Re: [pieml] Labiovelars versus Palatals + Labiovelar Approximant

Arnaud Fournet wrote:

> Latin can-is better fits kh2n-

But it doesn't fit *k^uh2(o)n-, and that's what you proposed. *k^uh2n-
would have given Lat. *cu:n-. Did the *u disappear just like that?

> I agree a PIE reconstruction has to fit IE data, and *k^un can
> certainly not be taken for granted.

Why "taken for granted"? Gk. kunos, Skt. s'unas, Lith. s^uns, OIr. con
simply point to *u in the weak cases.

> If you start saying that LAtin can-is is a peculiar development, why
> should your k^un not be the same thing ? How do you explain Latin
> can-is ?

I don't have to. I have made no claims about <canis> being a member of
this word-family.

> ======= No, Greek does not work, and looks like a Tocharian loanword
> Anatolian does not work and looks like an Iranian loanword.

Even if it's a loan in Anatolian, what difference does it make? It's
still a loan within IE.

> Indo-iranian words don't make phonological sense, unless you change
> the laws of Phonology.

What are those "laws of Phonology" and who dictates them? The
Indo-Iranian forms match those found in Tocharian, Germanic, Italic,
Celtic and Baltic.

Piotr