Hallo conlangers!

On Thursday 11 October 2012 08:16:04 Alex Fink wrote:

> On Wed, 10 Oct 2012 23:49:13 -0400, Alex Fink <[log in to unmask]> wrote:
> > I've heard it asserted that true palatals don't fall into velars, by e.g.
> > people disputing that the values of PIE *ḱ *k can have been [c k].
> 
> Having said that I may as well lay out my own idle (and non-original)
> feelings about the PIE dorsals.
> 
> The system we can comparatively reconstruct has three series *ḱ k kʷ,
> presumably more or less [c k k_w].  In the system *k appears somewhat
> marginal, mostly being a secondary delabialisation of *kʷ; however, there
> exists a set of roots with *k near *a that isn't so compellingly explained
> that way.
> 
> Now, there are two things about PIE phonology that, to me, seem to suggest
> uvulars might've recently been around (in a positive sense, as opposed to
> this silly impossibility of [c] > [k] claim; that part's just a matter of
> timing).  One is the above fact, that *ḱ seems to have been
> unconditionally fronted from an old unmarked **k.

Yes.

>       That kind of
> unconditional fronting mostly happens because of pushing by uvulars (also
> see the bottom of this message).  In fact, in Salishan langs and perhaps
> other languages of the area, we actually see *k kw q qw  ending up as [ts
> kw q qw] and such, so this process can clearly leave rounded members
> alone.
> 
> A second is that the main PIE vowel is *e.  Now "obviously" the main vowel
> "should" have been **a; and that situation looks like what we see in e.g.
> Maltese, where the allophone of Arabic /a/ unaffected by coloring by
> uvularisation and pharyngealisation has come out as /e/, while the
> affected one continues as /a/, with the loss of the consonant contrasts
> which caused the affectation.  Potentially, I suppose, it was only *h2
> which conditioned *a originally, but it feels weird that a single
> conditioner would've been frequent enough to push on the elsewhere variant
> like that.  It could more robustly have been a whole series of uvulars
> containing *h2, one of which would be responsible for those *k near *a
> roots.

Certainly, PIE *e was *a before the rise of ablaut.  A vowel
system */a i u/ makes perfect sense, while **/e i u/ doesn't.

My personal hypothesis is that several vowels collapsed into *a
in an earlier stage (in fact, all except *i and *u), and that
neighbouring velars became front velars when an adjacent
pre-collapse vowel was front (or when no vowel was there), and
labialized velars when that vowel was rounded.  This, however,
requires more elaboration; perhaps comparisons with Uralic are
helpful here.

I call the change described above the "Great Vowel Collapse"
(GVC); this seems to have been preceded by a change I call
"Resonant Conditioned Lowering" (RCL), which means that *i
and *u were lowered before nasals, liquids, semivowels and
vowels in hiatus (explaining the lack of *CeiR and *CeuR
roots).

A possible example:

*kulV- 'to turn' (appears to have cognates in Uralic and Altaic)
> *kol- (RCL) > *kʷol- > *kʷal- (GVC) > PIE *kʷel-.
 
> So I have an old **k kʷ q  becoming *ḱ kʷ k, with **q kinda sparse, and
> I've called *h2 **χ.  This has a gap at labialised uvulars, and I can fill
> it by calling *h3 **χʷ, as after all its colouring effects are like those
> of *h2 plus rounding.

Just what I think about it!  *h3 was a labialized counterpart
of *h2 (and *h1 a fronted one).  Indo-Uralic comparative evidence
seems to point at laryngeals often (but not always) corresponding
to Uralic *k, apparently the same way PIE *s sometimes corresponds
to Uralic *t.

>      This idea has been objected to on the grounds that
> *kʷ (taken by some of these uvular people to have been **qʷ) does not
> colour, so dorsal labialisation can't colour.  But in my system we can say
> that only uvulars coloured, and there may well have been a rarish **qʷ
> which caused *o-colouring then fell in with *kʷ < **kʷ.  Since a
> non-alternating *o could equally be a frozen ablaut variant, the traces of
> this wouldn't be as consipcuous as the *a near *k stuff.  But do they
> exist?

I don't know for sure; this requires investigation.
 
> (There's also this theory that *h3 was voiced.  But the only evidence I've
> seen advanced is *pibh3-something in the reduplication of 'drink', and
> that seems unreliable evidence for anything since it contains a *b and the
> irregular behaviour might instead be due to the process that (mostly)
> killed *b.  And anyway we don't know what the phonations of the stop
> series were this far back.)

I agree with you that the *pibh3- argument is a very weak one,
and I don't see why *h3 should have been voiced.
 
--
... brought to you by the Weeping Elf
http://www.joerg-rhiemeier.de/Conlang/index.html
"Bêsel asa Éam, a Éam atha cvanthal a cvanth atha Éamal." - SiM 1:1