Re: Partially-heard words. (or phrases?)

On Sun, 13 Jul 2008 15:43:37 -0400, Eldin Raigmore
<eldin_raigmore@...> wrote:

>Are there any natlangs in which, by hearing the last part of a word, you can
>get an idea how long the first (unheard) part was?
>
>Are there any natlangs in which, by hearing the first part of a word, you can
>get an idea how long the last (unheard) part was?
>
>(If "word" isn't the appropriate concept, substitute "phrase".
>Or, substitute "morpheme" or "syllable" or whatever applies in the natlang in
>question.)
>
>Same two questions for conlangs.
Had you not mentioned your inspiration I would've said I'd be floored to
find anything like this in a natlang.  What came first to mind was the
self-segregating morphology method exploited in Plan B and more recently
Joerg's X-1, number 4 on the list at
  http://conlang.wikia.com/wiki/List_of_self-segregating_morphology_methods

>(I was inspired to wonder by reading of languages whose words' primary stress
>was on the first or second syllable depending on whether the word had evenly
>many syllables or oddly many syllables.)
Yeah, not quite what I was expecting "how long the unheard part was" to mean.
- "How long do you think it'll be before we hit peak oil?"
- "Oh, an odd number of years, I reckon..."

Anyway, my (newly named) conlang Sabasasaj (till recently codenamed June 25)
does pretty much the flip of this.  In a word consisting entirely of light
syllables, the primary stress will fall on the penult in a word with evenly
many syllables and in the antipenult in a word with oddly many.  What's
actually going on is that words are parsed into trochaic feet from the left,
and then the last foot gets stressed on its head: so
lllll       llllll
become
(ll)(ll)l   (ll)(ll)(ll)
and then are stressed
(ll)(Ll)l   (ll)(ll)(Ll) .
The heads of the other feet get secondary stress.  In a word with heavy
syllables the foot formation isn't so regular, and the parity pattern gets
thrown off.

As far as I know the (an?) accepted theory explaining languages like the one
you mention is the same.  In that case words would be broken into binary
feet from the right, and then the left syllable of the leftmost foot would
get the stress.

I'd still be really surprised if a natlang was to be found whose prosodic
system does something length-sensitive much different than this, like the
stress being in a different position in <N and >=N syllable words for N
large, or being on the middle syllable, or something.  My intuition would be
that any possible stress system would be a regular language over the
alphabet of {syllable types} Cartesian product {degrees of stress} with a
pretty small bound on the degree of complexity of a recogniser.

Optimality theory folk: do OT treatments of prosody yield systems respecting
my intuition?

Alex