1. Why the glottalic theory doesn't work for PIE (as opposed to pre-PIE):
1.1. The justification for the glottalic theory is wholly typological: the theory explains the absence of *b (interpreted as [p']) and the phonotactic constraint prohibiting roots of the *ged- type (interpreted as [k'et']). On the other hand, the hypothesis is at odds with the comparative evidence of the documented IE languages: the actual reflexes of the proposed ejectives are voiced stops in most of the branches, while the alleged survival of the original phonation type is doubtful even for the Eastern Armenian dialects, the glottalists' pet case. The slightly laryngealized stops and affricates in those dialects are just one particular dialectal development of Old Armenian voiceless stops. They also reflect old plain voiceless stops in loans from Greek and Iranian. Incidentally, there is no tendency to avoid <p> in those dialects.
Since it's a patent fallacy to expect that any reconstructed language must represent the cross-linguistic average, typological expectations can't take precedence over "hard" comparative data. The shift of [t'] > [d] (presumably via an implosive) is possible, but the chances of such a change occurring independently in several branches (Greek, Indo-Iranian, Balto-Slavic, Albanian, Italic and Celtic, not to mention minor extinct groups) are pretty low.
1.2. The constraints explained by the glottalic theory were only tendential, rather than absolute, already in PIE. Morphemes like *ged- were rare but not impossible, and *b was rare but not totally absent. Crucially, combinations like *ded- and *did- occurred freely in reduplicated roots. Gamkrelidze is aware of this argument and has tried to counter it by proposing that the prohibition did not apply to ejectives with the same place of articulation. The ad hoc character of that qualification is obvious: other phonation-type dissimilations are not blocked in reduplicated roots (cf. Grassmann's Laws in Greek and Sanskrit).
2. Why it's hard to believe that *d = [t'] in pre-Germanic
In the earliest Celtic loans in Germanic and in a few Wanderwörter of other origin, original /b, d, g/ ended up as PGmc. *p, *t, *k, e.g.
*ri:g- --> *ri:k- 'king'
*du:no- --> *tu:na- 'fort, enclosure'
*baita: (Gk. [Att.] baite:, [Dor.] baita:, believed to be a loan from Thracian) --> *paido: 'leather coat, jerkin'
*kanabi- (Gk. kannabis) --> *xanapi- 'hemp'
Possibly also in some toponyms:
*da:nu- --> *to:nu- (--> Slavic river-name *ta:nu: > *tany/*tanUv- > Pol. Tanew)
It follows that at the time the borrowing took place, foreign voiced stops were identified with the pre-Gmc. *d series. This is consistent with the traditional interpretation of pre-Gmc. *d as a [d]-type consonant subsequently shifted to *t by Grimm's Law.
Piotr
----- Original Message -----
From: "tgpedersen" <tgpedersen@...>
To: <cybalist@yahoogroups.com>
Sent: Thursday, March 20, 2003 12:23 PM
Subject: [tied] Re: Germanic Scythians?
> I'm not up to date on the status of the glottalsation theory. Why
> doesn't it work for Grimm's law, and why doesn't G & I's version work
> for PIE? Early reports seemed so optimistic.