On Wed, 19 Dec 2001 14:20:03 +0100, "Piotr Gasiorowski"
<
gpiotr@...> wrote:
> From: Miguel Carrasquer Vidal
>
> > There are no voiced allophones in O-U.
>
> I wonder how you know that. The spelling wouldn't have shown such a purely allophonic contrast (as it does not in Old English, where <f> = [f/v], <s> = [s/z], <T> = [T/D]). Umbrian rhotacism proves that *s had a voiced allophone in pre-Umbrian.
If they were merely allophones, of course I can't prove anything. But
Oscan and Umbrian consistently use F, never B or D. Pre-Italic must
have had *f, *T, *x and *xW in all positions. /T/ > /f/ is a natural
development. /D/ > /f/ is not.
> > I accept *kH > x for Slavic (soxa ~ 'sakHa:). The argument itself is of course typological: we cannot have /dh/ without /th/. The Balto-Slavic evidence for *dh is even more indirect (it didn't lengthen a previous vowel like *d did, so assuming the lengthening was caused by the voicedness of *d, *dh must have been less voiced, as murmured stops are).
>
> The evidence supporting Winter's Law is indirect but _systematic_, while <soxa> : <s'akHa:> is just an isolated equation -- definitely not enough to establish a "correspondence". The form is otherwise OK (but not so OK if you want to derive <s'akHa:> from *k^nk-h2-), and the semantic match looks good. However, we have even more impressive semantic matches in pairs like "day : dies" or "deus : theos", and yet these equations are patently false.
There are some other examples of *kH > x (*ple^sI/*plIxU, *xorbrU),
besides some examples of *kH > k. Perhaps *kh2 > x and *kh1 > k, but
there is not enough material to be sure.
> The typological argument is only valid if you insist that PIE (or Proto-B/Sl) {dH} was distinctively aspirated, as in Indo-Aryan (that is, if the onset of modal voice was significantly delayed after its release), where {tH} appears as expected on typological grounds. But breathy voice is possible without audible aspiration. This, of course, is what makes the PIE triad (without phonemic aspirated {tH}) possible.
Ladefoged & Maddieson's discussion mentions that the breathy voice
stops in e.g. Owerri Igbo "appear to have a shorter breathy voiced
period, and to have stronger voicing than the corresponding sounds in
Hindi and other Indo-Aryan languages". But nowhere do they say that
breathy voice is possible without audible aspiration, nor do they
distinguish between "breathy voice" (a.k.a. "murmur") and what has
traditionally been called "voiced aspirates" in IE linguistics.
Therefore, the PIE triad is in my opinion impossible. Only the
emergence of voiceless aspirates (C+h2 or C+h1, apparently not C + h3
(?)) made it possible, at least in an (Eastern) dialectal area.
It is worth noting that the two branches that split off from PIE
earliest, Anatolian and Tocharian, do not have voice as a
distinguishing feature in their stop systems. This must represent the
PIE situation, and we can assign the values *dh = [th], *d = [t] and
*t = [t:], almost precisely as shown by Hittite orthography in medial
position (*dh and *d = <t>, *t = <tt>).
When the fortis-lenis contrast was transformed into a voiceless-voiced
one, the triad could naturally develop in two ways: (A) [t:] develops
the voicing contrast with [t], giving *t = [t], *d = [d], and *dh
remains as [th], or (B): [t:] contrasts with [th], giving *t = [th],
*d = [dh] and *d remains as [t]. If we add *tH to the equation, it
has no effect on (B) [where it simply merges with [th] = *t, except
cases like Armenian *kH2 = [kx], not [kh] > /x/)], but it transforms
system (A) into Brugmannian *t *th *d *dh.
=======================
Miguel Carrasquer Vidal
mcv@...