--- In
cybalist@yahoogroups.com, "mkelkar2003" <swatimkelkar@...> wrote:
> http://en.wikipedia.org/wiki/Indo-Uralic_languages
> http://brettkessler.com/McDonald/paper/Kessler--Multilateral.pdf
> "The other piece of bad news for the multilateralist research program
> is that this more rigorous version of the methodology failed to turn up
> any connection between Indo-European and Uralic (Kessler & Lehtonen,
> 2003 or later)."
Actually, I don't accept that the final comparison between IE and
Uralic in Kessler and Lehtonen was multilateral. The final step is
necessarily a bilateral comparion of IE and Uralic. That such a
comparison should be unconvincing should surprise no-one.
I believe the power of the statistical test is weakened by the method
of combining languages (partly?) given in the paper. The usual wisdom
is that when comparing the words for a meaning in two groups A and B,
we should compare the protoforms from each group. If all the words in
each group are compared, and half the languages in each group have
replaced the proto-form with their own independent words, then one
effectively gets a quarter of a match (plus positive noise bias)
instead of one match. This is a case of reaching down demonstrably
weakening an argument! Unfortunately, I don't have an alternative way
of weighting the forms from the different languages in each group.
Note also the statement in the paper, "We started with the Swadesh 200
list (1952) but omitted the concepts that are typically not fully
lexical: `and', `at', `because', `few', `he', `here', `how', 'I',
`if', `in', `not', `some', `there', `they', `this', `we', `what',
`when', `where', `who', `with', `ye'". That in itself eliminates
about half a dozen IE-Uralic matches.
I hope there was no biasing effect from stripping out presumed loans
from Germanic from the Uralic vocabulary and then having 4 Germanic
languages out of the 11 IE languages - or from stripping out
Balto-Slavic words and having 2 Balto-Slavic languages in the IE group.
Richard.