Re: ASJP - Automated Glottochronology: Dating the World's Language F

--- In cybalist@yahoogroups.com, "Francesco Brighenti" <frabrig@...> wrote:

>
>
>
>
> Dear List,
>
> A file I have just uploaded in the Files section (which can be also accessed through the link being provided in the announcement message at
>
> http://tech.groups.yahoo.com/group/cybalist/message/65117 ),
>
> contains a paper entitled "Automated Glottochronology: Dating the World's Language Families." Most of the dates they arrive at are about what one would expect (e.g., Indo-European, Indo-Aryan, Indo-Iranian, Iranian, Balto-Slavic), though many (e.g., Hungarian, Germanic) are lower than is generally thought. Note the low dates for languages that are often given very early dates (e.g., Turkic, Mogolic, Sinitic). In any event, a lot of people contributed to this research, and it should generate considerable discussion.

The dates on Chinese, Turkic (and Common Turkic) and Mongolic are input data, and Table 2 shows that the dates and similarities for Turkic and Mongolic are both discordant. I'm disappointed they haven't done much better than the Levenshtein metric (even adjusted for word length). It can be misled by the choice of citation form, e.g. infinitive in l- for Maltese and Hebrew v. perfective for Arabic and position shifts (e.g. Grimm's Law).

They've also turned up some contradictory dates -

1. South-Central Dravidian split before Dravidian.
2. Eastern Malayo-Polynesian split before Malayo-Polynesian (possibly an error in writing it up - the date is the same as that of Austronesian, but the similarities are different.)
3. Northern Hokan split before Hokan.
4. Oregon Penutian split before Penutian.
5. Northern and Southern Carib both split a tad before Carib.
6. Rama split before Chibchan.
7. North-Central Panoan split before Panoan.

I'm surprised straight lines fit the scattergram of Figure 2 as well as they do. The effect of decreasing percentage loss with increasing age as commonalities are reduced to conservative items barely shows up (and I'm sure what there is is not statisitcally significant). Perhaps there greated tendency to phonetic change has had a counter effect. The cognate Sanskrit, English, French and Russian 1st person singular nominative have zero similarity.

Richard.

Richard.