Re: [tied] About methodology...

----- Original Message -----

From: Glen Gordon

To: cybalist@egroups.com

Sent: Wednesday, August 30, 2000 12:30 AM

Subject: Re: [tied] About methodology...

Glen: True, it's fuzzy - so what? The point is to acquire greater and greater accuracy. We all know that the tree model isn't a _perfect_ way of looking at languages and language interaction but it gives us an immediate general picture of relationship...

Si parva licet componere magnis, you are arguing not unlike Einstein, who never came to terms with the uncertainty principle. Remember? "God doesn't play dice". "Greater and greater accuracy" is possible up to a certain limit.


Glen: ... However you seem to equate "worse" with "hopeless" but the two just don't go hand in hand. The difference between Nostratic and IE is only a matter of the level of accuracy but as accuracy increases within IE studies, so to will the accuracy increase with Nostratic studies and ultimately other studies like DeneCaucasian or Asiatic. On the one hand, we can't expect Nostratic to ever be as accurate as IE is at any given time, but on the other hand, it doesn't mean that Nostratic will never be just as accurate as IE now is currently. Get it?

By stressing the tree/wave duality and its consequences I don't want to argue that "deep" research is hopeless. I only say that the comparative method has its limits and once it breaks down you can't go on reconstructing protolanguages because there in no longer any detectable "genetic signal". But areal networks produce regional effects which sometimes remain stable even over many millennia (as in Australia). Ancient Sprachbunds may be as fascinating as protolanguages. As you say (and I agree), there is far too little interest in them, also among IEists. Dialectologists don't use trees at all and still have a lot to say about the history and evolution of languages. Historical linguistics is not only about splits and unitary protolanguages. The Inuit example in your posting is instructive. A dialect network model works excellently here, while a tree model is completely inadequate.

Piotr