Indo-Hittite Hypothesis

[Richard M. Alderson III]

On Fri, 25 Feb 2000, Steve Long (X99Lynx at aol.com) wrote:

> Actually, I've tried to figure out how the UPenn tree could possibly
> 'confirm' the Indo-Hittite hypothesis and I think that the term may have been
> misapplied in the papers that are available on this subject.

> You may recall that the top of the UPenn tree was diagrammed on this list as:

>>       PIE
>>      /   \
>>     /   Anatolian

> This is not the I-H hypothesis, which would yield something like this:

>>       PIH
>>      /   \
>>  PIE   P-Anatolian

> The premise being that PIE and proto-Hittite/Anatolian are sister languages
> with a common parent.

Let's stop here for a moment, and revisit what the UPenn tree purports to do,
vis-a-vis the standard models presented in linguistics texts.

The standard model has for years been a 10-way branching from a central node
labeled PIE, sometimes with minor branches provided for languages like Phrygian
and Thracian.  This model is always accompanied by text to the effect that we
know that there must have been binary or trinary branches at some point, but
that we don't really know where they go.  (NB:  That's 10 branches including
Anatolian, and accepting Balto-Slavic.  If we remove Anatolian--the "Indo-
Hittite hypothesis"--there are 9.)

The UPenn work is supposed to provide exactly the information we have lacked
till now, that is, where to put in branches.  Having generated their branching
structure, they then picked a root position such that Anatolian branches away
from the entire rest of the family, all of which can be viewed as going back to
a single proto-language.

This is *precisely* the position taken by Sturtevant and his followers:  That
we can reconstruct a "narrow PIE" (to use one correspondent's characterization)
that is opposed to Anatolian under a higher branching structure.  Under that
reading of the UPenn tree, ignoring the actual label placed on the root of the
tree, the Indo-Hittite hypothesis is *indeed* "confirmed"--or at least made
plausible if one accepts their placement of the root.

> My understanding is now that the difference between these approaches is not
> trivial.  The reconstruction of the hypothesized PIH gives substantially more
> weight to the Anatolian languages than does a reconstruction of PIE that
> makes Hittite et al a mere branch of Indo-European.

Actually, in its original form, the IH hypothesis *trivializes* the data of the
Anatolian languages with regard to "Indo-European proper" (the more common term
in IH writings):  We know everything we need to know to reconstruct the proto-
language from which all 9 branches derive, and need not look at those pesky
Anatolian languages at all!  Oh, sure, they're interesting, with those funny
obstruents that match up with Saussure's _coe'fficients sonantiques_, but they
have nothing to say to us about the wonderful PIE we've worked on for so long.

> And although the I-H hypothesis has been associated with, e.g., the new
> version of the IE obstruent system offered by Hopper, Gamkrelidze and Ivanov,
> I'm told that the actual scope of its possible ramifications for PIE
> reconstruction has not yet been explored.

I've just laid out its ramifications:  It *has* *NONE*!  It *doesn't* *MATTER*!

But it has nothing at all to do with, at least, Hopper's proposal regarding the
PIE stop system.  When Paul Hopper made his proposal about 30 years ago, two
things were true of Indo-European studies:  Almost no one accepted the Indo-
Hittite hypothesis, and a large number of IEists rejected the laryngeals at the
stage of IE immediately prior to breakup.  Hopper's proposal had to do rather
with the question of the phonological naturalness of the reconstructed stop
system, and nothing to do with "the Indo-Hittite hypothesis" or "the laryngeal
theory".

> I do not believe - again, from the papers we have - that the algorithm used
> on IE at UPenn ever produced an 'unrooted tree'.  Contrary to what has been
> said on this list in the past, the external adjustments appeared to have been
> made directly to the algorithm from the outset.  What we see in the papers is
> a model of a 'unrooted tree', but I could not find one that represents the IE
> languages.

Since the papers available on their website state that it was an unrooted tree
to which they added rooting information, you appear to be accusing the group of
deliberate falsehoods.  I'd be very careful of that, were I you.

								Rich Alderson