Siouan glottal stops, some stats.

[Robert L. Rankin]

I looked through the Proto-Siouan reconstructions and, of around 1200
reconstructed words, about 25 have an independent glottal stop (and I'm
not counting instances where the [?] is part of an ejective stop or
fricative like p?, t?, k?, s?, etc).  Of these, 12 are in initial
position in the word, 12 are medial, usually but not quite always
between vowels, while one is final.  Looking at morphemes instead of
whole words, the vast majority of the glottal stops are
morpheme-initial, all but about 3.

The Siouan ejectives may also involve independent glottal stops however.
This is because, whereas in Dakota and the other Mississippi Valley
Siouan languages there are ejective consonants, in the northern
languages (Hidatsa, Mandan with traces in Crow) those same consonants
are non-ejective, but the syllable ends with a /?/.  So:
	Dakotan  C?V	corresponds to
	Hidatsa  CV?	etc.

Thus it is unclear whether we should consider the proto language to have
had ejectives, like the Mississippi V. languages, independent syllable
final /?/ as in the North, or perhaps glottalized vowels [V?]
contrasting with the oral and nasal varieties.  The eastern languages
don't help here.  Sapir's few Tutelo transcriptions show that they had
/?/ but the rest of the transcribers didn't bother with glottal stops
unfortunately.  It is clear though that, if we add the ejectives, there
are lots more glottal stops.

I'm tempted to reconstruct the ejectives and consider the Northern
languages to have decomposed them.  This is because statistically the
consonants that turn up as Dakotan p?, t?, k?; s?, $?, x? fit Joseph
Greenberg's typology of glottalic consonants (his IJAL paper back in the
early '70's) so well.  There are many glottalized velars, fewer
glottalized palatals and dentals while labial p? is extremely rare,
occurring in only one or two stems.  This is just what we'd expect if
these consonants had started out as real ejectives (as in Dakotan).

Bob