Re: The role of analogy, alliteration and sandhi in counting

--- In cybalist@yahoogroups.com, "tgpedersen" <tgpedersen@...> wrote:

> I imagine that in a situation in western(?) PIE where both p and kW
> existed depending on dialect and sociolect there has developed a drift
> to prefer one over the other (but not necessarily consistently). Note
> that Latin replaces p with kW in the presense of another kW: quinque,
> quercus, coquere (*penkW-, *perkW-, *pekW-), so to speak as an over-
> reaction to the presence of p-varieties *penp-, *perp-, *pep-.
> De-Oscanization, one might call it.

But the Celtic & Italic changes occur in a more specific environment
than that and there's also an opposite rule:

p() > kW / _V/G/C+syl [] kW
kW > p / p() [] _V/G/C+syl

so prope but proximus not *kWrokWe or *propsimus.