Re: [tied] Thoughts on the existence of *H1

From: Piotr Gasiorowski
Message: 9478
Date: 2001-09-14

----- Original Message -----
From: Glen Gordon
Sent: Thursday, September 13, 2001 8:20 PM
Subject: Re: [tied] Thoughts on the existence of *H1
Cleverly argued!

> I think we agree that syllabic *r must surely be *[&r] and is ergo of a VC-form. So *CrC means *[C&rC], a CVCC-syllable, **not CVC**. Pretend syllabic liquids never existed.

I'd question "must surely", since syllabic liquids (and nasals) no doubt exist (did you see that Czech song I posted? The Czechs have words like [krk], [trh], [prst], [vlk], and no schwas at all). With regard to phonological processes like Edgerton's Law PIE *-CrCV- is treated as a sequence of two _light_ syllables (-CVCV- rather than -CVC.CV-). I have no problems with the idea that syllabic liquids and nasals existed in PIE. Otherwise it's difficult to explain why, e.g., in Greek or Sanskrit *-VN- combinations are retained, while syllabic nasals become [a] (not [an] or the like).
Of course in _most_ individual branches syllabic consonants came to be pronounced asynchronically as *-VN-/*-VR- or *-RV- combinations, but even so divergent *V values show that this development is post-PIE. E.g. syllabic *r happens to be reflected as such in Indo-Aryan (and possibly in Old Iranian), but otherwise gives Hittite and Armenian *ar, Italic *or, Germanic *ur, Celtic and Albanian *ri, Balto-Slavic *Ir/*Ur, Greek *ar/*ra, etc. It's far less trouble to derive all these reflexes directly from syllabic *r than from anything like *&r.
> So what if we find *-CerH1C-? This should be a trivial case as well. Here we have *CVCVC as always, but since *H1 equals the unmarked vowel *[&] in this case, "full" vowels like *e or *o take precedence over the schwa. Hence here, we should most often find *-CerH1C- > *-Ce:rC-.
Which is precisely what we don't find, except in Balto-Slavic; but there we also have *-Ce:rC- from *-Cerh2C- [-erx-] -- the amazing coincidence I mentioned. If you assume that compensatory lengthening is to be expected in *-C&r&C-, it's hard to see whay it doesn't happen in *-C&reC- or *-Cer&C- in the same branches. 

> Now for the attack... If *H1 were truely either *[?] or *[h], should we not find cases where *H1 has affected a preceding stop? If it were *[h] we should find sequences like *dH1 becoming *dh, nicht wahr? If a glottal stop, we should find correlations with Kartvelian ejectives in loans that demonstrate such a thing. If the latter were true, it would facilitate the development of ejectives in IE just like its neighbours, yet we simply don't see such a thing aside from highly tentatively in Armenian.
This is a valid argument. Alas, cases of aspiration in *-Ch1- combinations may be accidentally difficult to find, since *h1 is rarer than *h2 in positions where stops might be affected. In *dHeh1- *dH is already aspirated, and *deh1- or *geh1- (not to mention *beh1-) don't seem to be attested. I'll see if any examples can be found (maybe before the *-h1en- suffix?).