PIE e/o Ablaut

[Jens Elmegaard Rasmussen]

On Wed, 15 Mar 2000, Vidhyanath Rao wrote:

> []
> When JER suggested Sanskrit as an ``almost'' example several months ago,
> I looked for exceptions. I gave up because it is hard to do searches
> like these without a comprehensive word list (one exists, but I don't
> have access to it). But I am doubtful because pairs of type viyukta
> (unyoked) and vyukta (vi + (vac+ta), explained) must have existed even
> if I am unable to point to them on demand in RV.

> We can try to use accent to distinguish them, but then we run into the
> problem of how vyukta was actually pronounced (it seemed to have varied
> over time and space). []

Well, if there was no difference between the two words, it does not matter
how we write them. If there was, one may just write /vyyvkta/ and /vyvkta/
in accordance with the morphemic analysis, the former giving a reading
[viyukta], the latter, [vyukta], by the rule demanding that you begin from
the end of the word and syllabify every sonant not already having a
syllabic neighbor (that was also the PIE rule). A few subrules should be
added allowing specific sonant clusters (as for PIE), as here vyu-, not
+uyu-. It is not absolutely flawless for synchronic Sanskrit, given the
interesting syllabifications caused by laryngeals that have later
vanished. But if it is used to _recover_ the stage with segmental
laryngeals still around, it works remarkably well. That even demands such
pairs as vra- : ura- to be posited as /vra-/ and /vrHa-/, and the adj.
uru- as /vrH-v-/, just as Avest. vouro- shows it really was. For urV-, one
might even contemplate a synchronic notation /vrrV-/ which, with an
Edgerton-style development of syllabic + asyllabic shape of the same
sonant would give what we find. Typologically that may even be preferable,
for there is no guarantee that the old stage with laryngeals intact is not
in fact _so_ old that the monotonous /a/ was still the original triad /e/
: /a/ : /o/. In any case, it works best in cyberspace.

Jens