Re: [tied] Re: Caland [was -m (-n)?]

On Fri, 28 Nov 2003, Miguel Carrasquer wrote:

> > I do not think there are 10 %
> >relic forms, on the contrary I estimate there are some 10 % *innovative*
> >forms with the "wrong" suffix, i.e. with *-ro- combined with a sonant
> >nucleus, or *-u- after a vowel (vocoid) nucleus.
>
> That would be even more remarkable.

In what way? There used to be zero percent overlap, then it grew, and by
the split-up of PIE it seems to have been something in the order of 10 %.

>
> Maybe what I mean will become clearer if I use a purely phonetic example
> instead of a morphophonetic.
>
> Suppose we have a proto-language with the phoneme /k/.  The theory is
> that
> a soundlaw applies, turning /k/ into /c^/ under certain conditions (say,
> before front vowels).  It's supposed that there was no phoneme /c^/ to
> begin with, so it's a split, no merger involved. 
>
> Now what we actually find is one language, say Tofarian, which has /k/
> everywhere.  Another language, Lifuanian, has /c^/ everywhere.  The other
> daughter languages show the distribution /c^/ before front vowel, /k/
> before back, with perhaps some 10% exceptions. 
>
> How believable is the theory?

Do you seriously mean this? There is no denying that in PIE /u/ and /ro/
are *not* in complementary distribution, but they appear to be nearly so
when forming adjectives, so I try to explain how that can be. In this
theory it is *not* presupposed that there were no other segments /u/ and
/ro/ in the language, so the parallel is not parallel.

Jens