Medial *pr

[Koontz John E]

This is long and those not interested in historical phonology (or perhaps
numerals) may wish to skip it.

I thought I'd come back to the issue of medial *pr in IO, since from my
point of view the two main obstacles to a hypothesis of MI paraare ~
palaani < Pre-Chiwere **phataaniN are

(1) that the l in Tutelo *p(h)alaaniN is a better match for MI l ~ r, even
if Pre-Chiwere /t/ was as d-like as modern IO /t/ (usually written "d"),
and

(2) I can't be absolutely positive that the "br" in attested IO
gree'raabriN is non-native.

Of course, in the first case MI l ~ r is also not a very good match for
Ofo's t in pata'ni(N), either.  (Ofo, not Otoe.)  Only Tutelo really works
as far as attested forms, and, of course, an attested form, like a bird in
the hand, is an attractive option.  Defence attorneys can appeal to
hypothetical alternate murderers unknown in explaining a corpse, but you
can't convict a hypothetical murderer, and perhaps one shouldn't try to
etymologize on a basis of a hypothetical etymon.  I understand the
reluctance of folks to believe in a hypothetical IO or IO-like 'eight'
along the lines of *phadaaniN, especially if I can't be totally convincing
that the -raabriN part of IO gree'raabriN 'eight' is alien.

So let me return to contention number one, which is just that -raabriN is
a very odd-looking form of PS *raa'priN 'three' to find in IO.  If you
don't believe that, you're not going to believe that a hypothetical
earlier *pha'daaniN or *phe'daaniN or even *gree'daaniN would be a more
likely form, and you're not going to believe that any of these forms might
have existed to conceivable influence MI.  I hope it will be clear that
you can believe that -raabriN stands out like a sore thumb in IO without
getting into the question of where MI 'eight' comes from at all.

Point one is simply that IO 'three' is *not* *raa'briN.  It is daa'niN.
To be more accurate, it is [daa<ny>iN], with /n/ as [<ny>] (enye) before a
front vowel.  Moreover, Winnebago 'three' is not *raapiN'niN, but daaniN',
with a parallel development.

These forms that do *not* occur are what would happen if PS *raa'priN
'three' developed as expected, or at least if it developed in the most
obvious way.  Two things happen to prevent this.  One is that the initial
*r becomes *R and *R becomes d ~ j^ in IO and d in Winnebago.  It's *R,
not *t, because, though *t becomes d ~ j^ in IO, too, and is
indistinguishable from *R there, in Winnebago *t becomes invariant j^ (or
c^in final position), which is quite distinguishable from *R.  Winnebago
daaniN' 'three' (and not *j^aaniN') implies PS *Raa'niN 'three' (and not
*taa'niN), and the same PS form *Raa'niN can also account for IO daa'niN.

The second thing that happens tends to explain the first.  This is the
substitution of medial n for medial *pr.  In fact, what happens is that
*pr in the middle of 'three' becomes *R, too, and that *R looks like n
before nasal vowels.  In other words, Winnebago and IO 'three' are from
*raa'priN via an intermediate form *Raa'RiN.  While the shift of initial
*r to *R is unexpected and irregular, the shift of medial *pr to *R is
probably entirely regular.  I say this because a number of other *pr in
Winnebago and IO have taken this route.  In particular, this happens to
*pr in noun initials and to *pr where it occurs in the first persons of
*r-stem verbs.

Some sets:

         PMV        Te        OP         IO           Wi
HAND     *ru-       y         dhi-       ru- ~ ri-    ru-
'tongue' *ree'Si    c^hez^i'  dhe(e)'ze  re(e)'dhe    reezi'
'uncle'  *(i)Re'ki  lek(s^i') (i)ne'gi   ij^e'ga      dee'g(a)
'male'   *pro'(ka)  bloka'    nu(u)'(ga) doo'ge       (c^he)do'(k)
A1+r...  *p-r...    b-l...    b-dh...    (ha)-d...    d...
'flat'   *pra's(ka) blas(ka') bdha's(ka) bra'<th>(ke) para's

S = fricative showing multipel grades, e.g., *s and *s^.

In these sets:

- HAND ru- is the familiar hand or pushing instrumental, with a
well-behaved *r.

- 'tongue' *ree'Si, shows initial *r in a body part noun, a context where
Dakotan (Teton) c^h reflects *y instead of *r, a development in many
Dakotan body part nouns.  In the other stocks *r develops as *r would
normally.  Conceivably the noun is really *yee'Si and *y > *r between
third person possessor *i- (etc.) and the following noun.  This is not
clear.

- 'uncle' (mother's brother) *i-Re'ki shows *R.  Dakotan (Teton) l, OP n
before an oral vowel, and Winnebago d that doesn't become j^ are the
particular marks of this correspondence.  In other MV Siouan languages *R
is indistinguishable from *t.

- 'male' *pro'-ka (often with the noun former -ka in one or more of its
reflexes) shows initial *pr in a noun, which behaves as *pr in Dakotan,
but is elsewhere *R.

- A1 + r..., or the first person of *r-stems like those formed with *ru-
shows *pr in Dakotan and Dhegiha, but *R in IO and Wineebago (with added
regular A1 ha in IO).

- 'flat' *pra's, sometimes extended with *ka, shows *pr at the start of a
verb stem, i.e., where first person inflection does not provide the *p,
which is instead an organic part of the stem.  This behaves as *pr across
the board.

We have to assume that the pattern of behavior with *pr above involves
some degree of analogical levelling.  It is entertaining to imagine that
Siouan sound laws might be conditioned by morphological and/or lexical
considerations, but it seems more likely that *pr became *R in initial
position in early Dhegiha, Winnebago, and IO while remaining *pr
non-initially and that the various morphological contexts sorted out
differently in terms of whether the resulting *R ~ *pr alternation
levelled to *R or *pr.  For example, *R won across the board with nouns,
lost across the board with verb stem initials, and won in inflection in
Winnebago and IO, but not Dhegiha.

It's been suggested that some of these *pr might be *wr or *br, but the
patterns of correspondence, whatever the reconstructions may be, fall out
exactly by the morphological contexts, not randomly mingled, and I am not
easy with a scheme that restricts, say *pr to nouns, *br to inflection,
and *wr to verbs.  Besides, there are reasons why we might expect the
initial labial to come from *wV- in all of these cases.  So, I prefer to
see these cases all as *pr (maybe *[br]) that does not contrast with *wr,
and that in many cases comes from *wV-r, but develops differently in
different contexts.

So far I have only considered *r, *R, and *pr before oral vowels.  It
turns out that *r before nasal vowels develops more or less as an n,
except that sometimes it stays r (or whatever *r becomes), usually in
verbs.  If we try to see nasalization of *r to n as not occurring in verbs
(in the relevant languages, e.g., OP) then the exceptions are those verbs
where *r is always n.  I conclude that I don't quite understand the
interaction of *r and nasality, but I won't go into the examples.  As far
as *R before nasal vowels, it seems not to occur.  Maybe that explains the
anomalous n-verbs - *R/__VN is always n, though *r is not?  Again, this is
something of a side issue, and I'm not going to go into it here.

So, that brings us to *pr before nasal vowels, and here we have some clear
examples.

         PMV        Te        OP         IO           Wi
'water'  *priN      mni(N)'   niN(iN)'   niN(iN)'     niNiN'
'smell'  *-praN     -mna(N)   -bdhaN     -braN        -paNnaN

This latter example here is the stinky kind of smell, or at least 'to have
an odor of', which is often found with the *o- locative.

I'm not thinking of any definitely *rVN... verbs to do inflection with.
The cases that come close seem to be really *yVN... or to have epenthetic
*r in the third and sometimes second persons on a stem *(?)VN...  Passing
regretfully on to conclusions we can see that *pr in verb stems behaves
pretty much the same before nasal vowels as before oral ones.  But in
nouns it is mn in Dakotan, where it was bl (Teton of course) before oral
vowels.  And in nouns elsewhere it looks like an n, or just like *r looks
when a nasal vowel follows it.

So, we have our choice of saying that *pr in nouns before nasal vowels
becomes *r which naturally behaves as *r would behave before a nasal
vowel, becoming n, or that *pr in nasal nouns becomes *R as it does in
oral nouns, and that *R behaves like *r before nasal vowels and becomes n.
(Maybe it makes sense to think of anomalous n in verbs as *R/__VN?  But I
digress - as usual.)

On that note, I return to 'three'.  It seems that *pr PS *raa'priN becomes
n, leading to *Raa'niN because *pr/__VN in medial position acts like n.

There's still that nagging initial *R in *Raa'niN (or really *Raa'RiN)
which derives irregularly from *r.  Perhaps this is because
Pre-Proto-Winnebago-IO had *praa'priN for 'three'?  The initial *pr would
need explaining, but the phonology would be regular.  The initial *pr
might be influence from the medial *pr, or, there are some constructions
in which numrals receive *wV prefixes, and perhaps this is the source of
the contamination.  The former seems more likely because there's no trace
of *wV- prefixation running amuck on other Winnebago-IO numerals that I
know of.

Based on the foregoing, I am proposing that medial *pr behaves like
noun-initial *pr in Winnebago and IO.  But what is medial *pr and are
there other cases of it?

The whole concept of (root) mediality in Siouan is fraught with
difficulties.  The cannonical Siouan root is a (C)CV or (C)CVCV form, and
the last V is often a bit uncertain or arbitrary or missing in the
bisyllabic cases.  There just aren't many (C)CVCCV stems.  Anything with a
heavier or longer structure than (C)CV(CV) is highly suspect, as a root.
Even forms like (C)CVCVN are likely not to be a single root.  You can, of
course, get much longer forms by compounding, derivational prefixing,
enclisis, and inflection - or combinations of these.  Longer forms are
also sometimes loans, as in cases like Teton s^agla's^a 'British person'
or OP kku'kkusi 'pig' or kku'kkumiN 'cucumber'.  When not loans they must
be suspected of involving a compound, possibly an old compound no longer
transparent, of the form (C)CVC-CV or (C)CV-(C)CV.  In many cases
inflectional patterns suggest that to native perceptions "no longer
etymologically transparent" and "not obviously a compound" are two
different things, since forms like this are often the infixing or multiply
inflected verbs.

With this in mind, I have the following candidates for medial *pr in mind,
though I am not sure some or all of them do not involve compounds or
derivational prefixing and in some cases I am essentially positive that
they do.

         PMV         Te              OP           IO          Wi
'three'  *raa'priN   ya'mni(N)       dhaa'bdhiN   daa'niN     daaniN'
'bean'   *huN'priNka omni(N)'c^a     hiNbdhiN'ge  uNuN'niNge  huNuNniN'k
woB      *(i)hti'pro thiblo'         (i)tti'nu    (i)c^hiN'do (hi)c^ido(ra)

'ten'    *kyee'praN  (wi)kc^e'mna(N) gdhe'b(dh)aN gree'braN   kerepaNnaN'

Note that Winnebago 'ten' usually has (h)iz^aN 'one' appended to it.
'Bean' is suspected of being a loan, perhaps ultimately from Uto-Aztecan,
and the initial syllable is very irregular across the Siouan languages
that have versions of it, so the loan was probably fairly late, i.e., not
into Proto-Siouan or even Proto-Mississippi Valley.  The abbreviation woB
stands for 'woman's older brother'.  This form in particular may be a a
compound 'house' + 'male', i.e., something like 'household (or lineage)
male'.  OP has simplfied gdhee'bdhaN 'ten' to gdhee'baN since the earliest
recording of OP 'ten', probably sometime in the middle 1800s.

Three of these four forms have *pr becoming *R (or n before a nasal vowel)
in IO and Winnebago.  The exception is 'ten', and, of course the final
-raabriN in IO 'eight' is exceptional, too.  I would expect 'ten'
*gree'naN in IO and 'eight' *gree'daaniN.  I do not know why either is
exceptional, though my hypothesis with 'eight' has been that it is a loan,
even though no other Siouan language has a gree-N construction for any 5+N
numeral, let alone 'eight'. It would be difficult to maintain that 'ten'
was a loan, too, because it is also exceptional in Winnebago, which has an
innovated form haruwaN'k for 'eight'.  So, if 'ten' is a loan, it would
have to have been an early loan, which might prove problematic in
explaining the Miami-Illinois form as a (fairly) recent loan of a
preceding form.

One might explain retention of *pr as br in 'eight' as due to the rhythmic
structure of 'eight' (CVV'CVprV), but the rhythmic structure of 'three'
and 'ten' is much the same (CVV'prV), and 'three' changes *pr to *R (n),
while 'ten' does not.

I will leave matters at that point for now.  Additional examples of medial
*pr would be welcome!