Indian perfume set.

[Koontz John E]

On Tue, 7 Oct 2003, R. Rankin wrote:
> As I mentioned, the *o > u change may affect locative o- when other
> elements are grafted on as prefixes or proclitics.  This simply hasn't
> been determined yet. Jimm may know.  The problem is that relatively few
> elements can precede the locative prefixes.

Jimm confirmed that *o- appears as u- in all cases of incorporation of
some element before it that he could think of.

> Also it would be good for us to keep in mind that the sequence John
> reconstructs as *pr has two (at least) proto-Siouan sources.  One is *wr
> (most of the instances, actually) and the other may be *pr (but there
> were precious few *p in these sequences, 'flat' may have been one).  So
> *py may behave like one or the other or differently because it had a
> different chronology.

We've discussed this before.  Bob prefers to distinguish *wr and *pr,
while I consolidate all these sets phonologically.  There are sets of this
type or types arising in nouns root initials, in verb root initials, in *r
with first person pronominals, and medially (basically, in 'three').

Dakota pretty nuch treats all of them as *pr, which becomes bl/bd/md/mn
before orals depending on the dialect, and mn before nasals generally.

Dhegiha treats the nouns (*wr in Bob's analysis) as *R, n/d ~ j^/t ~
c/t depending on the dialect, but n before nasals, and the rest appear
as *pr (bdh/bl/br/pd, depending on the dialect, and no nasal variant,
e.g., bdhaN 'to smell of something').  \

IO and Wi treat the nouns and the inflected verb forms and the medial form
as *R (d with nasal variant n), but retains *pr in verb stem initials
(br/pVrV, depending on the dialect and no nasal variant).

So, the verb-stem initial forms (like 'flat' *pras(ka)) are the only ones
that never get reduced anywhere.

Of course, though I don't find any reason but this last to distinguish *wr
from *pr, I don't claim that there are not multiple origins for *pr, or
perhaps we should always write *wr.  In particular, most of the
polymorphemic cases of *pr (or *wr) clearly have a paired uncontracted
variant *wV-r, suggesting that these are from *w + *r or perhaps that a
lot of earlier **pV sequences "soften" to *wV.