Rob wrote:
>>Strangely enough, Lubotsky doesn't even mention some of the most
>>commonly quoted roots and words with *a, such as *kap- and *kan-.
>
>
> Those could very well be loanwords, IMO.
_Anything_ could be a loanword, but if you claim that PIE had no *a and
then dismiss any piece of counterevidence as a loanword just because it
shows an *a, that's circular.
>>Some of the analyses (e.g. that of *k^aso-) strain all credulity.
>
> How so?
Lubotsky develops a long argument trying to prove that *k^as- is really
*k^h1-es- (itself a strange form, ablaut-wise, and derived from an
uncertain root), and paints himself into a corner as this "improved"
reconstruction creates more problems than it was to solve -- the wrong
vowel in Old Prussian, the wrong initial in Indo-Iranian. He desperately
posits proterodynamic *keh1s- with the weak stem *k^h1es- _or_ *k^h1os-
(or "secondary -o-vocalism") and the weak grade *k&1s-, and still has to
rely on complex analogy to account for all the cognates. It's surely
less trouble to posit *k^as-, which accounts equally well for Germanic
*xas-, Italic and Celtic *kas-, OPr. sas- and Indo-Iranian *s'as- (and
for the failure of Brugmann's Law).
> Can you explain that counterargument in more detail?
>
> How realistic would a form *lexiwos be in IE?
Less realistic than *laiwos, given the circumflex intonation, which
indicates a laryngeal-less root. Derksen posits *lh2aiwo-. Brave boys,
laryngealising to the last ;-).
Piotr