Re: Hiatus within words

H.S.Teoh wrote:

>On Mon, Oct 30, 2000 at 08:38:12PM -0500, Roger Mills wrote:
>[snip]
>>         I can't think of with a single form in my (fairly standard
Midwest)
>> Engl. with [...aV...] except furrin loans like "paella>
[snip]
>
>Interesting. I observed English L2 speakers saying "idear" instead of
>"idea" too.>
In this country? or in Malaysia?  Over there, I'd imagine it's a relic of
British English.

Further to my idea that Engl. forms with /...aV.../, like "paella", are
foreign-- it strikes me that all Engl. words with final -a (/@/) are
probably foreign too-- sofa, paella, pizza, diva, mantra, panda, idea,
etc. etc.  We can form -ish and -y derivatives of these-- "pizza-y taste",
"diva-ish behavior" etc.  I find I use the "voiced h" in these, or more
rarely a very weak glottal stop.
>[snip]

>Yes, [dwa] occurs in fast speech. Interestingly, [duwa] is the most common
>pronunciation of "dua"; I'd say [du?a] never occurs except in poetry where
>syllable breaks are emphasized.>
Agreed.
Of interest to all inquiring minds:  The native Buginese script writes "w"
in "..u/o_V..".  It uses the same character for the initial of e.g. w@rr@?
'rice'', wai 'water', etc., also medially in -awa-, -iwi-, -iw@- etc.  As
far as I could tell, this /w/ is [w] in all cases, likewise in Matthes'
1850s description.  BUT:  Bug. /w/ has three sources-- 1) automatic after
rounded vowels 2) AN *w, as in 'water', and 3) AN *b, as in 'rice' .  There
is a dialect of Bug. where (1) and (2) can be pronounced [h], but (3)
remains [w].   There are dialects of closely-related Sa'dan Toraja with the
same rule; yet other dialects retain [w] in (1) and (3), but change it to
[h] only in (2).  Clearly, at some point the /w/ < *b and the /w/ from *w
had to be pronounced differently (and automatic w isn't there at all,
phonologically).

Many Indonesian languages have /w/ < both *b and *w; but any instance of
modern /w/ preceding /u/ or the local reflex of *@ can be traced back to *b
with confidence-- so far the sequences *wu and *w@ are not reconstructible
for AN.  Kash will prove to be quite similar.  Any such restrictions in
other conlangs?