>
> Basque has ur "water" but no trace of *wer- (pace Miguel's *(w)ur-
).
> I think that is a pointer that this loanword from wherever was
> originally *ur, as in Basque, and that it was borrowed into PIE as
> such, which then manhandled it according to its ablaut rules. Thus,
> no PIE *wer-, but PIE *ur-.
>
The Basque combining forms of 'ur' are uh-, ub-, ug-. The /r/ of 'ur'
is difficult to reconcile with whatever the consonant it was that
produced the various last consonants of the combing forms. I believe
the /-r/ is a locative suffix, found also in Sanskrit 'va:ri' (with
extra locative -i), nominally a nominative, and the word very likely
once a r/n heteroclitic (gen.sg. 'va:rin.as'), the nom. /-r/ of which
(as also in 'va:ri') I believe once was a locative marker; all those
r/n heteroclic neuters designate materials or objects of which it is
odd to assign a role of acting agent, rather than location of the
action (of the verb). Others have proposed that 'mare' "sea" is also
a locative in -r of a word that's found again in Egyptian (forgot the
details, can't find the reference).
As I probably told before, a spanish hitchhiker we once picked up
told us, he was from 'þara?oþa', I believe it was. I was glad I
wasn't doing linguistic fieldwork, since I wouldn't know whether it
was 'þaraGoza', 'þaraGWoza', 'þarawoþa'. The problem here is that /?/
is the 'low' or 'sloppy' version of /G/, /GW/ and /w/ and these in
turn of the 'high' or 'exact' /g/, /b/ and /v/. Thus, when a language
is available only in the 'low' form, when 'higher-ups' redefine it,
they have no clue, literally, so they come up with /g/'s, /b/'s
and /v/'s with no system to it (this trend is why the /w/'s have been
purged from the European continent, replaced by the 'exact'
French /v/). And, to conclude the thread, that's why I believe that
the origin of the Basque combining forms uh-, ug-, ub- is *up-,
weakened to *u?- and thence "reconstructed" with a proper consonant.
Torsten