cantonese usually learn the word combinations rather than any rule.

Mr.Zhang is Zoeng1 Sin1 Saang1. In general, the tone number one can be either high flat or high falling, depending on the word. In Hongkong today, this difference has all but disappeared. It is always pronounced high flat. This has nothing to do with tone sandhi, i.e. the same word does not change from high flat to high falling depending on the sentence. Instead, there is a shift in popular speech patterns such that the high falling tone is not used anymore. Words that had been pronounced with high falling tone before are nowadays generally pronounced high flat (with few exceptions). Many text books still quote high level and high falling. However, they often do not agree on when to use high falling and when to use high flat. I found 4 different results in 5 text books for Zoeng1 Sin1 Saang1.
Posts: 134
Joined: Thu Feb 03, 2005 3:53 pm

Re: Tone Sandhi

Post by Mark » Tue Jul 09, 2002 4:30 am

As for Xiansheng (for why I'm saying it, tone isn't extremely important), seeing the Cantonese equavalent (Sinsaang) made me think-- that's like Japanese in that in Japanese you say Sensei.

About the Matthews/Yip book: I actually had that checked out from my local library recently, I might actually still have it. However I haven't yet looked at it much. If you want, when I get my computer-with-scanner working, I might OCR (is that a verb? couldn't OCR also stand for optical-characterally recognise or something like that? lol) the 4 pages of which you are talking for the benefit of James (if he can't find it) and those others who don't have access to the book.
James Campbell

Re: Tone Sandhi

Post by James Campbell » Tue Jul 09, 2002 6:25 pm

Thanks for everybody's input.

There was a good book written about morphophonology specifically in Korean, I think the author was Samuel Martin. Very detailed and good read but getting to be dated.

Thank you Helmut for your info. Those books are readily available, but I never took a close look at their contents. I think there are two books now: a beginner's and intermediate one (pink cover?). Do you know which one it is mentioned in and the page numbers? Sounds like you've got the book with you, if you don't mind letting me know. In fact, mentioning tone change in contrast to tone sandhi as being a separate phenomenon really interests me. In some languages and dialects (I think Mike has already mentioned this) some tones undergo sandhi into contours that are not shared by another tone, so one cannot say this tone becomes that tone, instead a whole new tone is created. Much of Shanghaiese is this way.

But I'm sure what happens in Wu would be completely different than what you are describing for Yue dialects. I wonder if the various kinds of tone sandhi themselves can be differentiated into categories. The tone sandhi in some dialects like Ningbo Wu are quite extreme (perhaps more than a hundred different combinations?), but it goes to an even more extreme in examples like in the dialects of Jin, notably Pingyao (central Shanxi province). Taiyuan Jin also has this but not to the same degree. There are perhaps hundreds of different tone contour combinations in these Jin dialects.

In other cases, like standard S. Min (Xiamen), I would consider in the same class as that of Mandarin. I like to think of these languages as having an accent as well (it's not real accent but some kind of strong accent-like attribute that syllables can have). Maybe I use accent for a lack of a better term. In these languages, preceding vowels have a weak accent and therefore undergo sandhi, and the final vowel has a strong accent and is not affected. In Wu, I think of the accent falling on the first syllable as this is what affects the tone sandhi of the whole phrase even though the first syllable undergoes tone sandhi as well. But there is also some kind of real accent that is revealed over the length of several syllables (rising-falling), especially in 3, 4 and 5 syllable phrases.

Helmut states that there are no easy rules to apply, which means that there must be rules: they're just complex! I thought the same about Shanghaiese, just as many other Shanghaiese speakers feel (they state that there are no tones!), until I saw a chart, which just confused me even more. But after learning how to speak the language more and reading up as much as I could about it, I learned that there really were patterns that I could follow and I could finally decipher the charts. This improved my ability to be understood more. And in Shanghaiese, I don't really feel it's the tones that are important, but the voicings of consonants that are important (vd/unv). Once I understood that, all the tones fell into place. So it's learning how to look at it from a completely different perspective--and it was much easier to explain.

The Pingyao tone sandhi charts definitely are an intimidating sight. But I'm sure there's a hidden key beneath it all--or else how would anybody be able to speak it naturally?

Now back to Cantonese. If this is really a case of haphazard tone change rather than tone sandhi, my belief is that it may be something that we already know about in another dialect somewhere; some kind of special tone sandhi case like I just mentioned. Otherwise let's get a study done on it and figure out where these tones are going.

And it seems that Cantonese has been undergoing a period of big change during the last few decades. A lot of sound shifts happening and even tone shifts. I never realized that tones 4 and 6 could change to 2nd tone--it feels quite unnatural to me. I never even saw this mentioned in any book either.

I just remembered: I recently bought 第七屆國際粵方言研討會論文集 which has 500 pages of in-depth Yue research (published 2000.12), so I'll search through it for any mention of tone research. Yes, I just found that page 172 has an article by 林建平 titled 香港粵語陰平調值商榷, so I'll read through it and get back to you. But I can't find anything else in this book about Cantonese tones or tone changes like what Helmut was mentioning.

Ichi: sorry for getting on your back. I was just trying to start a more intelligent linguistic discussion, and then you come and state the obvious and even some obviously incorrect information ("3rd followed by 4th"--at least I never caught myself speaking that way), so what do you expect me to say? "Tone sandhi is documented in Mandarin; I believe a textbook..." that's just asking for it. But I'll drop it, ok?

James Campbell
Thomas Chan

Re: Tone Sandhi

Post by Thomas Chan » Tue Jul 09, 2002 9:58 pm

James Campbell wrote:
> Thank you Helmut for your info. Those books are readily
> available, but I never took a close look at their contents. I
> think there are two books now: a beginner's and intermediate
> one (pink cover?). Do you know which one it is mentioned in
> and the page numbers? Sounds like you've got the book with
> you, if you don't mind letting me know. In fact, mentioning

It sounds like section 1.4.2 "Tone change" in Stephen Matthews
and Virginia Yip's _Cantonese: A Comprehensive Grammar_
(New York: Routledge, 1994). The "Basic" and "Intermediate"
ones are textbooks rather than grammars, I believe.

However, phonetic and phonological information is scant in such
a work. You can find a lot more in sections 2.5 "Tone change"
and 2.8.1 "Tone sandhi" of [Anne] Oi-kan Yue[-]Hashimoto's
_Studies in Yue Dialects 1: Phonology of Cantonese_ (Cambridge:
Cambridge University Presss, 1972) and the appropriate sections
of Robert S. Bauer and Paul K. Benedict's _Modern Cantonese
Phonology_ (New York: Mouton de Gruyter, 1997).

Yue (1972: 112) only gives the following for sandhi:
53 -> 55 / __ 53/55/5
21 -> 22 / __ 21/22
(i.e., the 53 variant of yinping becomes the 55 variant when
preceded by 53, 55, or 5 yinru; 21 yangping becomes 22 yangqu
when preceded by 21 or 22--I know you can read the notation,
but I just wanted to apply the traditional names to the tone

Yue goes on to explain (the section is only two pages total)
that it can be chained, e.g., 21 21 21 21 -> 22 22 22 21; that
the 21 -> 22 change is seldom discussed compared to the
other change because it " seldom noticed partly because
the pitch level of that tone is so low that that the distinction
between a level and a falling variety often escapes the ear,
and partly because many speakers probably do not pronounce
the tone normally with a falling contour..." (113), etc. I'm sure
there's more information in the Bauer and Benedict book, but I
don't own a copy to quote from at the moment. (Yue, on the
other hand, is out-of-print and the print run was low--it was
expensive to produce--let me know if you need more information,
examples from the text, footnotes, etc.)

I believe with your "Mr. Zhang' question, you were looking to the
answer that is the first of Yue's rules, but without an actual
analysis, I believe I only use the 55 form of the yinping tone, so
I'd say 55 55 55, or each syllable separately as 55, 55, 55. I seem
to recall the Bauer and Benedict book described the the 55 and 53
forms of yinping as the different contours that HK and Guangzhou
have taken on for that tone.

> Helmut states that there are no easy rules to apply, which
> means that there must be rules: they're just complex! I
> thought the same about Shanghaiese, just as many other

The rules for the sandhi (see Yue above) motivated by ease of
pronunciation are easy, as the rules for tone change can be
described, but when to apply the latter has to be learned lexically,
although it can be generalized. e.g., for many, the changed tone
conveys familiarity or diminutiveness, and some can be optionally
changed by speakers such as jung55man21 vs. jung55man21-35
'Chinese language' while others are obligatory because the changed
form has a different meaning like tong21 'sugar' vs. tong21-35
'candy'. The entire process is about as predictable as the application
of the neutral tone of Mandarin (but if you've figured out any rules
to make the latter easier, I'd like to know!). Yue's section is about eight pages in length.

> And it seems that Cantonese has been undergoing a period of
> big change during the last few decades. A lot of sound shifts
> happening and even tone shifts. I never realized that tones 4
> and 6 could change to 2nd tone--it feels quite unnatural to
> me. I never even saw this mentioned in any book either.

It feels pretty natural to me. :) BTW, the citation form of my
surname is chan21 (Mandarin chen2), but for familiarity when
prefixed by "Ah", you get the form a33 chan21-35.

I believe this has been going on for a long time, but I do recall
seeing some articles on how items have gained or lost changed
tones over time.

Books not mentioning it might be the result of listing the changed
tone alone and not mentioning the base tone (in some cases,
the changed tone has overwhelmed the base tone and speakers no
longer always know what the base tone originally was, particularly
in the colloquial strata where reference to rhymebooks and other
dialects cannot always be made).

Thomas Chan