Funny W

[Koontz John E]

On Thu, 2 Nov 2006, Rory M Larson wrote:
> I'm not as familiar as you and Bob are with the total distribution of these
> sets over Siouan.  Could you suggest some actual problems that would
> challenge a nasally-released stop interpretation for *W and *R?

Distributionally, I don't think nasally release stops are particularly
common compared to prensalized stops.  Bob's example of forms like /ob/
common compared to prensalized stops.  Bob's example of forms like /ob/ as
[obm] and, hypothetically /el/ as [edn], are more common, but notice that
these follow oral vowels, whereas *R and *W precede vowels, generally oral
ones.  In fact, with only a very few exceptions *R and *W are strictly
word initial.

Apart from the last observation, within the context of Siouan, since there
are such strong analogies between *R and *r in second position in
clusters, we'd have to wonder if *R = tn meant that *pr tended to develop
as *ptn and *Sr as *Stn.  I suppose that might explain why *p and *S tend
to dissappear in these contexts.  :-)  (Taking *S to be s^ ~ s ~ ...)

> Or which would favor an alternative interpretation?  (If I'm
> understanding aright, you favor prenasal stops: *W = *mb, *R = *nd,
> while Bob favors ...

I'll leave Bob the task of explaining his own views, but it would be fair
to say that I suspect *R and *W to be [nd] and [mb], though not as
contrasting segments, i.e., not *r : *t : *R as r : t (or d) : nd.  I
suspect that *r represents something like [n] in initial position and [r ~
n] medially, with medial n occurring when one of the adjacent vowels was
nasal (and with that nasality tending to spread to the other vowel).  If
*r occurred initially before an oral vowel, it may have been orally
released, leading to something like /ra/ = [nda], while /raN/ = [na(N)].
Medially /ara/ = [ara] while /aNra/ and /araN/ = [aNnaN].  In fact, I
suspect that /aNra/ = [anda ~ aNna ~ aNnaN] and /araN/ = [adna ~ aNnaN ~
anaN], but that for some reason [nd] only came to contrast with [r]
initially.

Where *r and *R contrast initially, the reflexes of *R tend to be more
stop like.  Usually the stop portion prevails, but sometimes the nasal one
does, as in OP.  In phonetic terms it is a question of timing.

Where *t and *p are final in Dakotan they tend to become *r and *w, or, in
practical terms [d] and [b] or [n] and [m].  As Bob reports, if [n] and
[m] are preceded by an oral vowel, there can be some tendency to [dn] and
[bm].

In *pr most languages tend to [bd] or [md] and even [d], which parallels
the tendency of *R [nd] itself to become [d].  Where *s^r > *s^R and
this appears as s^n it may be no accident that this is encountered as hn
> n.  Perhaps the phonetics of the cluster in these cases is essentially
a breathy [nd], i.e., [hnd].  To some extent I suspect that what I analyze
as *pr > *pR and *s^r > *s^R is really just a way of saying that *r tends
to be longer and less *r-like - more stop-like - as the second element of
a cluster.  I don't know why *s^r should be more prone to this than *sr or
*xr.  As far as I know there are only three languages where *R in or out
of clusters ends up contrasting with both *r and *t.  Most places it
becomes one or the other.  But in Dakotan *R yields the l ~ d ~ n
shibboleth, while in OP it is n before oral vowels and in Winnebago it is
/d/ (often written "t") which is distinct from /r/ and from /-j^- ~ -c^/
which is the reflex of *t.  In Osage *R merges with *t.  In Ioway-Otoe it
merges with *r.  This is essentially what happens outside of Mississippi
Valley, too.

The extent to which and appearance of a contrast like *r : *R : *t seems
to endure until the very last minute, only to resolve into something like
*r : *t just before we make it's acquaintance is something of a clue.  In
effect *R is the cases where *r is more *t-like, and these cases represent
a particular set of environments for *r - initial position and certain
cluster-final positions.  Or, think of *R as something like Dakotan l,
etc.  Teton l is an alternate form of t in final position, the position
where *t becomes more *r-like, and an alternate form of y (*r) after b and
s^ and in initial position, contexts where *r becomes more *t-like.

It's true that there are initial cases of *r before oral vowels that
appear as *r, not *R, but these are mostly cases where there is a frequent
set of prefixes.  So when we get *r > y initially in Dakotan, this is with
inalienable nouns or with verbs, where there are (C)V prefixes alternating
with zero in front of the *r.  Perhaps this originally led to *r = [n] ~
[nd] alternations, but it looks like the *r-like range won out when there
were prefixes.