Gray & Atkinson (Nature)

I'm quite impressed after a cursory initial reading of the article. It's
far better than any computer-aided linguistic phylogeny I've seen
before. If it were not for the position of Balto-Slavic as a sister
branch to the Germano-Italo-Celtic clade and of Greek and Armenian as
sisters (thus making Satem a polyphyletic set), and the position of
Polish as a sister lineage to East Slavic, it's very similar to a tree I
might sketch myself. All these controversial groupings, however, are
rather weakly supported by the data and my guess is that they are
mirages produced by areal diffusion (the same may be true of the
position of Frisian). The exclusive use of lexical data (cognate sets)
makes the method sensitive to this kind of distortion, since words are
easily borrowable and old loans may be indistinguishable from shared
inheritance (and can't be removed from the analysis). The possibility
that Germanic may be more closely related to Italic than to Celtic is
something I've been thinking about for some time. Also the proposed
dating of splits is roughly consistent with my own ideas. A date of ca.
6500 BC or slightly earlier for the separation of Anatolian is what I
have long advocated. However, I derive the date from the hypothesis that
the expansion of the Linear Pottery culture north and west from the
Danubian area should be correlated with the spread of non-Anatolian IE,
so my scenario doesn't start in Turkey.

Piotr