THEORY: ambisyllabicity & gemmination (long)

On Sat, 7 Oct 2000, Raymond Brown wrote:

> At 11:14 pm -0600 5/10/00, dirk elzinga wrote:
> >(Some have even argued that ambisyllabic
> >consonants are really geminates in disguise since they also share the
> >property of belonging to two syllables at once; interesting that in
> >English many of these are written with two consonant letters, as in
> >the word 'happy'.)
>
> Right, I'd like to come back to that last point - not to be contentious,
> but because I genuinely seek enlightenment.
>
> English ambisyllabic consonants masy be "geminates in disguise", whatever
> that means, but they most certainly are not geminates.  The /p/ in _happy_
> /'h@pi/ is one of these so-called ambisyllabic consonants.  It is very
> different from the /p/ in Welsh _hapus_ (happy) /'hap1s/ where /p/ is
> pronounced [pp_h] and gemination is as clearly marked as it is in, e.g.
> Italian _cappa_ (cape, cloak).
Is _hapus_ the spelling of the word in Welsh? If so, how is gemination
marked? In Italian, gemination is marked in the orthography by
doubling the letter, but I don't see this doubling in the Welsh word.
Or are you using "marked" in the Praguean sense?

> Ignoring the question of morae, and just thinking in terms of syllabic
> onset, nucleus and coda, presumably Dirk's analysis above would mean that
> the Welsh word would be represented thus:
>          s     s
>         /|\   /|\
>        o n c o n c
>        | | |/  | |
>        h a p   1 s
>
> Does that mean that the English word is:
>          s     s
>         /|\   /|\
>        o n c o n c
>        | |  \| | |
>        h @   p i 0   [0 = zero element"]   ?
I'm not sure I understand the significance of the representational
distinctions you make here between

        c o       c o
        |/   and   \| ; they are the same.
        p           p

Also, I don't usually represent empty syllable positions. (Government
Phonology has a well-articulated theory of empty syllabic positions,
but I'm not familiar enough with it to say anymore about the idea.)
Otherwise, I think I agree with the representations.

> Before I'd read Dirk's letter I'd have shown the Welsh as:
>          s     s
>         /|\   /|\
>        o n c o n c
>        | | | | | |
>        h a p p 1 s
In a theory which does not recognize moras, I would insert a "timing
tier". The timing tier is a level which represents segmental "place-
holders". So your Welsh representation would be minimally altered to
the following:

          s       s
         /|\     /|\
        o n c   o n c
        | | |   | | |
        x x x   x x x
        | |  \ /  | |
        h a   p   1 s

Gemination is shown in such a theory by doubly linking a single
feature bundle (labial, voiceless, stop; here represented by /p/)
with two timing units.

> Also, is the /p/ in English _happy_ really ambisyllabic?
Well, this is one of the Big Questions of English phonology. I'm
inclined to think that /p/ isn't ambisyllabic, and that there is no
such thing as genuine ambisyllabicity. More on why below. (Of course,
I've been looking at Shoshoni phonology for so long that
ambisyllabicity just looks weird. :-)

> The argument, as
> I understand it, is that the lax vowels (/@/, /E/, /I/, /O/ and /U/) never
> occur in word final position, therefore they do not occur in syllable final
> position.
Yes, this is the argument.

> But this is not entirely the case. In the above I denoted _happy_ as
> /'h@pi/ because that's the way it's usually shown on this list and, as a
> southerner, that's the way I pronounce it.  But over much of northern
> England one hears ['hapI]. The network manager at my College is a young
> lady from Manchester and she invariably pronounces the final -y in sych
> words as [I].  In some non-rhotic varieties of English final -er is
> pronounced [@].  I've heard, e.g. _father_ pronounced ['fAv@] in certain
> London dialects.  When we were school kids we had no difficulty in
> reciting the ancient Greek definite article as [hO], [hej], [tO] where [O]
> is the British [O] in _not_, not the American [Q] or [A].  I _know_ that's
> not the correct ancient Athenian pronunciation - but that's not the point
> here - the point is that we had no problem pronouncing short, lax [O] in
> word final position.
>
> In other words I do not find the argumernt that the lax vowels /@/, /E/,
> /I/, /O/ and /U/ cannot occur in English except in blocked syllables well
> founded.  Therefore, I am not convinced (yet) by the ambisyllabic argument.
> We did discuss this once before on this list.  Certainly I cannot accept
> the syllablic division [h@p.i] as some suggested.
The felicity of the argument will vary then according to regional
variety. While that argument for ambisyllabicity may not be convincing
for those varieties which allow lax vowels in open syllables, it is
weightier for other dialects in which lax vowels are prohibited in
such syllables. Of course, this leaves the ambisyllabicity question
open for the Northern dialects.

In his book, _The Phonology of English_, Mike Hammond marshalls a vast
array of distributional evidence to argue for the representation which
you reject; viz. [h@p.i]. I find his arguments compelling. However, he
explicitly states that his analysis of English is based on his own
variety of English--he was born and raised in Southern California.
While that particular variety does not vary significantly from General
American, it will differ in important ways from Insular varieties.

> I am open to persuasion over the ambisyllabic argument, but personally I am
> still inclined to: [h@.pi] (or [ha.pI]), i.e.
>          s     s
>         /|\   /|\
>        o n c o n c
>        | | | | | |
>        h @ 0 p i 0
Again, I'm not sure that I would include "zero elements" in the
syllabic representation, but this would seem to be a sensible solution
*for the northern dialects you cite*, not necessarily for RP or GA. I
personally don't believe that these medial consonants are
ambisyllabic but that they should be affiliated to the syllable on the
left.

> I would dearly have liked to hear what Mark Line would have to say on these
> matters also.  Is anyone still in contact with him?
Yes; I would also be interested. Knowing his theoretical
predilections, I would suspect that his analysis would resemble in
broad outlines (if not in particulars) that given by Hammond. I say
that only because Mark seemed to rely heavily on distributional data,
which is also a notable feature of Hammond's book.

Dirk

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