lexico-statistics

[H. M. Hubey]

----------------------------Original message----------------------------
Larry Trask wrote:

> -As for the universal validity of Swadesh's lists, this has been very
> severely questioned on a number of grounds.  People continue to use
> Swadesh's lists, since no other list appears to have a better claim to
> universality, but specialists in particular families or areas have
> sometimes drawn up their own lists.  For example, somebody (I forget
> who) has drawn up a list of words appropriate for working with in
> southeast Asia.

The simplest way to cure this problem, using common sense, is to make the
list longer. That is how the effect of errors is minimized. There are
basically two principles in use:

1. Some words should not be used (i.e. high-tech words)
2. We should use words that likely existed very long ago.

Both of these are the opposite sides of the same coin. If words like snow and
ice are no good for the tropics then these words should not exist. If they
exist then they were borrowed or the people possibly kept memories of the
words alive. If they were all borrowed from related languages then they will
resemble each other, but if the languages are being compared only to each
other then it does not really matter. It would matter if we were comparing
some of these languages to other language families (such as IE) from which
they could have been borrowed.

The basic concepts of statistics are based on this. The greater the sample
the smaller the uncertainty in the result. The smaller the sample, the
greater the uncertainty of the result.


--
M. Hubey
Email:          hubeyh at Montclair.edu    Backup:hubeyh at alpha.montclair.edu
WWW Page:       http://www.csam.montclair.edu/Faculty/Hubey.html