Funny W

[Koontz John E]

On Fri, 3 Nov 2006, Rory M Larson wrote:
> I don't think the first is a problem.  We're dealing with a specific case,
> not a probability.  World-wide, clicks are rare too, but that doesn't deter
> their presence in Khoi-San.

Still, we don't usually hypothesize them when they are not attested in the
daughter languages.

> Being rare and distinctive would be an advantage to preserving *W and *R
> and keeping them from merging sporadically with other sets.  If, on the
> other hand, they were pre-nasalized stops, then what would consistently
> distinguish them in their reflex pattern from all the epenthetically
> pre-nasalized interior stops following nasal vowels?  If the Dhegihan
> positional *aWa' was actualized as *ampa' or *amba', why wouldn't the
> interior consonant in 'day', *aN'pa or *aN'ba, have developed as *a'Wa ?

I wondered if this case was bothering you, since you mentioned it earlier.
This is pretty much the major case of "word internal" *W.  Of course, it
is word internal mainly by the grace of how we now analyze and write the
elements involved.  My take on this is that akha and ama < *akha and *aWa
are the result of lumping two elements following a noun into one:

nu=akha 'the man' < *pro-a=kha
nu=ama 'the men'  < *pro-a=Wa

My gloss for ama above is farily notional.  It is awkward to add or "the
moving man" or to write "the slightly off topic/newly topicalized man."

Historically I believe these forms both involve an underlying *pro-a, with
an appended -a, analogous in structure to Dakotan relict forms like
he-(y)a 'louse', wiN-(y)a(N)  'woman', iN-(y)a(N) 'stone', thi-(y)a=ta 'in
(the) dwelling', etc., not to mention s^uNk-a.  The same morpheme appears
as =ra 'the' in Winnebago hee=ra 'the louse', etc.  And, of course, it
also appears inOP tti-a=di 'in(to) (the) dwelling', tti-a=tta 'to(ward)
(the) dwelling'. That is, in all these forms we see an old thematic
element -a added to the stem.  This element is mostly lost in modern
morphology, but it is preserved in Dakotan when the thematic form has
replaced the underlying form (heya replaces he, except in compounds, etc.)
or when something else follows, as with Dakotan thiyata or OP nu=(a)kha or
tti-(a)di.  Sometimes it becomes a morpheme itself - or remains one? - as
in Winnebago.

So, the element following *pro-a > OP nu-a is either *kha or *Wa.  I don't
have any concrete explanation for these elements in terms of obviosu
correspondents in other branches.  A simple match like PreDakotan *k-ha
and *Wa would be nice of course!  I assume that some sort of highly
reduced morpheme sequence is a possibility instead.  Assuming something
matching the other languages is a desideratum, how about:

*kha < *k(iN)-ha
*Wa  < *p(i) -ha

About all I can say in defense of these is that

(a) an element A1 he, A2 s^e, A3 -- is found after the obviative
articles, even =ma obviative collective, which is =ma=s^e in vocatives
addresed to collected groups, and

(b) we see -p acting like *W in Dakotan, in forms like hakab 'afterward'.

In these -b cases I think the usual assumption is that -b derives from
=pha, not =pi, I think, and I'm not trying to contradict that, only argue
that -p(h)# in Dakotan comes out like *W.

If there's anything in these latter musings, it would mean that forms like
nu=akha and nu=ama derive from something like *pro-a=k(i)=ha and
*pro-a=p(i)=ha with quite a stack-up of enclitic elements.  I'd have to
assume that *pro-a=pi was well on the way to becomming *Ro-a=W by the time
the =ha (or whatever it is - ha 'declarative'?) was added (for whatever
reason).  I also have to assume that a fairly verbal style of presenting
nouns was once the norm in Dhegiha.  (Before you object, think of people
who say "(if) you know what I'm saying" after every new topic sentence (?)
in English, usually reduced to /(fyu)na m se:n/ or /namse:n/.)

I do have one possible confirmation for this.  The quotative (or
reportive) =ama could conceivably come from *a=pi=ha 'they say' via
a=W=ha > ama.

In this little exegesis it may have to think of *R and *W as d ~ n and b ~
m, but in a language where the unaspirated stops are t and p.  This is
pretty much the situation in Dakotan, for example.

> Consideration 3 is one that might raise a problem for the nasally released
> stop model.  If the stop is released nasally, wouldn't we expect the
> following vowel to be more likely nasal than oral?

That's what I'm thinking.

> construed as a nasal allomorph of *w or *r.  Thus, e.g.:
>
>   Nasal           *NN
>   Oral         TTTTTaaaaaa
>   Laryngeal         VVVVVV

I'm thinking more along the lines of

   Nasality  Tier       N
                        |
   Segmental Tier     # C            V
                        [+ resonant]
                        R            V
                        n            u

In effect "orality" spreads outward from oral vowels and pushes into the
inherently nasal resonants, producing prenasalized stops initially, and,
presumably, in the reverse fashion, post-nasalized stops finally.

This helps explain how *pr "bd" becomes *R "nd"  If we think of "bd" as
essentially "md" then the difference between md and nd is basically one of
assimilation of place in "md."  The asimilation occurs most easily in
strict initial position (in nouns) and least easily in internal position
(in stem initials of verbs).  Where md occurs in first persons the
environment is somewhat intermediate - and there's a reason to be more
aware of the m- - so we get preservation of md (now bdh) in OP, etc., and
loss in Chiwere and Winnebago.

Incidentally, the same languagfes that tend t reduce *pr to *R in cases
with a following oral vowel tend to reduce *pr to just n before nasal
vowels, e.g., *pr > nu, to, etc., 'male' (also inthe homophonpus root
'potato') and *priN > niN 'water'.  Dakotan, which keeps *pr as bl across
the board has mniN 'water'.  And, of course, when A => iN before ktA
IRREALIS, you get ble 'I go' changing (historically) to mniN as in
mniN=kte 'I will go'.

So, *r and *R and *w and *W end up contrasting sets, and it helps to think
of them in that sense, but, to a fair extent, *r and *R and *w and *W may
not have been contrasting sets in Proto-Mississippi Valley.  The
difference between the "normal" and "funny" resonants is that the latter
have been subject to a morphologically significant and analogically
revised set of sound changes that draw them closer to the corresponding
stops, but not always all the way.  In many cases some stops (e..g, final
ones in Dakotan, or initial ones in *Cr clusters) get drawn into the *R/W
pattern from the other direction.