On Wed, 26 Feb 2003, Glen Gordon wrote:
> [...]
> The ejectives would have first eroded to *t: and *k:, perhaps
> being spoken as unaspirated [t] and [k]. Then, they would become
> voiced *d and *g in a later stage. Throughout, the absence of the
> correlating labial stop would remain constant until the fracturing
> of IE when various dialects would adopt their own *b's.
This is a point where I agree almost entirely with Glen. If the glottalic
values belong to some prestage of the protolanguage, they will still leave
an uneven system, i.e. a language with *very few* cases of /b/. I find it
relatively easy to imagine that some b-words did come into being in the
final prestages of the protolanguage. Since there are also processes
creating voiceless aspirates, the result was a stop system much like that
of Sanskrit, which is certainly a state of affairs permitted by typology.
The exact route from *t? to *d does not matter much - I have always
pictured a change from ejective to implosive, i.e. [t?] > [?t] which, I am
given to understand, is often accompanied by voicing. The final steps
would then be [?t] > [?d] > [d]. This is what has been observed as a
frequent line of change by Haudricourt.
The important thing is of course that we do not have to invent
independent, but identical, Lautverschiebungen for the majority of the
languages: IE /d/ was [d] in PIE, and there is no change when we find it
as Skt. [d], Gk. [d], Lat. [d], Celt. [d], BSl. [d], Alb. [d]. For
Anatolian and Tocharian, Sturtevant's law and the loss in Toch. clusters
certainly point to a lenis consonant, so there are only Germanic [t] and
Armenian [t(?)] to cast a vote for a voiceless fortis. And if both harbor
loanwords reflecting pre-shift phonetic values, not even these witnesses
can be called.
Jens