Re: tonal languages

On Wed, Jan 08, 2003 at 11:17:25AM +0800, Florian Rivoal wrote:
> Yesterday, i bought a book on shanghainese language. Living in shanghai,
> i thought i could be a could idea to take some of the local language in
> addition the the widely spoken "common language" that mandarin is.
> shanghainese is generaly called a dialect of mandarin, but it is wrong.
> it is a dialect of the wu language, which is one of the many languages
> spoken in china, together with mandarin, and cantonese...
Well, the so-called "Chinese dialects" are properly languages in their own
right. They are grammatically different (although similar), exhibit
different (though similar) vocabulary, and are mutually unintelligible
(except for the writing system). Even Spanish and Portuguese are closer to
each other than two of these so-called Chinese "dialects".

> As a language learner, i was horified by this book, but as a linguist, i
> found it pretty interesting. If anyone here is into tones, I guess you
> may be interested by the following.
Ahh, to be so fortunate as to grow up with a tonal L1. :-)

[snip]

> Tones is about changing the height (and possibly the lenght) of a
> syllable. kind of singing, if you want.
More like a sophisticated pitch-accent than singing, IMHO. When singing in
a Chinese language (or any tonal language for that matter), you have the
very interesting phenomenon that tones are lost and replaced by musical
pitch, often leading to ambiguities that can only be resolved through
context or reading the written lyrics.

> i will note tones using a 5 degree scale. 1 is low, 5 is high.  The
> tones in mandarin are phonemic, and linked to morphems just like vowels
> and consonants. It counts 4 tones plus a neutral tone.  Long 44, short
> 14, long 213, very short 41. neutral tone is a not accentuated one used
> in quite a lot of compounds.  To make it more intersting, there are some
> rules of sound change when the syllabs are not isolated.
That is called "tone sandhi". Most of the Sinitic languages (the "Chinese
dialects") have this phenomenon. If tones weren't enough to scare the wits
out of non-native speakers, tone sandhi will definitely drive you crazy.
:-P (Douglas Koller can tell you stories about this, I'm sure.)

> The number 3 looses the raising part (thus becoming 21)  when it
> prececeds any other tone than 3 or neutral. before an other 3, it
> changes to tone 2. More over, there are some exceptions. for exemple,
> the word "bu4" (marking negation) gets second tone before any other 4th
> tone.  On the first look, it can seem troubling, but actualy, this aint
> so complicated.
When analysed, sandhi rules look horrendously complicated; but actually
they are just the tonal equivalents of the euphony in ancient Greek.
Certain tones don't sound well together (at least to the native speaker);
so when they must appear next to each other, one or both must change to
preserve tonal euphony.

> so here are the basic tones of shanghainese
> 1st : long 53
> 2nd : long 34
> 3rd : long 13
> 4th : short 55
> 5th : short 22
My L1, Hokkien, has 7 tones, which are basically 4 basic tones plus
"short" (or "clipped") versions of three of the basic tones.
Unfortunately, being a native speaker, I have great trouble detecting
which tones are long and which are short--it's just a spontaneous,
unconscious thing to me. Douglas Koller, who is familiar with Hokkien /
Taiwanese, would be able to give you the full descriptions of these 7
tones.

> the entering word of a compound can get a final glotal stop(as explained
> later) noted "?"
I conjecture this must happen when the word ends with a "short" tone?

> the initials of the first tone (53) and 2nd (34) are voiceless sounds
> the initials of the 3rd tone (13) are all voiced sound exept nasals
> sounds and the lateral "l".
Interesting. Hokkien probably also has similar constraints, but I can't
tell 'cos it's completely unconcious for me. :-)

> The tone will change when two or more words are pronouced in series. The
> tone change in serial pronounciation can be divided in two groups. the
> extensively used form, and the infrequently used form.
>
> The characteristics of the exensively used form are: in a serialy read
> word group, the first word is generaly not prounouced in original tone,
> but in a fixed tone. for example: the tone 53 is transformed into 44, 34
> into 44, 13 into 33, 55 into 44 and 22 into 11. the tone change of the
> first word affects the tone of the following words which also lose
> original tone, in despite of rising and falling tone. the fixed form of
> the tone change is formed by combining the first word with the following
> words. according to the various tones of the first word, the tones for
> two-word group, three-word group, .., five word group tend to be
> changed.  general rule are shown in the following table:
[snip]

Interesting. This is probably caused by the vocal apparatus taking the
shortest path between two tonal peaks.

[snip]
> ok, here you are. above is roughtly an cut from my book, and since i
> haven't started learning, i guess that i wont be able to explain much
> more that what i just wrote. i just wanted to show you this system,
> cause i think it is shows that natlangs are quite good at doing
> complicated things when they want. Does anyone has a conlang (or
> natlang)  with tonal rules as twisted as this?
I suspect my L1, Hokkien, is a LOT more twisted than this. Unfortunately,
being mostly unconcious of tone changes in my own speech, I'm probably
unaware of most of it. Also, my dialect of Hokkien has diverged from
mainland Hokkien and Taiwanese; the original high-falling (51) tone has
mutated into a high-rising tone (35), and a lot of borrowed words have
crept in from local Malaysian lingos. There are probably lots of weird
rules that it has picked up along the way that I'm not aware of. :-)

But coming back to what I *am* aware of... some examples of tone sandhi in
my Hokkien dialect include the following: (sandhi'd words marked with '*')

1) be21 sai35 ("not allowed to") + a33 ne33 kuan35 ("in this manner")
   --> be21 *sai33 a33 ne33 kuan35 ("should not be like this")

   [On a related note, when be21 sai33 occurs in isolation, it means
   "to sell [a] tiger". Just think of the potential embarrassment for
   a non-native speaker who can't get the sandhis right.]

2) be21 seng13 ("cannot be accomplished")
   --> be21 *seng21 kong13 ("will not succeed").

3) tso21 ("to do")
   --> *tso33 kang33 ("to do work")
   --> *tso33 siao35 ("to make fun")
       --> *tso33 *siao33 su21 ("to be involved in funny business")
           (notice how both initial words have been sandhi'd)
   --> *tso33 lang13 ("be a person")
   --> *tso33 bang35 ("be a mosquito")

   It may seem as if tso21 consistently becomes tso33 when it occurs as
   the first word in a compound; but this isn't always the case:
   --> *tso21 liao35 ("done")

   What's worse is, both bang35 and liao35 have the same tone, but when
   compounded with tso21, tso21 behaves differently. And to throw a few
   more monkey wrenches into the mix:
   --> *tso33 ho35 ("to do good")
   --> *tso21 ho35 liau35 ("to have completed")

   So apparently the sandhi can be caused by words not even adjacent to
   the sandhi'd word. :-)

Of course, there are probably many, many more examples. Maybe Douglas
Koller has a few words to say about why the sandhi rules are this bizarre.
:-)


T

--
Indifference will certainly be the downfall of mankind, but who cares? --
Miquel van Smoorenburg