Piotr:
> See what a mess Old English velar/palatal obstruent were, in
> phonological terms; I wouldn't expect much neatness of PIE either.
Alright. Thanks. However, while I must agree that "the comparative
evidence takes precedence over any a priori expectations of phonological
symmetry", I thought that the evidence only _hints_ to us that *h3 is
likely to have been voiced. This still doesn't show us what position
it was in either. Hence, in the absence of enough comparative evidence,
I feel the urge to side with phonological symmetry to keep things
grounded. This is why, I can go so far as to accept a voiced *h3 if one
must, but /G/ is not my first choice.
Without voicing, the best choice appears to be /hW/ based on not only
symmetry but also the rounding effects *h3 has on *e. Since we already
have the *k-*q-kW series, it stands to reason that we should have
*h-*x-hW as well. I've already explained how *hW could have rounded
neighbouring *e independently in the various dialects during post-IE
times as it degraded to null. This appears to be the optimal theory
if we dismiss the "bibere"-evidence.
However, if we accept a voiced *h3 as virtual fact, the optimal theory
appears to be just the above conclusion with added voicing. Hence /HW/,
not /G/. In this way, we account for everything based both on evidence
AND at least partial symmetry. Without symmetry, imagination gets the
better of us, I think, because there's nothing to ground us when trying
to make a sensible choice from the existing possibilities. Sure, /G/
is "possible" as well, but I don't see how it's the best choice over
/HW/. It doesn't come as close as /HW/ in both explaining the evidence
and conforming as well to logical symmetry.
Conforming to symmetry doesn't mean that I'm ignoring the many
assymetrical languages out there, but we can't prefer chaos over order
here! We stick to a pattern no matter how "simplistic" until we have
evidence to proceed further. That's always been my methodology.
> Such things happen -- cf. the change of Slavic *g into fricative /G/
> and further into breathy-voiced glottal /H/ in Czech or Ukrainian,
> leaving an ugly gap in the stop system.
Duely noted. I think of the lack of "s" in Hawaiian too. It happens
but I still don't see how that should affect our pursuit of the
_optimal_ choice. We should be siding with symmetry in cases where
nothing else can be used to extract a single answer. If we have a
choice between, say, /G/ and /HW/, what else can be used to determine
the best course of theory except a logical pattern?
> The problem is that the /h/ in <ahead> becomes "voiced" (actually,
> breathy-voiced) by assimilation. This kind of allophonic voicing cannot
> spread to any adjacent segments (the latter must be voiced in the first
> place to make /h/ voiced!). In order to have caused the effect we're
> talking about *h3 must have been _distinctively_, not _contextually_
> voiced.
Oh yes, of course. But what I meant was more intended for the issue of
pre-IE stages, not IE itself. I feel that there is a pre-IE reason as
well to side with a laryngeal system that mirrors the velar system.
Both these series should have a plain-uvular-labial pattern at some
point in IE's past, if not in IE itself, so I figure. This is because
the very reason for labialized phonemes in IE would strongly appear to
be due to an early reduction of the vowel system. Others have discovered
this long before I was born, but I have to agree with that idea because
it can account for why we see such an overabundance of *e or *o without
many examples of *a. A two-vowel system for pre-IE seems unavoidable.
An early syllable *kuC would become *kWeC, and likewise *huC transforms
into *hWeC. This is evocative of the similarly vowel-reduced Abkhaz-Adyghe
languages which I sense evolved the exact same way, and perhaps at the
same time through areal influence. Allan Bomhard already has hinted at
such a possible pre-IE contact with (pre-)NWC but this is naturally
unproven as yet.
In line with what you've just stated but in keeping with my current
views of IndoTyrrhenian, I'm perfectly comfortable in accepting the
idea that *h3 = /HW/ (or even /G/ if further evidence shows this is a
better idea) while continuing to reconstruct *h3 = voiceless [hW] for
the earlier MIE stage. My idea so far is that some time around Syncope
(the beginning of the Late IE period), the laryngeal system also changed
such that plain MIE *h was uvularized to *x. The reason for the "Laryngeal
Shift" is probably because the laryngeal series, while at least
originally mirroring the velar series, lacked a uvular counterpart.
In Old IE, the laryngeal series would have been *?-*h-*hW (no uvular
phoneme **x, although certainly uvular _allophones_ of plain *h and
*hW).
You see, if *k was uvularized to [k.] when found next to *a in MIE and
earlier, it's reasonable that *h was likewise uvularized to [h.]. So far
it looks like it's not that simple and that uvularized [h.] became null
or *w in early stages of IE. So, this is what I have for the laryngeal
series for each stage, tracing the evolution of the phonemes from Old
to early Late IE.
OIE MIE eLIE
-------------------------------------------
*? *? *?/*-h-/*-h
*h [h] *h *x
*h [h.] ZERO/*w ZERO/*w
*hW [hW] *hW *hW
*hW [h.W] ZERO/*w ZERO/*w
In other words, OIE has the series *?-*h-*hW. MIE has the same laryngeal
series, still lacking a uvular laryngeal, but it loses some uvularized
allophones of *h and *hW along the way due to erosion. Uvulars, being
automatically rounded to begin with, merge with *w in many instances.
Then Late IE innovates by aligning the laryngeal series better with the
velar series by shifting former *h to a uvular all-round. This shift
encourages mediofinal instances of *? (*h1) to weaken to lazy aspirates.
Believe me, I started out with a simpler system but this is where it took
me <:)
So MIE *hW then would simply remain *hW in early Late IE while becoming
voiced sometime between 5000 and 4000 BCE, let's say. Nifty!
= gLeN