Re: `Akupa, was: 5 phonemes

On Sat, 22 Jul 2000, Danny Wier wrote:

> --- Vasiliy Chernov <bc_@...> wrote:
>
> > No, I wasn't satisfied with Miituu. Danny was right saying that
> > 'true'
> > geminates could be counted as phonemes. And the other complications
> > with syllabication can be suggestive of something like deep-level
> > glottal stop.
>
> Your task is daunting.  I couldn't come up with anything that detailed.
>  Maybe it's because I've studied Farsi and a little Arabic, but I see
> syllables as being C, CV or CVC -- V and VC just begin with a "null
> consonant", which could be the glottal stop (lenited to a mere hiatus),
> or a voiced or voiceless [h].  All utterances must have a beginning and
> an end.
>
> And syllabries do have V and/or VC representations.  Just add one to
> the number of consonants.
>
> Gemination need not double the number of consonants, if you treat a
> geminate CC as being XC, where X is a "dummy consonant" that
> assimilates to the following consonant.
and later he said:

> So a "deep" phonlogy needs only to use an existing consonant, which
> automatically has to cause gemination, or just add a wildcard phoneme
> which has no value of its own but causes gemination.  (Or better yet,
> use the glottal stop without a following vowel, since you don't have
> preglottalized or implosive consonants.)
Exactly. In current phonological theory, gemination is handled by
having an extra slot for some segment with a rule saying that extra
slots need to be filled by pronouncable material. This slot could be a
mora, root node, or a spot on the "timing tier" (either a C, V, or X).
The choice between these options depends on the particular brand of
phonology and the language under investigation. For instance, I use
moraic representation in my analysis of Shoshoni rather than strings
of Cs and Vs or Xs (since stress is determined partly by syllable
weight), but since geminates don't contribute to syllable weight in
that language, the geminates are represented by features linked to two
root nodes:

        [moppo] 'mosquito'

             $         $   syllabic tier
            /|        /|
           / M       / M   moraic tier
          /  |\     /  |
         R   R R   R   R   root node tier
         |   |  \ /    |
        [m   o   p     o]  features

In a language in which coda consonants contribute to syllable weight,
the first half of the geminate is linked to a mora as well (and there
need be only one root node for the geminate). Thus:

        [moppo] 'mosquito'

             $       $   syllabic tier
            /|\     /|
           / M M   / M   moraic tier
          /  |  \ /  |
         R   R   R   R   root node tier
         |   |   |   |
        [m   o   p   o]  features

Personally, I find the structure with moraic geminates to be more
elegant, but Shoshoni is just not that way.

So there could very well be a language which interprets empty root
nodes or moras as glottal stops; this would be the situation discussed
by Vasiliy and Danny.

Dirk

--
Dirk Elzinga
dirk.elzinga@m.cc.utah.edu