> Piotr wrote:
> Mind you, what I said (after Hamp) was that the ending *-i: in
> something like *ac^í: was _replaced by_, not _extended with_ a
> reflex of *-o:u. An archaic, non-productive dual ending was
> replaced with one borrowed from the numeral "two", which is a
> perfectly natural process.
Thanks again, Piotr for these additional clarifications.
On my side, I hope that I well understood that your explanation
was "the ending *-i: in something like *ac^í: was _replaced by_, not
_extended with_ a reflex of *-o:u"
My question back (or at least my intention for this question)
was "why we need to completely reject a different idea: that the
ending *-i: was _extended by_, and not_replaced_with_ a reflex of *-
o:u (or at least 'by a source of an u(:)"
I have asked this because:
1. to suppose that a "non-productive dual ending was replaced with
one borrowed from the numeral "two", which is a perfectly natural
process" => is fully logical.
2. but on the other hand:
a) the supposed phonetic rule: o:u > u: (>y) is not 'well
reflected' in Albanian (based on this I have doubts also about
Hamp's derivation of dy...)
b) even your proposed contraction "e.g. the contraction of *du[w]
o: > *dúö: > *du: > dy" is a 'local' solution.
Viewing these issues my proposal/my question was : why not an
extension in "u" in PAlb (like *ac^i(:)-u(:) ), directly sourced in
PIE? (even this extension was added later (not in PIE times))
I asked this because: i(:)u(:), u(:)i(:), u: regular gave y in
Albanian.
Thanks again,
Marius