>
> You've never come across a sporadic change? Neogrammarian regularity is
> a useful idealisation, indispensable as a heuristic, but hardly
> realistic. There are untidy "facts of life" like, for example, the
> unpredictable development of Middle English /x/ after back vowels (it
> was sometimes lost, lengthening the vowel, and sometimes changed into
> /f/), or the a-colouring of Middle English /e/ before /r/ (we have
> <star, heart, start, yard, barn>, but <earth, earnest, herd, birch>).
Piotr
==============
This proves English is not the right place to look
for this particular point.
I still believe the best method is to look for a language
where changes are consistent and traceable.
Arnaud
=============
> If it has a "real" *a, then it doesn't matter what the initial is,
> especially as *h1 doesn't colour vowels.
Piotr
=========
H1 colours inherited *a into PIE *e
H2(s) colour inherited *a into PIE *a
There is no contrast in (pre)PIE between *e and *a
I will add that the contention PIE had *e and *a
BUT no contrast between *e and *o
conflicts with macro-comparative data.
But this is what we are discussing.
Arnaud
=========
Georgian zhva and PIE sa(:)l
===
All words I know for "salt" have a connection
with words meaning "sea"
English brine
Cf. Arabic baHrain
Root is *bh_H-r
Semitic yam
Cf. Sumerian TiAmat "lake of salt"
Root is *y_m?
PIE sa(:)l
KArtvelian z_hva
Root is *z_h-
I suppose some word like *zah
was first borrowed as *sa:
then it wandered within IE bubble.
Arnaud
================
>
>