On 2006-10-01 08:59, P&G wrote:
> To repeat myself, root "constraints" are only tendencies, not rigid
> inviolable rules.
That should go almost without saying, especially if the constraint was
the consequence of some special phonetic properties of pre-PIE rather
tha PIE itself (and if most PIE roots were inherited from those times).
If we accept the traditional reconstruction of the mediae as
more-or-less plain voiced stops, it's hard to see what could have
prevented them from co-occurring. Indeed, nothing prevents them in
reduplications, though typical Grassmann-type dissimilatory processes
operate normally in such environments, cf. Skt. dádHa:mi, Gk. títHe:mi.
I just agree with Torsten that it doesn't seem impossible for a
"neo-root" to follow a conspicuous dissimilatory pattern even if it is
no longer phonetically grounded. I have myself toyed from time to time
with the idea that the PIE 'thrush' word comes from *dru-sd-o- >
*truzdo- (with various branch-specific reshapings), describing the
bird's habit of singing from a high perch; cf. RV dru-sad- 'sitting in a
tree'.
> There are 9 fairly certain roots of the form DeD in PIE:
> *bed to swell Pokorny p96
> *bend tip Pokorny p96
> *de:g to grasp Pokorny p183
> *g'eid to suck Pokorny p356
> *geid to stab Pokorny p356
> *geig' to stab Pokorny p356
> *gleg soft Pokorny p 401
> *gred to sratch Pokorny p 405
> *gWreig' to sleep Pokorny p485
"Fairly certain" is, however, an overstatement with regard to at least
some of them. For example, *gleg^- (sic) 'soft, tender' is attested from
Balto-Slavic and North Germanic only (a distribution that hardly
guarantees PIE status), shows no Winterian effects in BSl. (which casts
doubt on the reconstruction of a final media rather than media
aspirata), and the Germanic cognates have either -nk- or -kk-, never -k-
alone (in a root which should not invite a nasal infix, so *-g^Hn- >
*-kk- by Kluge's Law and *-nk- by pre-nasal strengthening and pre-Kluge
metathesis is at least a possibility). So, if a common proto-root can be
reconstructed at all, the optimal solution is something like *gleg^H-.
Piotr