--- In
cybalist@yahoogroups.com, "P&G" <petegray@...> wrote:
>
> This, however, does not answer your basic question. A more
interesting
> souce for the same question is the root (1) vr. = "cover". It
shows both
> vr.noti and u:rnoti. The first is presumably from PIE *vl-ne-u-ti
(Pokorny
> page 1140). I don't know why the second has a long vowel, and
reverses
> syllabicity. Perhaps someone else can explain?
I believe so. I posit the root as *(H)welHw- 'envelop' or *(H)werHw-
'defend' which both exist and would merge in Indo-Iranian. The
regular nasal present would be *(H)wl.H-né-w-ti >
u:rn.óti 'envelops, defends'. The aorist was *(H)wélHw-t > PIE *(H)
wél-t, Skt. á:var. On the basis of the aorist the present was
reshaped to vr.n.óti (like kr.n.óti to á-kar).
The one-vowel system breaks down when loanwords and names enter the
language, such as toyam 'water' from Dravidian which should be *
[tavyam] if analyzed /tavyam/. There is a systematic ambiguity with
vra- which is opposed to ura-. However, the latter may still
synchronically be analyzed as /vrra-/ (as per Edgerton),
supposing /rra/ to be realized "vocalic r + consonantal r + a", then
changed automatically into "prop-vowel + r + a", followed by a
change of "v + prop-vowel" into [vu] with subsequent loss of [v]
before [u]. All of this would be free of contradiction and quite
well-motivated (which is not even really a requirement).
It also breaks down when analogy forgets that [y] and [i], etc. are
not distinct phonemes. But that goes for PIE also, so the parallel
is still a viable one, in fact it is surprisingly close.
Jens