Re: Glottochronology.

From: John Croft
Message: 2220
Date: 2000-04-27

Andrew wrote

> The fact that something involves many bodies does not mean that
mathematics
> cannot treat it, at least in the large. For example, the movement
of
air
> involves very many millions of molecules. In some cases (eg design
of
> aeroplanes) we have good mathematical models, in others (eg weather
> forecasting) our models are less successful.
>
> Dennis also commented:
>
> >Chaos Theory claims to describe systems with integral feedback
(the
output
> >of one state is the input of the next), where stability is
dependent on
> >highly complex and sensitive factors, a slight change in any one
of
which
> >may (or may not) lead to states of wild unpredictability until a
new
> >stability is arrived at. Doesn't this perfectly describe language
change?
>
> There is a risk that chaos theory is seen an a panacea. The chaos
label is
> an easy dustbin class into which we can throw any unsolved puzzle.
When I
> have failed to solve a problem, calling it "chaotic" implies that
nobody
> else could solve it either. I can then feel better inside because
my
failure
> to solve a problem becomes the problem's fault and not my own.
>
> But technical work on chaos is much more precise than this vague
feeling
> that a problem is intractable. The nub of chaotic systems is extreme
> sensitivity to starting conditions - with particles which start
very
close
> ending up far away, and vice versa. If language evolution was like
this,
> then we should not be able to disentangle (for example) Uralic
languages
> from Indo-European languages. In fact, we can and do separate
languages in
> this way. This is prima facie evidence that we are not dealing with
chaotic
> systems here.

We need to be precise about our use of Chaos Theory.
Chaos/Complexity
Theory systems give the illusion of being chaotic, but they are
underlain by certain regularities, based upon "complex atractors".
Thus what appears to be chaotic is in fact quite ordered. Language
systems are good examples of "atractors", and so they could possibly
allow a Chaos/Complexity treatment.

> Current glottochronology models may seem unreliable, but they are
also very
> simple. There is a trade-off. Given painstaking empirical studies,
we may
> find more sophisticated metrics of language similarity, which
better
allow
> us to corroborate archaeological data on language divergence. We
can
guess
> what these measures might involve - for example, as well as common
> vocabulary, we might give some weight to common grammatical or
phonological
> features. We might also expect to trace language divergence more
accurately
> if we take account of several languages at once, rather than just a
series
> of comparisons of language pairs.

Glottochronological studies have been used to give dates for the
split
of PIE (4,500-3,500 BCE) for example. Given this example it would
seem to be based on some evidence. Others have used it with
profitability on Austronesian languages too. I know Guthrie and
others
have used Glottochronology in their comparative studies of Bantu
Languages.

Glottochronology was used by Karl Frankin quite profitably for the
splitting of the Central Highlands Family of the Trans Papuan
Superphylum in Papua New Guinea. I think on specific cases, when
supplimented with other evidence it can be a useful technique. When
used alone (like any of the methods we use) it needs to be taken with
a grain of salt.

Hope this helps

Regards
John