[b-hebrew] Recreating the Origins of Language

Will Parsons wbparsons at alum.mit.edu
Sun Feb 24 13:57:27 EST 2013

On Sun, 24 Feb 2013 01:42:54 -0800, "Rev. Bryant J. Williams III" <bjwvmw at com-pair.net> wrote:
> Dear List,
> BAR has the following article on Recreating the Origins of
> Lanauges. This article is a small summary of a larger BBC article at
> http://www.bbc.co.uk/news/science-environment-21427896. This BBC
> article is actually a reporting of the larger Proceedings of the
> National Academy of Science
> (http://www.pnas.org/content/early/2013/02/05/1204678110 ). Sorry, I
> am not a subscriber to this journal.
> I find it interesting that the picture accompanying the BAR article
> is of The Tower of Babel. Interesting that Polynesian languages are
> used. What would be nice is to try other language groups. Some think
> Africa, because it is thought that languages and mankind descended
> from the African continent. Genesis 2 indicates that mankind was
> from the area of the Euphrates and Tigris Rivers. Genesis 11
> indicates that languages were a direct result of man's disobedience
> at the Tower of Babel. Furthermore, I would like to see the
> assumptions or presuppositions, methodology, etc. used in the
> study. I remind myself of GIGO. 85% of the language can be
> reduplicated. Not bad. But it is definitely not good. 15% cannot be
> recreated unless there is a lot of GUESSING, albeit educated
> guessing, but still GUESSING going. Oops! That should be
> "speculation" NOT "GUESSING. Scholars don't guess they
> speculate.:))))
> Any comments? Does any one have the full PNAS article?

I'm skeptical.  It seems that the only kind of check on the accuracy
of the results are if the results match what historical linguists
reconstruct, which can be quite speculative.  There's plenty of
uncertainty in the reconstruction of Prot-Indo-European, probably the
most thoroughly studied language family of all.  When one reconstucts
a language supposedly spoken 7000 years ago solely on the basis of
modern spoken languages, how is one to judge whether the reconstuction
is good or wildy off the mark?

It would be interesting to see what the software does with a case that
can be independently verified.  I know of only one good case, and
several not-so-good:

The good test case is that of the Romance languages.  This is only
case that I am aware of where one has a variety of modern languages
and dialects, and the common ancestor is well known with a lot of
written evidence.  How good would the software be in reconstructing
Latin on the basis of modern Romance dialects?

Not-so-good cases would include:

The Germanic languages.  We do not have written evidence for the
ancestor of these, but we do have Gothic, which is a lot closer to the
ancestor language that other Germanic written languages.

Similarly with the Slavic languages.  Old Church Slavonic is not the
common ancestor of modern Slavic languages, but is closer to the
ancestor than others.

Perhaps the modern Indo-Aryan languages is a similar case, with
Sanskrit serving as the not-quite-but-close ancestor.

There may be a handful of others that I can't think of right now.

Will Parsons

More information about the b-hebrew mailing list