Hallo!

On Mon, 7 Dec 2009 21:45:45 -0500, Vasiliy Chernov wrote:

> On Fri, 4 Dec 2009 22:44:42 +0100, Jörg Rhiemeier wrote:
> 
> >Perhaps the IE system of ablauting vowels developed from an alternation
> >between */a/ and */@/ (as still found today in Northwest Caucasian)
> >which resulted from the weakening of unaccented vowels - which is
> >rather to be expected from a *stress* accent.  There are good reasons
> >to reconstruct a pitch accent for Late PIE, but in an earlier stage,
> >the language may have had a stress accent.
> 
> But this would imply that two non-zero-grade vowels would be disallowed to
> co-occur in one word. 
> 
> Also, what do you think of quantitative Ablaut (vrddhi and plain 
> lengthening)? 

These are indeed problems I haven't solved so far.  Indeed, the whole
thing is just an *idea* of which I know that it is not the definite
answer.  If it was, it would be a major breakthrough in Indo-European
historical linguistics, and my name mentioned in every handbook of this
field to be written in the future.  I frankly don't expect to attain
that fame anytime in my life.

> >Some */@/s, however, remained, especially
> >where no neighbouring segment could take over the syllable peak (stops
> >and sibilants could not be syllabic), and total loss of the vowel would
> >have left behind inadmissible consonant clusters.
> 
> There might be also a contrast between schwa and zero... 

Sure.  That contrast would lie beneath the difference between o-grade
and zero grade.

> Why not just assume that w and j were deleted before another sonorant? That
> is, that all the w's and j's in contact with other sonorants (mostly on
> morpheme boundaries) are produced by schwa-deletion? 

I tried, but it doesn't work.  One would expect *CeR~*CiR and *CeR~*CuR
ablaut pairs resulting from that, but they simply do not occur.

> >In some instance, the pre-GVC vowel can be recovered form the secondary
> >articulations of neighbouring velars.  Late PIE had three series of
> >velar obstruents: plain, palatalized and labialized.  Pre-GVC Pre-PIE
> >had only one velar series, as Uralic and Altaic still do, but velars
> >neighbouring front vowels were palatalized, and velars neighbouring
> >rounded vowels were labialized. 
> 
> This is remarkably similar to Illich-Svitych's reconstruction, and the more
> elaborated version by Aaron Dolgopolsky. 

I don't know those reconstructions well, as most of the stuff they
wrote is in Russian and has never been translated, and I don't master
that language.

> >Well, I am not sure about that at all, and consider it merely a
> >theoretical construct to get some order into it.  A better way is
> >perhaps this:
> >
> >1. Stressed vowels were phonetically lengthened (not very unusual,
> >   I think).
> >2. [i:] and [u:] diphthongized to [ai] and [au] (quite common).
> >3. Short [a] was weakened to [@] (also not uncommon).
> >
> >This way, there never was a one-vowel stage, however brief: unstressed
> >/i/ and /u/ never changed.  There may also have been other lengthening
> >factors in addition to stress, which could explain cases where full
> >grades occur that are not predicted by the penultimate stress accent rule.
> 
> Thank you, Jörg! This is probably the clearest explanation I've ever seen of
> the point that there was no "vowelless" PIE.

Yes.  I doubt that there is any natlang with only one vowel phoneme, or,
for that matter, none at all.

> > On Sat, 5 Dec 2009 00:19:32 +0100, Jörg Rhiemeier wrote: 
> 
> >One idea I had, but am no longer sure about, is that the thematic
> >vowel once was a transitivity marker which evolved from a 3rd person
> >object suffix identical with the 3rd person perfect ending *-e.
> >I assume that Pre-PIE was an active/stative language in which one
> >set of endings (the later present/aorist endings) were used for
> >agents and another (the later perfect endings) for patients, with
> >transitive verbs taking a patient ending followed by an agent ending. 
> 
> But one would expect, then, some correlation between thematic/athematic
> conjugation and the verb's semantics (or transitivity, at least)?

This is indeed a problem, and the reason why I am no longer sure about
that idea.  If thematicity would correate with transitivity, it would
be as good as a proof that the thematic vowel once was an object marker,
but that is not the case.

> Also, how does it fit with the set/anit thing?

Isn't that a matter of laryngeals?

On Mon, 7 Dec 2009 21:47:52 -0500, Basilius wrote:

> On Fri, 4 Dec 2009 22:44:42 +0100, Jörg Rhiemeier wrote:
> 
> >This kind of connection between PIE ablaut and eastern Siberian root-
> >retraction vowel harmony has been drawn by Greenberg (who also claimed
> >that IE had /e/~/i/ and /o/~/u/ vowel alternations, which actually do
> >not exist), but this is not very convincing. 
> 
> Was it really what Greenberg meant? That is, are /i/ and /u/ above just what
> they stand for in standard PIE reconstructions?

At least, that's how I remember it.

> On Fri, 4 Dec 2009 23:02:36 +0000, David McCann wrote:
> 
> >/e/~/i/ and /u/~/o/ alternations are discussed in 19th century writers
> >like Brugmann; it's their lesser 20th century followers, like Lehmann,
> >who've swept them under the carpet. From my notes,
> >
> >Gk nux, Lat. nox; Gk prumnos, Lat. pro
> >
> >Gk Hestia, Histia; OCS večeru, vĭčera; Skt śiras, Av. sarah; Skt čana,
> >Av. čina; Lat. ex, OCS izŭ; Proto-Slav. *ismi, *esmi. 
> 
> All of these except _Hestia, Histia_ (and "Proto-Slav. *ismi, *esmi", "Skt
> čana, Av. čina" which I don't recognize) seem to illustrate known things:
> vowel accomodation to labialized consonant in Greek; schwa reflected as _i_
> in Skr. but _a_ in Avestan; Slavic *ĭ as a possible reflex of zero grade
> ("OCS izŭ" is P-Sl. *jĭzŭ). The vowel in Lat. _pro_ is probably an 
> inflection. 

Indeed.  These have nothing to do with an "e/i and o/u ablaut", which
is entirely fictional.

> >Consider:
> >Gk peitho, pepoitha: /e/~/o/ < *ə ~ *a, with reduplication in *e
> >matching the RTR *a
> >Gk tithemi < *didəʔmi, with reduplication matching the ATR in the stem. 
> 
> But these seem to be different reduplications; in Greek, one always has -e-
> and the other -i-. And if you don't mean just Greek, you'll need a few "ad
> hoc correspondences" already on the PIE level :)

Right.  These are different reduplications with different meanings:
the _pepoitha_ type forms the perfect, the _tithemi_ type is a way of
forming the present stem.

On Tue, 8 Dec 2009 18:42:20 +0000, David McCann wrote:

> > On Mon, 2009-12-07 at 21:49 -0500, Basilius wrote:
> > >The problem with most current long-range comparison work is that it
> > >relies almost exclusively on lexical data with virtually no heed taken
> > >of morphology.
> > 
> > But maybe that's just how the reality looks - morphologies don't survive 
> > at 
> > such time depths? 

I doubt that, but of course, morphology can be lost.

> Actually a lot do:
> -i plural in IE (demonstratives), Uralic, Chukotian (pronouns)
> -k diminutives in IE, Uralic, Altaic
> -m 1st person
> -n genitive in IE (pronoun), Uralic and Altaic (nouns and pronouns)
> -s plural in IE and Mongolian
> -s desiderative/future in IE and Altaic

At least some of them are likely to be valid.  But the problem is that
we are dealing with short morphemes, which incurs the possibility of
chance resemblances.

> I think the real problem with long-range comparison is a question of
> philosophy, rather like the distinction between "clumpers" and
> "splitters" in paleontology.
> 
> The "splitters" in this case treat everything as unrelated until
> conclusively proven otherwise. I'd say this is rooted in a 19th-century
> concept of multi-linear human evolution.

Which is now championed primarily by racists and nationalists.  Most
scientists in the field now consider multiregionalism disproven.

>       The "clumpers" (this one, at 
> least) argue that everything is ultimately related; we are all descended
> from one woman, so all our languages are descended from hers: we can
> safely rule out any community-adopted conlangs :-)

Yet, it is illegitimate to posit it because of belief in it.  Like any
scientific theory, it must be proven, and many historical linguists are
of the opinion that the techniques we have today are simply insufficient
to deliver such a proof.

>       Hence, the task is to 
> look for evidence of how close that relationship is between various
> groups. IE is clearly closer to Uralic and Altaic than to Amerind or
> Austric, and furthest from Khoisan.

I consider it *very* likely that IE and Uralic are related, and also
quite likely that Altaic is related to them as well, and that the
indigenous languages of the Americas as well as the "Khoisan"
languages (actually three families and two isolates) are only very
distantly related (via Proto-World, which I consider very likely to
have existed but beyond reconstructability) to them.  But I don't
think I will ever be able to prove even an IE-Uralic relationship.

At any rate, the establishment of a language relationship needs an
understanding of the languages involved; it is not sufficient to
compare dictionaries, which is what many long-range comparativists
seem to do.

... brought to you by the Weeping Elf
http://www.joerg-rhiemeier.de/Conlang/index.html