*** MERRY CHRISTMAS ***
[Piotr]:
<<If Dolgopolsky has a hundred really convincing equations, why doesn't he
publish THEM in the first place? I'd love to see some of them. What he DOES put
forward is what I'd call poor-quality stuff. Is he hiding his light under a
bushel?>>
[Alexander]:
A reasonable comment.
[Piotr]:
<<If in somebody's list of etymologies only 5% look good, they may well
represent the expected occurrence of chance similarities or ancient loanwords.>>
This is the question I have been thinking about. Is there any quantitative
criterium to apply and thereafter to say: "The probability, that the presented
massif of potential cognates can be explained just as a coincidence, is less
than 5%" or "Such a probability is higher than 10%, therefore the hypothesis of
common origin (due to kinship or loans) is not proved"? Whithout such a
criterium we have to relay only on expert estimations of linguistic authorities
but the opinions differ...
Piotr, have you asked yourself why 2 groups of scholars come in this case to the
opposite conclusions analizing the same evidences? I don't belive that it caused
by the malicious intent or the ignorance of one of the groups. Nobody argues
whether the Hittite language belongs to the IE family or not. The answer is
obvious for (almost?) everybody. But when we consider connections BETWEEN
families the meanings split. Can't it happen because one of the scholars groups
tries to apply here the same criteria as for analyzing the language relatedness
INSIDE a family? Look, if we use as THE ONLY index of genetic kinship the degree
of similarity inside a linguistic group (say, Slavic) we would have to deny
Grimm's law and recognize that different groups inside "so called IE family" are
unrelated! Maybe a new methodology for interfamily investigations should be
worked out?
[Piotr]:
<<As for the "regular" sound changes, they turn out to be less regular if you
look at the etymologies in detail and see how much liberty Illich-Svitych and
Dolgopolski take with individual items. Dolgopolsky uses so many cover symbols
for classes of phonemes that occasionally he gets confused himself. As for
Nostratic *l, it sometimes gives Dravidian *t as well; *r also suffers an
unconditioned split in Tungusic (*r or *n).>>
I can add here an argument against my position. If a large part of word
comparisons is wrong it will prevent obtaining correct sound correspondences.
The ratio signal/noise will be too small.
[Piotr]:
<<To account for various correspondence patterns Nostratic consonants are
multiplied beyond belief (of course ANYTHING can be compared with ANYTHING ELSE
if your inventory of protophonemes is sufficiently large). The current version
of Nostratic has 50 (FIFTY) consonants, including eight coronal fricatives and
twelve coronal affricates.>>
This confuses me. It resembles the situation in theoretical physics where more
and more new particles are postulated to bring experimental results in
accordance with the theory.
Nevertheless believing in the Nostratic theory,
Alexander