> There's no Centum/Satem division, but arguably there is a Satem taxon
within IE. Since it >isn't a primary subfamily, its existence doesn't entail
the establishment of "non-Satem", i.e. >"Centum" as a valid genetic unit.
The existence of a Satem group based on a common innovation in its members
would indeed be as you say, Piotr, but that is not quite what we have for
PIE. There is a Centum innovation as well - the collapsing together of
palatal k' and plain k.
We can in fact see two parallel innovations:
Collapsing k' and k over against kW (Centum)
Collapsing kW and k over against k' (Satem)
The probably later sibilisation of k' is not what actually makes the Satem
group, even though it gives it its name. Thus the Centum groups has as much
right to be called a "taxon" as the Satem.
Peter