>Is reconstructed PIE supposed to be exactly as related to one branch as it
is to any other? >That expectation seems a bit silly to me.
And rightly so. But depending how you count, there are 6 - 12 main
branches, and it would be equally silly to reconstruct PIE without including
all of them, and also silly in a purely statistical computer algorithm to
include 20 examples from one branch and only one from another. Real
linguists can take into account the fact that evidence from Italian and
Spanish cannot count as heavily as evidence from Italian and Russian, but a
simple statistical computer program can't. Whe we learn that the entire I-I
branch was omitted, we have some questions to ask about the validity of any
result.
Peter