The UPenn IE Tree (Celtic as PIE)

[X99Lynx at aol.com]

In a message dated 9/9/99 12:05:18 AM, alderson at netcom.com wrote:

<<>No it doesn't.  Not in the scenario I gave you.  Celtic IS PIE!  It
>distinguishes only two.

Your experiment falls apart at this juncture.

The evidence for three different series of dorsals is pan-IE:  The canonical
examples are things like Skt. _kravis._ "raw meat", Latin _cruor_ "gore, blood"
in which the "satem" data lead the linguist to expect a labiovelar in "centum"
dialects, while the "centum" data lead the linguist to expect a palatal or
alveolar fricative in "satem" dialects.>>

What "data" is that?  The data used in the approach that created the Stammbaum?

If you are saying that the whole IE canon stands behind the Stammbaum, well
then yes I guess my experiment falls apart.  But only because it was
addressed to what the Stammbaum was saying.  Not the IE canon.

I want to remind you that one of the statements made about the data behind
the Stammbaum is that it includes no reconstructions.  And the only basis for
chronology was attestation.  Aren't we talking about what that limited data
yielded?  Can its conclusions stand on its own?  Please?

I have too much respect for the study of IE and the people on this list (and
too much respect for my own ignorance) to challenge anything you've said
above.  But I don't think it's even-handed or necessary to place the entire
IE canon between my points and the Stammbaum's approach.

If the approach used by the Stammbaum were testing the PIE 3-obstruent
reconstruction by going to the documented data, and disagreed with it, I'm
sure you would be one of the first to ask what assumptions were behind the
program.

This is really no different.  Chronological conclusions, conclusions about
what constituted "innovations", conclusions about when individual languages
branched off from "the proto-language" are quite clearly represented in the
Stammbaum.  Were those conclusions stricly based on the data, with no
reconstructions and limited chronological assumptions?  Then let the process
support itself.

I don't believe the approach behind the Stammbaum can contradict the Celtic =
PIE assumption.  Not by itself.  You do have to bring in the IE canon to do
that.  But the existence of the three dorsal obstruents series in PIE is a
reconstruction.  So why bring it in?  It's not part of the protocol.  Please
don't give this approach eyes where it's blind.  That will not allow us to
see what it is really capable of doing.

<<There is no way to predict which velars in Celtic will be represented as
such,
and which as palatals, in the "satem" dialects.  Therefore, if you insist on
placing Celtic at the top of the tree, you make incorrect predictions of the
facts in evidence, and you fail.>>

This is a whole different matter.  How else would you test a reconstruction
than by how well it predicted the actual outcome?  In the Stammbaum approach,
that should not be the test.   The process should start with the data and go
where the data sends it.  Please understand this.  If you found certain
evidence that PIE had only two obstruents tomorrow, what would you think?  It
would simply mean that the "incorrect predictions" were irrelevant and the
evidence for three obstruents would have to be explained in some other way.

Would a logic program be able to yield that kind of conclusion from the data
if you told it to assume ahead of time that there were three obstruents in
PIE?  If you wanted an honest analysis, wouldn't you want it to go to the raw
data without that assumption?

I pretty much believed (mainly on authority) that PIE had three obstruents
before I thought up the scenario.  I think I still do.  Whether it did or not
does not matter for this purpose.  It only distracts from seeing what the
process can do for itself.

<<(The fact of Luwian maintenance of the three-way distinction is simply
icing on the cake of the comparative method.)>>

You know of course I  didn't say that.  What I wrote was that Luwian was the
only documented instance of 3 obstruents.  To use anything else would be to
use a reconstruction.  The Stammbaum approach supposedly does not use
reconstructions.  Luwian is all you get as direct (not reconstructed)
evidence of three.

You wrote:
<<The "innovation" is not unshared, but rather occurs across the spectrum.>>

So therefore Celtic, which is part of that spectrum, could at one time have
had 3 obstruents?  That would also take care of the unmerging problem.

<<It is not a matter of assumptions about PIE, but of real data in multiple
languages.  Your insistence otherwise reveals nothing about the Ringe tree.>>

If you take a second look at this, I think you may change your mind.

Regards,
Steve Long