Marcus Smith wrote:

> Tom Wier wrote:
> >
> > > On Sun, 12 Aug 2001, Thomas R. Wier wrote:
> > >
> > > > The lengthening here is very normal when there is loss of some
> > > > kind.   However, geminate consonants are basically defined by
> > > > a break in syllables lying between them.
> > >
> > > This is not true. There are languages in this world that have word final
> > > geminates, which hardly spans a syllable break.
> >
> >You know, I seem to remember reading about this too, but then my
> >phonology professor said something to the effect that I mentioned
> >above;  perhaps I misunderstood or misheard her.
> Or perhaps she used the distinction that someone on this list made (Nik?)
> between geminates and long consonants.

Nope, but she has some interesting things to say.  I emailed her afterwards,
and here's what she had to say (I also mentioned Nik's distinction between
long and geminate C's):

In languages where there is a contrast between long and short consonants, (i.e.
you have [bota] and [botta] and they are also semantically distinct), the
difference is taken to be phonemic, i.e., is generally taken to be represented at
input/underlying representation.  The claim (i.e. Hayes 1989 and others, I think
also Hyman 1985, cited in Hayes) has been that the geminate consonant is
represented with some unit, for example a mora, in UR that the nongeminate doesn't
have at that level.  (Note that geminates don't always have to be singled out in
UR: just like a vowel can be systematically lengthened in a language in a specific
context, so can a consonant.  Thus, it isn't always the case that if a language
has both geminate and nongeminate Cs, that the difference is phonemic.)  Because a
moraic coda consonant is usually made "long" by linking it up to a following
syllable as an onset, most geminates discussed in the literature span a syllable
boundary on the surface.  But, where a long consonant is created by something
other than onset formation, you may have a long consonant/geminate that doesn't
span a syllable boundary.  Luganda (a Bantu language) even has word initial

I don't make a distinction between long consonants and geminate consonants the way
you seem to be making; for autosegmental phonology, the distinction would be
whether the consonant is multiply linked or not--what most phonologists call a
gemninate is long because it is linked to more than one prosodic unit.  Some
phonologists might insist that a geminate be linked to at least one mora.  To call
a consonant "long" implies the same.  (By contrast, a nongeminate, or fake
geminate, would be two identical consonants linked to different prosodic units,
superficially resembling a geminate, but with a different representational

So, could those Pima [tt] s be geminates?  Maybe.  I'd want to know more about the
prosodic structure of Pima before saying yes or no.  But it could well be.

There's a nice paper by bruce Hayes called, I think "Geminate Inalterability..."
that appeared in _Language_  (I think) in 1986.

So, in other words, she seems to deny my original claim -- that geminates are
simply *defined* by a break in syllables.  On the other hand, she does say that
"what most phonologists call a geminate is long because it is linked to more
than one prosodic unit", and of course, a syllable is one type of prosodic unit,
the one mostly commonly associated with gemination.

> I find that a very arbitrary and useless distinction, but some distinctions
> are that way.

She doesn't make it sound as very arbitrary as you say it would be -- it depends
on whether you insist that geminates are linked up to a mora or not, and the
way she puts it, informed phonologists can disagree on this matter.

Thomas Wier | AIM: trwier

"Aspidi men Saiôn tis agalletai, hên para thamnôi
  entos amômêton kallipon ouk ethelôn;
autos d' exephugon thanatou telos: aspis ekeinê
  erretô; exautês ktêsomai ou kakiô" - Arkhilokhos