It must be noted that the tree was derived using a large number of
morphological and lexical characters. However, vocabulary and
morphological patterns diffuse easily, so some of the "clades" in the
tree may be nothing more than areal clusters. Old English and
Armenian, in particular, seem to be misplaced because of the failure
of the method to distinguish between genuine synapomorphies and
diffused characters. The "horizontal" spread of grammatical and
lexical traits from language to language can be quite massive, and
it's wrong to assume that all types of characters are equally valid
as evidence. My suspicion is, for example, that the more lexical data
you feed into the algorithm, the more "areal" and the
less "phylogenetic" cladogram will result. I'd be reluctant to accept
any genetic grouping that is not supported by a single phonological
character (the less commonplace the better), unless the case is
really well argued. Because of its nature, phonological change is of
greater probative value in determining genetic relationships that
_any_ other kind of evidence.
I'll upload a schematic phylogeny showing my own views on the IE
subgroupings to the group's Files.
Piotr
PS. My e-mail server is temporarily out of service; I have to post
via the group site, which is not very conveniet, so I'm not as active
in this discussion as I'd like to be.
--- In cybalist@..., x99lynx@... wrote:
> This is the later IE tree that Ringe et al generated by computer
using a
> larger group of shared and unshared characteristics than in the
earlier tree.
>
> You can see this tree at http://www.cis.upenn.edu/~histling/