Re: [tied] Olsen's Rule [was: Ploughs]

From: Piotr Gasiorowski
Message: 29874
Date: 2004-01-21

21-01-04 02:18, Jens Elmegaard Rasmussen wrote:

> Hey, you don't *really* mean "*-dHlom" with a voiced aspirate? Has my poor
> wife Birgit Olsen been preaching in vain that the IE alternation was
> *-tlo-/*-thlo- (after roots without liquids) ~ *-tro-/-thro- (after roots
> containing a liquid), in both cases with the aspirate restricted to the
> position after non-syllabified laryngeals H1 and H2?
>
> Balto-Slavic has generalized -l- and so is of little value for the
> assessment of the distribution of the liquids. Since Baltic has only
> *-tl-, I would suppose Slavic -dl- is the product of IE *-tl-.
>
> There are no examples of *-dhlo-/*-dhro- following a voiced aspirate known
> to me. The reconstruction with a voiced aspirate seems only to be a myth.
>
> Don't get the tenor of this note wrong. Birgit's thesis is not above
> criticism, so if she is wrong, let's hear what exactly is amiss.

I'd like to discuss it, actually. It's ingenious, but I have my doubts.
One uneasy aspect of this explanation, as far as I'm concerned, is that
it depends so heavily on the testimony of Latin (<po:culum> vs.
<pa:bulum>), but doesn't work consistently even there. Apart from the
*-tlom- suffix and the suggested interpretation of <ple:b-> (and Gk.
ple:tHu-) as *pleh1-tw- (I have to say I particularly like this one), is
there any independent reason for thinking that *tH developed into a
voiced sound in Latin? One has to resort to analogy to explain away
quite a lot of counterevidence. OK, let's admit that *mah2te:r and
*bHrah2te:r may owe their unaspirated *t to the analogy of *p&2te:r and
*dHug&2te:r (already in PIE). But why don't we ever find *-to:r or
*-tr-ih2- with reflexes of *tH, in Latin or any other language where
they could be seen (Greek, Sanskrit)? Lat. terebra apparently shows
vocalised *&1 (if what we see in the second syllable is a laryngeal at
all). What is it analogous to? *tr.h1-tro-, unattested in Latin? What
about <stabulum>, where vocalised *&2- must be assumed?

Slavic has generalised *-dlo-. Since it's hard to believe that Slavic *d
should have developed out of *tH, you seem to be suggesting that *-dlo-
is a reflex of *-tlo- (with assimilatory voicing?). However, apart from
this morpheme, old *tl survives unchanged in Slavic, and even in the
instrumental suffix we have occasional traces of *-tlo- in words with
obscured morphological divisions: *veslo 'oar' (which I'd suggest comes
from *vez-tlo < *weg^H-tlom), *c^islo 'number' < *c^it-tlo, *teslo
'adze' *tes-tlo < *tek^T-tlom (there are only isolated traces of
*-tro-). Isn't it more realistic to assume that Balto-Slavic had both
*-tla- and *-d(H)la-, and that either the one or the other was
generalised in the descendant groups? Which of course would mean PIE
*-tlo-/*-dHlo-.

Piotr