According to an article in Nature, the “English language [has been] traced to Turkish farmers.” Stop presses, &c. Professor Russell Gray, evolutionary biologist in the psychology department, and a PhD student Quentin Atkinson at Auckland University say that Proto-Indo-European came out of Anatolia almost 10,000 years ago. Gray also admits that his findings may be controversial.
Gray and Atkinson had analysed thousands of words from 87 languages (past and present) to find out when the various branches of the Indo-European family tree started diverging.
“We looked at words from different languages that were clearly related and grouped them in sets.”
Gray said a simple example was that five was cinq in French and cinque in Italian.
“We built matrices of all our information, gleaned from the Internet and every obscure etymological dictionary we could find.”
I wonder if the data they entered into their matrices was of the order of cinq and cinque or s̃ɛk and tʃɪnkʷɛ? I have often wondered whether Gaulish would end up being more closely ranked with Latin than Irish based on these algorithms? [via Desbladet]
[Addendum 12/02/03: Languagehat has added to the bloggage on this, as well as Phluzein, whose entry predates Des’ and mine.]
[Addendum 12/12/03: Bill Poser has a great critique of the Gray and Atkinson article over on Language Log.]
Posted by jim at December 1, 2003 07:25 AMWhile this thing is clearly a total train-wreck, it is not clear which would be preferred. Reputable scholars start reconstruction from the earliest known sources, and orthography is likely to be a fertile source of phonologically older material.
Anyway, Bobby Bryant on sci.lang reports:
"""
It happens that their primary source of data is "Dyen et al.", a data file
that's listed as being on line at
, with the three
dead languages added from unspecified sources.
"""
Des, I'm not saying we should not attempt these kinds of computer-generated models of language change, just—as you say in your entry—that it'd be nice to involve some historical linguists into the mix. The genetic model these sorts of things seem to be based on deal mainly with DNA from the same period, the present, but the linguistic data comes from the present back to 5K years before present.
Posted by: jim on December 1, 2003 09:28 AMOh, absolutely. Do it, but do it _right_.
Posted by: des on December 2, 2003 01:59 AM