> > > A similar case would be IE *penkWe- '5' ~ NEC *fimk?wV 'fist'. If I'm not mistaken, Petr suggested that Starostin's f should be replaced by X\W or XW.
> > >
> > What are the attested words on which this NEC reconstruction is based?
> >
> See here: http://newstar.rinet.ru/cgi-bin/response.cgi?single=1&basename=/data/cauc/caucet&text_number=1008&root=config
This example seems to come up frequently around these parts, so I took a look at the other examples listed as containing PNC *f- here.
I agree that a labialized back spirant of some sort seems like the better reconstruction here. The only subfamily showing reflexes more front than velar is apparently Avaro-Andian, and then only for the cluster (?) "*f_h-".
Worse yet though, essentially no regularity arises.
- West Caucasian shows two reflexes, both *xW-.
- Nakh shows two reflexes, one *H-, one *X-.
- Avaro-Andian shows five reflexes (out of five altogether), including three cases of *xW-, one *x-, and one *h-.
- Tsezian - 3 reflexes, all different: *Xw- *X- *h-
- Lak - 2 reflexes, both *h-
- Dargwa - 3 reflexes, two *X:-, one *x:-
- Lezgian - 5 reflexes, two *X:-, one *Xw:-, two *Xw-
Including cluster and medial cases only seems to complicate the picture (introducing reflexes like Lak -X-, Tsezian -x-), but it seems this "*f" is, despite the assertion of regular correspondences, nothing more than a grab bag of comparisions difficult to fit together with the more securely reconstructed comparisions. Even if there IS something here, it seems that most likely *xw- should be the regular AA reflex, which means the PEC status of "fist" (which has h-) will have to remain dubious.
_j.