--- In cybalist@..., Piotr Gasiorowski <piotr.gasiorowski@...>
wrote:
> Let me return to this affiliation question (see also
http://www.cis.upenn.edu/~histling/ ).
>
> To begin with, as regards the UPenn family tree, I entirely share
Jay Jasanoff's scepticism. The choice of linguistic "characters" for
a cladistic analysis is an extremely tricky business. I'm
particularly suspicious of the use of grammatical and lexical
innovations, which diffuse easily. Phonological innovations (the more
uncommon the better) are much better indicators of genetic
relatedness. The problem in linguistics (as opposed to biology, where
the effect of horizontal diffusion between branches is practically
negligible) is that the number of such reliable characters for a
given family may be frustratingly limited, and that if we try to make
up for that by using less reliable ones as well, it becomes
impossible to rule out areal interference. I agree with Jasanoff that
the unstable behaviour of Germanic (and the indeterminate position of
Albanian) reflects a fundamental weakness of the method, not the fact
of Germanic being a "special case". Its allegedly close relation to
Balto-Slavic is an illusion, and the "consistent explanation for
Germanic" is a face-saver proposed post factum by the humans involved
in the project, not a computer-generated result.
>
That's not nice. They make the assumption all branches stayed put,
model doesn't work, they assume one of the branches changed
neighbors, model works. If that's a "face-saver", what is then the
proper of correcting one's erroneous assumptions?
> Despite "the dizzying combinatorial explosion" of possibilities the
results are quite unexciting. Don Ringe may have been surprised by
the primary split producing Anatolian, but I suppose most historical
linguists would actually have been surprised if Anatolian had _not_
branched off so early. Nobody doubts that Baltic and Slavic or Indic
and Iranian are closely related; a special Italo-Celtic relationship,
while debatable, is also an old idea. So is Graeco-Armenian, except
that it has very few supporters now.
>
OK, mr. Ringe is surprised where he shouldn't have been. Not good.
> Though I doubt if the Aestii could really have been "Celtic"
If (at least) the leaders of the Cimbri and Teutones (note: no
Verner) in Jutland were Celtic, why not?
(who they were is a different question; West Baltic, perhaps), it's
thinkable that there were some non-Germanic centum enclaves along the
Amber Road even as late as the first century. Some of the hydronyms
of the Oder system (including "Viadua" and "Adora") and along the
Baltic coast can't have been adopted by Germanic speakers before
Grimm's Law (they don't show the shift), which may mean that they
were used by speakers of some other language long enough to be
eventually absorbed by Germanic in unshifted form.
Nice, if the Grimm shift took place in the invading Bastarnean-
speaking army around 0 CE under the impression of similar
developments (before consonant) in the language of their Iranian-
speaking leaders.
The language in question was not Celtic either if names like Parse,ta
(< *pars-ant-) are anything to go by (note the un-Celtic initial *p),
but it might have been some kind of para-Celtic or para-Italic, or
indeed para-Germanic (without Grimm's Law; by "para-" I mean "closely
related to but not identical with").
Judging by pwmp "five", Welsh is not Celtic either. Suppose Aestic
was p-Celtic?
Another idea: Iranian?
>
> Piotr
>
Torsten