On Tue, 27 Mar 2001 04:17:48 +0200, Miguel Carrasquer Vidal
<
mcv@...> wrote:
>The parallel with Indo-Iranian
>s.as.-tí, ..., nava-tí- "60..90", made with a fem. collective *-ti-,
>however attractive, is impossible because of Grimm-Verner (we would
>expect Goth. *sibunde:-hund and OHG *[hund-]sibundeo), and there
>exists no collective suffix *-di-, as far as I know. Supposing an
>ad-hoc lenition of *t > *d would maybe be too ad-hoc, even given the
>lenitions that IE numerals are prone to (Lat. -ginta, Grk. hebdomos,
>ogdoos etc.).
Come to think of it, maybe not *that* ad-hoc. The numerals in
question are 7, 8 and 9, where lenition *is* found, not only in Greek
hébdomos, ógdoos, but more importantly in Slavic sedmI < *septm-ti
(and analogical osmI < *osdI?). So maybe for Germanic we *can* posit
lenited *sep(t)m-di- > *sibunt- and *ok^tu-di- > *ahtot-, plus
analogical *neun-di- > *niunt-. This means that the ordinals Goth.
*sibunda, ahtuda, niunda, with the proper reflex of *-t-, are to be
taken as later regularized innovations, but that's to be concluded
anyway (the Latin pattern septimus, octa:vus, no:nus is the archaic
one).
=======================
Miguel Carrasquer Vidal
mcv@...