Re: Vowel length near-minimal pairs in Tirelat

ROGER MILLS wrote:
> Herman Miller wrote:
>
> (snip examples)
>> So it looks like vowel length (and not stress) is distinctive.
>
> Yes, from that data.
>
> But
>> there's a complication: an unwritten schwa sound in some words, which
>> is always unstressed. E.g.:
>>
>> dbaxa /d@'baxa/ "to resist" (not /'d@baxa/)
>> knagi /k@'nagi/ "brass" (not /'k@nagi/)
>> tezn /'tEz@n/ "transparent plastic ball for gerbils" (not /tE'z@n/)
>> zgaki /z@'gaki/ "similar" (not /'z@gaki/)
>
> That isn't phonemic, merely a phonetic/sub-phonemic transition sound in
> the surface structure, to facilitate the cluster. In my favorite
> generative terms, [@] would be a very late (maybe even the last) rule in
> the derivation, surely _after_ stress has been assigned.  Is penultimate
> stress the rule? no problem in that case; or is it "stress the first
> (phonemic) vowel of the word?" or maybe something else-- still no
> problem, since whenever stress is assiged, [@] "isn't there yet" in
> phonological terms.
>
I'd agree that this isn't phonemic; in fact this epenthetic schwa is one
of the oldest features in the language, from before Tirelat even had a
phonemic schwa vowel. If you just look at the surface forms, though, you
have an apparent stress distinction between words like /k@'nagi/ and
words like /'s@lagi/.

But then, if you add a possessive prefix it becomes apparent what the
actual stems are. E.g.

lknagi /l@k'nagi/ "my brass"
lsəlagi /l@'s@lagi/ "my cod".