Ablaut et al

[Rankin, Robert L]

>> It is there at least in the orthographic renditions that linguists have built for these languages, and quite possibly in the heads of their speakers as well.

No, not “orthographic”, phonological.  I think this is maybe where our greatest disagreement lies.  There is an actual literature in linguistics that discusses all of these questions going back 125+ years.  Modern phonology begins with Ščerba, Kruszewski, de Saussure and others in the 19th century, and treatment of these issues like faithfulness, invariance, pattern congruity, generality, degree of abstraction, etc. go way back and have been pondered for a long time by linguists working with a variety of languages.

If you look upon this as a simple matter of orthography for a couple of Mississippi Valley Siouan languages, you’ll waste your time trying to “invent” phonology while just “thinking about” one particular problem.  It seems to me that saying that the sound [e] everywhere but at the ends of words is the phoneme /e/, but word-finally it is just a “piece” of an allophone of the previous consonant phoneme – which is what you’re saying here – is going to be in big trouble no matter what phonological theory you adopt.  We can make a case for epenthesis (of /e/ or /a/, not some vague burp) in Dakota where vowel length has been lost, but it lacks phonological motivation everywhere else.  We are actually responsible for hewing to good phonological practice in Siouan linguistics.  I used to spend entire semesters trying to get that across.

>  But the dichotomy does not hold if we assume that proto-Siouan words could end in phonemic consonants characterized by an unmarked vocal release.

That’s where “respect for the data” in those cognate sets comes in (all the –e up and down the line).  The phoneme correspondences show a reconstructible open syllable language.  Period.  Pattern congruity dictates a simple structural conclusion unless there is strong motivation that militates for some additional consideration.  Only in Dakotan is such a factor found.  From a historical perspective that motivation disappears when we restore long vowels.  

> I am looking for active verb roots of CVCv type where v is something other than -e.  I don't think I've found any yet.  That leaves -e predominating statistically at somewhere close to 100%. Certainly there are many nouns, and five out of ten of the basic numbers in Omaha, that have unaccented final vowels with phonemic values that contrast with -e.  Many stative verbs have unaccented final -a, at least in attributive usage.  Among active verbs, I can think offhand of ttaNriN, 'run', with accent on the first syllable, but this is surely a compound of ttaN 'ground' + riN 'move', i.e. CV+CV, not CVCv.  We also have bexiN, 'sweep', with accent on the first syllable, but I suspect this is underlyingly ba-i-xiN, with three separate morphemes. 

So you started with the entire vocabulary but quickly found that virtually any vowel can occur in final position unaccented.  So you restricted your search to verbs.  But there were so many pesky stative verbs ending in –ka/-ga, which seems to be a morpheme.  The stative roots include lots like žį ‘little’, htą ‘big’, etc.  So now it’s just active verbs?  Add gą́ąða ‘want’, another bimorphemic stem.

We’re talking about phonology primarily, not morphology.  The syllable structure ought to be uniform.  And, in fact, it demonstrably is.  You’re just ignoring that fact, assigning “morpheme boundaries” wherever you encounter problems.  You believe (or hope) that ttąąðį is bimorphemic.  OK, let’s say it’s possible -- if it is conjugated 1s ttąblį, 2s ttąšnį (or ttąhnį, whatever), but in Kansa it’s a unit: attąyį, yattąyį, and there’s no reason to believe that ttą is ‘ground’.  Some people used to think that mąąðį ‘walk’ was ‘earth-move’ until we discovered mą ‘go’ in Catawba.  That’s the kind of comical etymologizing philosophers used to do in the Middle Ages:  vulpe ‘fox’ must be vol-ere ‘to fly’ plus pe-dem- ‘foot’, because the fox is fleet of foot.  

A parallel example:  In Spanish all the verbs end in –r.  That may make it a morpheme, but it doesn’t make it epenthetic, does it?  As we discussed last time, the –e in Siouan active verbs might even be a morpheme, but if it is, it would be the end of the epenthesis hypothesis.  

I think we just need to take the nouns, verbs and everything else at face value.  The phonology of syllable structure is largely autonomous.  For example, morpheme boundaries don’t play a role in syllable structure in the cases in which there’s a boundary between a consonant and /h/ or /ʔ/.  Mįkhé syllabifies as /mį-kʰé/ even though morphemically it is m-įk-he, and -he is a (conjugated) auxiliary.  And waną́pʔį syllabifies /wa-ną-pʔį/ even though -ʔį is the ‘wear’ morpheme.  

Bob

________________________________________
From: Siouan Linguistics [SIOUAN at listserv.unl.edu] on behalf of Rory M Larson [rlarson at unlnotes.unl.edu]
Sent: Friday, September 09, 2011 9:42 PM
To: SIOUAN at listserv.unl.edu
Subject: Re: Ablaut et al

>> It is there at least in the orthographic renditions that linguists have built for these languages, and quite possibly in the heads of their speakers as well.  But the dichotomy does not hold if we assume that proto-Siouan words could end in phonemic consonants characterized by an unmarked vocal release.  If such a release were reinterpreted as a phonemic vowel by later speakers or their linguists, the vowel chosen would most likely be -e, and next most likely -i or -a.  It would probably not be o, u, iN, aN or uN, because those sounds are marked, either by rounding or by nasalization.  The pattern we see in your table is mostly -e, some -i, and possibly one case of -a, which squares well with that expectation.

> What I'm saying is that this is precisely tantamount to reconstructing *-e, which is exactly what I do.  Trying to push -e back to a "consonant characterized by an unmarked vocal release" merely adds an unnecessary, and unjustified, step.  Why not use Occam's razor and reconstruct what is actually there?

I don't think we're in disagreement at this point in terms of Occam's razor.  If our models are "precisely tantamount" to each other, then mine has no extra step.  We are both in agreement that there was a vocalic sound following the final consonant of CVC-E type verb roots, and that the phonetic quality of that vocalic sound was closer to [e] than to any of the other four Latin vowel sounds.  The question is whether the speakers at the time ablaut developed in Siouan recognized that sound as phonemic /e/, and hence as a separate contrastive sound constituent of the root, as you advocate, or whether that sound was simply a necessary artifact of ending a word on a consonant, and hence non-phonemic, as I am suggesting.


>> This model does not imply that "all 7 other vowels (i a o u iN, aN, uN) can occur unaccented word-finally", but that "the most common one, (e)" cannot.

> Sure it does.  ALL verbs in what I reconstruct as CVCe and you reconstruct as CVC behave the same phonologically.  It is not the case that some of them "ablaut" while others don't.  So there's no reason to say that you can have both CVCe and CVC.  I think you're building an elaborate "pre-final E" phonology where there's no need.

We're both building a pre-final E phonology here, and if mine is slightly more elaborate than yours it is justified by the fact that you are claiming certainty for your model by excluding alternatives, where I only need to show a reasonable alternative that you cannot exclude.  Specifically, you are making the strong claim that a CVC model for verbs of your CVCe type is untenable because it would necessarily exclude primary CVCe roots while allowing all 7 other vowels in final position in CVCv forms.  I proposed the obvious possibility that both CVC and CVCe roots existed primarily, but collapsed together at an early stage because they were rather similar phonetically.  The CVC roots were much more common, and the CVCe roots were perhaps reanalyzed morphologically as CVC.  With this very reasonable adjustment to the CVC model, your argument against it as excluding primary final -e loses all force.


>> 2. It explains why -e is, I believe, not only the most common, but overwhelmingly the most common, ending we find, at least on active verb roots.  To the CVCe roots would be added all the presumably numerous CVC roots as well.

> But we don't need both CVCe and CVC roots, because there is no difference in behavior to motivate them.  I do take your point that -e is the most common -V by far.  However, SOME vowel has to predominate statistically.

If the final vowel of CVCv roots is phonemic, we should expect it to contrast frequently so as to distinguish words.  The argument about the final vowel above presumes that it does.  A couple of messages ago, I asked you to offer a few examples to help guide our argument.  You suggested instead that I consult the CSD PDF file and search on "PSI[ *", which moves me along one word at a time.  (Yes, in fact you did share it with me, back in 2006.  It's a wonderful resource.  Thank you very much!)  I have been doing this for a while, and admittedly have not yet got far through the file.  I am looking for active verb roots of CVCv type where v is something other than -e.  I don't think I've found any yet.  That leaves -e predominating statistically at somewhere close to 100%.

Certainly there are many nouns, and five out of ten of the basic numbers in Omaha, that have unaccented final vowels with phonemic values that contrast with -e.  Many stative verbs have unaccented final -a, at least in attributive usage.  Among active verbs, I can think offhand of ttaNriN, 'run', with accent on the first syllable, but this is surely a compound of ttaN 'ground' + riN 'move', i.e. CV+CV, not CVCv.  We also have bexiN, 'sweep', with accent on the first syllable, but I suspect this is underlyingly ba-i-xiN, with three separate morphemes.

So my question is whether we even have any irreducible active verb roots in common Siouan of form CVCv where final unaccented v is other than -e?  If so, can we roll a few out on the table?  If not, are we left with only CV, CvCV and CVC(e) patterns?  If the latter is the case, then the whole argument above about the final vowels possible for primary CVCv roots becomes moot.


> John Koontz and I discussed this quite a bit.  I think he believes that -e and -a have/had morphemic status that explains their prevalence.  I have tended to resist that analysis since I don't see the semantic relationship.

I'm open to that possibility too, and have generally speculated in John's direction in the past.  The defense of CVC that I'm currently throwing up is largely motivated by trying to make your suffixed-particle-with-initial-a model work more smoothly in my head.

Rory