Re: Lots of Questions About Tones (more questions)

On Sat, 12 Jul 2008 13:24:45 -0400, Eldin Raigmore
<eldin_raigmore@...> wrote:

>On Thu, 10 Jul 2008 14:38:36 -0400, Alex Fink <000024@...> wrote:
>
>>Can there, or will there, be default numbered pitch levels for the tones in
>>a contour(s1) tone language?
>
>I think that will depend on the IPA and maybe on the linguist discussing the
>language.
>The IPA has
>1 eXtra Low
>2 Low
>3 Medium
>4 High
>5 eXtra High
[...]
>So maybe you want to use {1, 2, 3, 4} or {2, 3, 4, 5} for a language with four
>relevant pitches, but no relevance to "absolute pitch".
Right, wouldn't be the first time the set of distinctions the IPA provided
weren't the most convenient for the analysis of some particular language.

>>Or are the realisations of the tones supposed to be completely in free
>>variation?
>
>I'd imagine they're in complementary distribution and are conditioned, at least
>partially, by other factors, including perhaps both tonal sandhi and co-
>occurring segmental features.  I'm not sure that counts as "completely free
>variation"; does it?
If it's indeed conditioned, then no.

>The point is they're "allotones", or at least that's what I expect and
>understand.
Right.  What I was driving towards was asking what sort of schemes of
conditioning factors for allotony actually turn up in contour(s1) langs, and
in particular whether there can be said to be underlying pitch levels at
which each of the tonemes is realised, say when words occur in isolation.

Do you have any examples to hand of (claimed) contour(s1) languages and
their systems of allotony?

>>If you're just reporting the results of a discrimination task on single
>>words each from different speakers presented in isolation, I could see this
>>even for speakers of a language without a contour(s1) tone system.  Surely
>>register(s1) tonality doesn't imply perfect pitch.
>
>I'm not sure what you said.
>I doubt I'd disagree if I understood perfectly, but I understand only
partially.
>Could you explain, please?
Suppose we speak a language in which, say, /ma5/ and /ma3/ are
distinguished.  I wouldn't expect the tonemes /5/ and /3/ to correspond to
fixed pitches, but to respectively high and neutral points within a given
speaker's range, given that different speakers have different vocal ranges
(and this is the case even if any single speaker's deployment of the tones
uses absolutely constant pitches).  So if you came up to me and intoned /ma/
and nothing else, I wouldn't know whether that was /ma5/ and your neutral
pitch was lower than that utterance, or /ma3/ and your neutral pitch matched
it.  But it would be unwarranted to attribute to our language a contour(s1)
system on the strength of this alone.

Perfect pitch would be, in a weak sense, helpful in the understanding of a
tone language in which tonemes did in fact have constant speaker-invariant
frequencies.  At least, in this case, it would allow me to figure out which
word your /ma/ was.

Now, to introduce a question of my own, it seems to me that the actual
benefits of perfect(s1) pitch in understanding a register tone language
would be extremely marginal -- it's only necessary to distinguish between a
family of utterances whose tone-patterns are all transpositions, in the
musical sense, of the same melody.  For any utterance that's not extremely
short, most or all of these transpositions would contain a tone outside the
tonal range of our language, and of the ones that don't presumably all but
one would be quite likely to be lexically or gramatically or semantically
incoherent.  In a contour(s1) pitch language, the benefits of perfect pitch
would be precisely nil (only interval discrimination would be required).

Nonetheless, one hears claims -- such as the one at
  http://www.universityofcalifornia.edu/news/article/6750
-- that speakers of Mandarin are "almost nine times more likely" to have
perfect pitch than speakers of unspecified non-tonal language.  This
surprises me, given what I speculated above about vocal range, and
especially given that Mandarin seems one of the more likely langs to fit
into the contour(s1) framework: of its four tones that can appear on strong
syllables, if one regards tone 3 as dipping, none is a transposition of any
other.  So what's going on here?  Is this correlation true?  Is my
speculation misled?  I've not heard analogous claims about tonal languages
outside of East Asia, in particular IE langs; is that 'cause they don't hold
there, or just 'cause they haven't been investigated?

>And, by the way:
>
>Thanks!
No problem.