Stephen Oppenheimer wrote:
> The recently published estimates of time-depth in I-E come not from
> linguists, but primarily from mathematicians. The raw data were
> respectively obtained with permission from Don Ringe's and Isidore
> Dyen's I-E cognate sets, but neither of these linguists had their names
> as co-authors on the papers (Gray, Russell and Atkinson, Quentin (2003),
> ‘Language-tree divergence times support the Anatolian theory of
> Indo-European origin’, Nature, 426: 435–9. and Atkinson, Q., Nicholls,
> G.,Welch, D. and Gray, R. (2005), ‘From words to dates:Water into wine,
> mathemagic or phylogenetic inference?’, Transactions of the Philological
> Society, 103(2): 193–219.).
> I happen to find their methods and results rather convincing;
As a mathematician with experience in the area of phylogenetic tree reconstruction, I happen to find their methods and results completely unconvincing. The tree reconstruction methods even in the more quantifiable area of DNA analysis are full of uncertainties as to what the correct model should be, and all practical methods are heuristic.
Having said that, I very much support research in this area and would like to see better methods developed. But we have a long way to go yet.
Here are some specific objections to P Forster & A Thot, Toward a phylogenetic chronology of ancient Gaulish, Celtic, and Indo-European. Proc Natl Acad Sci USA 100:9079-9084 (2003), available at http://www.pubmedcentral.gov/articlerender.fcgi?artid=166441&rendertype=abstract.
0. What is the definition of "cognate" being used? It seems to be done by visual similarity of the written forms of words - obviously a bad procedure.
1. The Latin and Greek "god" words should not be equated
2. Most of the Gaulish forms occur once (or very few times) only, and mostly on pottery graffiti, where the interpretation is unclear. The method gives too much weight to this sparse and noisy data.
3. There is an arbitrariness in the translations. For example, under ALLOS one could have put Latin alias, and then we would suddenly have a match. The concepts "second"="following" and "other" overlap.
4. Why should primus and protos be counted as a match, but not cyntaf? Although the latter Welsh word has a different root, it shares a common superlative ending (the clue being the -m- in Latin and the -f in Welsh). The method hides the really interesting information, just because it is not obvious at first. Similarly filia:hija is a hidden relationship, not a mismatch.
5. Welsh "mae e wedi gwneud" is a modern colloquial periphrasis for "he has done", of not much interest for understanding historical relationships (it means literally something like "there is he after doing"). Welsh also has simple past forms (aorist and imperfect) which it would have been better to use here.
6. In the case of "has made", the translations don't match exactly: fecit can mean he/she/it made, but the Welsh form only "he made".
7. There are many case of borrowing, such as Welsh ffwrn from Latin. These will obscure the true relationship.
Keith
|