the UPenn trees /PIE chronology

X99Lynx at aol.com X99Lynx at aol.com
Mon Oct 18 05:20:53 UTC 1999


On Thu, 7 Oct 1999, Stanley Friesen (sarima at ix.netcom.com) wrote:

<<I question the assumption that the Anatolian languages necessarily split
off earlier than the others.  Given the linguistic and archaeological facts,
I suspect that the northern European languages, and probably Proto-Tocharian,
split off at least as early as Anatolian.  Thus I see an original three or
four-way split, not a simple bidding off of Anatolian...>>

Sean Crist replied (message dated 10/15/99 4:36:35 AM):
<<You must be new on this list; during August and September, the question of
the phylogeny of the IE languages was hashed over in great detail.  I'm
curious as to what linguistic and archaeological evidence you have in mind;
your claim is at odds with the work of Ringe, Warnow, and Taylor which I have
been discussing.  If you like, I can forward you some of the emails which I
sent out as a part of this dialog.>>

Just a couple of notes:

1. Whatever was "hashed over in great detail about the UPenn tree," there
wasn't much if anything that necessarily contradicts Mr. Friesen points.
Since the UPenn tree provides only "relative chronologies" there's nothing
that conflicts with Mr. Friesen's apparent point which might be that PIE
dispersed first and then "split up" into different languages or groups of
languages.

2. The "phylogeny" (more properly "trees") produced by the UPenn group are of
two different kinds -

One is the so-called "pure phylogeny" which does NOTHING but "measure of the
consistency of the historical linguist's linguistic judgements,..."  Note
that the measure is of "CONSISTENCY" not ACCURACY or VALIDITY.  What the
"pure phylogeny" does not do is say whether any theory (Ringe's or someone
else's) is correct or not.
As the web site that reports on this tree explains the methodology can be
used "to quantify the support for different evolutionary hypotheses."  What
it basically does is rearrange pre-selected data until it matches the
methodologist's notion of "consistent" groupings
of data - called "convexity" which essentially means nothing more than that
common PRE-SELECTED data should fall together in certain favored
sub-groupings.

(BTW, I'd strongly urge Mr. Friesen or anyone interested on the list to go to
website itself for a more through and accurate explanation of the method than
is available in any past postings on this list - pace Sean Crist.  The site
begins at http://www.cis.upenn.edu/~histling/home.html
If anyone has a problem opening or printing the papers - the files are
unfortunately in postscript form - let me know and I may be able to provide
you with Acrobat or text versions.)

3. The other "phylogeny" referred to on the website may reflect - I believe -
the point of view of Ringe rather than Warnow, who is a computer specialist.

Adjustments were made (it is not clear when) to the "IE" data which was ONLY
then processed.  There are all kinds of questions that can be raised about
the validity of these "directionality" corrections as well as to the basic
data itself.  I have now read the papers myself and asked a number of experts
to look them and I will be getting to those problems in a future post.

Suffice it to say, for the purposes of Mr. Freisen's points, that the tree
actually does no more than reflect Prof. Ringe's (et al.) positions on what
is appropriate data and what is not - and it does not reflect simple raw data
in any sense.  The processing does nothing more than provide an arrangement
of those judgments (about 350 pieces of "data" derived from various sources)
that is consistent in a certain graphical way.  Matters of pre-attested dates
are still entirely dependent NOT on the algorithm, but ON THE DATA which is
determined BEFORE any processing occured.  Most of this data - including the
"directionality" adjustments are NOT available to us, so we simply do not
know if they are speculative, valid or otherwise.

So with regard to Mr. Friesen's statement that he questions <<the assumption
that the Anatolian languages necessarily split off earlier than the others>>
I sincerely doubt whether the UPenn trees or our discussions of them give
here will enlighten him much.  If anything related to the trees might, it
would be Prof Ringe's explanation of why Mr. Friesen's position would or
would not be the case - since this explanation would be what is reflected a
priori in the trees.

Regards,
Steve Long



More information about the Indo-european mailing list