--- In
cybalist@yahoogroups.com, "Glen Gordon" <glengordon01@...>
wrote:
> There is too much variety in this rate for such a calculation to be
> meaningful. A word
> could be replaced now, two hundred years from now, or 10,000 years
from now.
> There's no limit to when this might occur.
This is what a model of random changes predicts!
> And let's say we only concentrate on first and second person
singular
> pronouns.
> How does one calculate this rate of change given our limited 6000-
year
> knowledge
> of language change?
Take a large family, look for the differences, and apply the
appropriate statistics. So long as you can identify cognates
reliably, you don't need to know the history of the languages. It
helps to have a few dates though, or otherwise all you get is
relative rates. (You can also mix families, so long as you don't do
direct comparisons of unrelated languages.) Given the involved
statistics, you'd probably settle for extracting a maximum likelihood
estimator. That's what the big Austronesian study did. The bad news
is that the data shows, though this conclusion was not drawn in the
original paper, that the rate of change varied from place to place,
being particularly high near New Guinea. Such a variation is not in
the statistical model used for glottochronology.
Richard.