Newsgroups: sci.lang
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!europa.eng.gtefsd.com!howland.reston.ans.net!gatech!rutgers!argos.montclair.edu!hubey
From: hubey@pegasus.montclair.edu (H. M. Hubey)
Subject: Re: More Proto-World
Message-ID: <hubey.782937996@pegasus.montclair.edu>
Sender: root@argos.montclair.edu (Operator)
Organization: SCInet @ Montclair State
References: <37pqr1$ffn@tardis.trl.OZ.AU> <37sb2u$jgj@tardis.trl.OZ.AU> <38cds2INNbgr@SUNED.ZOO.CS.YALE.EDU> <hubey.782882530@pegasus.montclair.edu> <Cy4EL5.2uI@inter.NL.net>
Date: Sun, 23 Oct 1994 18:46:36 GMT
Lines: 107

mcv@inter.NL.net (Miguel Carrasquer) writes:

>In article <hubey.782882530@pegasus.montclair.edu>,
>H. M. Hubey <hubey@pegasus.montclair.edu> wrote:
>>
>>All joking aside, do you believe that humankind came from a single
>>source, say about 100,000-200,000 which is more or less the
>>standard belief these days?
>>

>Are you referring to the transition from Homo Erectus to Homo
>Sapiens (which I find dated at 400,000-300,000), or to the transition 
>from "archaic" Homo Sapiens (e.g. "Neanderthal man") to Homo Sapiens 
>var. Sapiens (which my sources put at roughly 40,000 BP)?
>The second date clearly looks more promising for Proto-World
>(at least for our chances of recovering some of it)... but it is
>by no means necessary that the spread of genes be equivalent to
>the spread of language.  It's a possibility to be investigated.

Well, that was the essential question.  When is it that this thing called 
language got started. At what point could it have been called language.
I've seen dates as much as 2 million years ago (Deacon?) as well as
dates of about 100,000 years. Isn't this the beginning point of
whether one believes that one can find traces of this protoworld
language. If it was about 2,000,000 years ago then we might not
have any chance of even finding statistical traces of any correlation.

If it was only 100,000 years ago it's not impossible to find some traces. 
I know that much fun can be made of efforts to put some agreable
sense into the methodology of pronouncing cognatehood.  What other
bases can there be except a statistical one. Or do people prefer to
only allow intution to decide what's truth? Is statistics only good
enough for the other sciences but not linguistics?

As for the specifics:
It doesn't surprise me that the word "flea" shows up all the time
or that it was chosen. Most animals including primates are infested
with fleas and picking fleas (i.e. grooming) is practice among
primates. Similar arguments can be made for picking body parts etc
for looking for cognates.

Is there a more commonsensical or a more scientific approach?

			
>>Aside from the above belief (or nonbelief) is there any difference "in
>>kind" of what the proto-world, Nostratic, etc schools do that
>>is not done by the orthodox/standard schools? Or is the
>>difference only one of degree?
>>If there is no difference "in kind" but only a "difference of
>>degree" are there any statistical measures of how likely such
>>apparent cognates are either real cognates or chance occurrences?

>I think it's best to leave statistics out of this, unless you can
>come up with a measure for the _quality_ of a reconstruction.
>If you feed all the possible, attested sound changes into a
>computer program, it will probably come up with "Guy's Law" or
>worse.  You can prove that any sound can turn into any other sound,
>which is not very helpful...  The same applies to semantic shifts.


A measure of the quality of construction itself is a statistical
test. The statistical test itself has to be tested on what knowledge
we are already sure of. This is how knowledge gets extended.

There's nothing wrong with "Guy's Law" or "Zipf's Law" being
subjected to statistical/probabilistic tests.  There are constraints
on the problem. There are a limited number of sounds, and the probability
of some sounds (i.e words) constructed from this limited choice 
matching some other words (of another language) purely due to
chance can also be calculated. And even the probability that this
number is correct (i.e. confidence levels) can be calculated with
the same technique. At least people would have some standard tests 
and numbers to argue about instead of gut feelings.



>work on Nostratic and Proto-World is comaparable to Schleicher's, or
>to that of other philologists who at the same time were still trying


The problem is that all of these are recursive/iterative. BEfore you 
look for cognates you already have to be convinced that the two
are related but this also means that you should have already been
convinced of the relationship because of the "cognates". 

On the other hand, once the pronouncement that two languages are not 
related is made, then any resemblance becomes immediately due to chance
so that no one bothers/attemtps to make a collection of possible
cognates.

So then, if some do make a comprehensive effort to collect possible 
cognates, then the effort is ridiculed. Why not at least have a 
comprehensive list, try some statistical tests and then see how things
fare compared to what is already "known" or "strongly believed" to
be true?

I don't see any difference between the efforts of the Nostratic,
proto-globe etc and the others. They seem to employ similar tools
and get similar answers. But there's no number or test to fight
over.



--
						-- Mark---
....we must realize that the infinite in the sense of an infinite totality, 
where we still find it used in deductive methods, is an illusion. Hilbert,1925
