Re: [tied] Re: PIE's closest relatives/SIBUN

10-01-04 10:40, P&G wrote:

>> > > And we can add to this oddness the greek hebd- besides hept-.
>> Of course. So, there is a triplet of variants: *septm / *sepm / *sebdm.
>
> The Greek voiced form may simply be by analogy with the voiced forms in
> "eight": ogdoos etc. The origin of the voicing in 8 remains obscure, but
> laryngeals have been dragged into it. It is one of the bits of evidence
> sometimes used to prove that h3 was voiced. The presence of a laryngeal at
> the end of 8 seems clear (Sanskrit shows -au and -a, and cases with
> both -a- and -a:-), and the ending -o: points to -h3.

Slavic *sedmI '7' and *osmI '8' are both irregular; the latter owes its
*-mI to being paired with '7' (cf. also the almost complete assimilation
of *deve~tI '9' to *dese~tI '10', and similar pairing in East Baltic).
The origin of Slavic *-d- in *sedmI is unclear, but given Baltic -pt-
I'd say that it's a late development. On the other hand, Kashubian
setmë, which has been considered an archaism by some is more likely a
case of final devoicing (in monosyllabic *sedm < *sedmI in
Proto-Kashubian). It has been proposed that the voicing comes from the
ordinal *sedmU < *sebdm(.)o-, but I'm not convinced that the combined
evidence of Slavic and Greek suffices to reconstruct old voicing in the
ordinal. It is worth noting, in any case, that there is no voicing in
Slavic *osmU '8th'.

Piotr