--- In
cybalist@yahoogroups.com, "Brian M. Scott" <bm.brian@...> wrote:
> Sally Thomason has a nice skeptical piece on this in
> Language Log:
>
> <http://languagelog.ldc.upenn.edu/nll/?p=4612#more-4612>
It's a lousy riposte. The counter-arguments are so holey that it almost reduces one to ad hominem arguments.
She complains that using Eskimo instead of Eskimo-Aleut makes the study unlikely to be useful. In fact, so doing should merely make the study less likely to come up with a positive result. (I say 'should' because adding weakly informative data has the curious effect of weakening the conclusions of statistical exercises in linguistic phylogeny.)
The validity of her complaint about using Altaic remains to be determined. If, as is quite possible from what I've seen in discussions of core Altaic, the relevant words turn out to be Turkic, then there is no problem.
The weakness is liable to be in the statistics. Having read the paper, it's not at all clear which or how the per-group reconstructions were chosen, or how inter-group cognates were identified.
Richard.