Hamp analyses it as the old
> neuter dual *h3okW-íh1 which fell in, formally, with a feminine,
> imitating the pattern of the numeral "two", i.e. f. dy: < *dy-ë, a
> femininising extension of <dy> from *du: < *d(w)o:u (monophthongised
> after the unrounding of inherited *-u: in monosyllabic words). In
other
> words, the replacement of expected *si by <sy:> is due to analogy,
not
> to regular phonetic development.
>
> Piotr
First of all I want to thank you, Piotr. You have the patience to
look back among all past messages (and there was many) and to identify
my question. So thanks for the answer.
a) If I remember well (but I need to check), Orel suspect that an i>y
rule exists for some contexts in Albanian. Also I saw that Abdullah
suspect the existence of this kind of rule, too. On my side, I cannot
see any i>y rule in Albanian (there is an u:>i rule in some contexts in
place of u: > y (see bisht): but this is another rule). If I understood
well you have also the opinion: that there is no i>y rule in Albanian.
b) On the other hand, I want to add an additional question:
Why in your opinion we need to reject the idea that a final u(:)
existed in PAlb form *ac^-i(:)-u(:) with an additional corresponding
ending to PIE *h3okW-íh1 + 'source of this u(:)'? Only because the
other cognates don't show us a similar ending?
c) Talking about singularities: Did the Latin oculus shows an initial
PIE long vowel (even the initial o in the Lat. oculus was a short
one ?) Something like: oculus < PIE *h3o:kWe-los?
Thanks again and Best Regards,
Marius
P.S. A not related question: why you preferred *h3okW-ih1 in place of
*h3ekW-ih1?