Re: THEORY: Allophones

Pablo Flores wrote:

>I know we discussed this before, but I have a doubt
>concerning allophones and phonetic change rules.
>Suppose there's a language allowing a syllable
>structure CVF (F =3D fricative or liquid), where you
>can compound roots not necessarily conforming to
>CVF, and you have, say,
>
>        /kak/ + /tat/ =3D /kaxtat/
>
>(there's /x/ in the language independently from the
>one related to /k/). Now, what's the /x/ in /kaxtat/,
>an allophone of /k/ or merely an /x/ coming from /k/
>via a phonetic change rule?
Don't pin me down on this, but I think the second one=20
is right. It is not an allophone, but an /x/ coming=20
from /k/ via a change rule. Isn't this essentially what=20
consonant gradation is all about?

I have seen this phenomenon in Nivkh (aka. Gilyak), a=20
language isolate spoken on Sakhalin Island just north=20
of Japan. Many consonants change to other consonants=20
that already exist in the phoneme inventory when affixes=20
are attached. The type of change depends on the=20
environment that results from affixation. AFAIK, the=20
phenomenon in Nivkh also extends across word boundaries=20
(much as it would in Celtic languages)!! So depending on=20
what the neighboring words are, a specific word is=20
hardly ever spelled the same in a sentence as it would=20
on citation form. The description I saw of Nivkh called=20
it consonant gradation. I suppose this is the same thing =20
in Celtic languages - although, AFAIK, the consonant=20
gradation in Celtic only applies to certain words like=20
articles or some other, while in Nivkh all words apply.

-kristian- 8)