The Indo-European Hypothesis [was Re: The Neolithic Hypothesis]

[Peter &/or Graham]

Ray talked of models for predicting language development.

I think Ray has missed the point.   Statistical models remain only that -
statistical.   They might be useful for predicting or explaining language
behaviour, or human behaviour, in general terms over large "population
samples", but they are of no value whatever for the other major purpose of
historical linguistics, which is explaining the actual state of actual
languages.   Statistics might tell me that 50 % of people who smoke die of a
smoking related cause, but it does not tell me why the particular smoker
Fred died from that cause.   Likewise, statistical models of the kind Ray
suggests have their function, in helping us understand the processes
involved in language change, but they cannot explain why Sanskrit gacchati
is related to Greek baske: and Latin venit.

To be specific, Ray said:
> others have pointed out the failings of IE to
>explain how other languages developed.

He means how the family tree model is inadequate on its own;  but he forgets
that the combined models we actually use are remarkably successful.
Various correspondents pointed out to him the inadequacy of the simple model
he first posited.   He should not think that poor linguists have his simple
model as their only available tool.

He also said:
> it
>is not a relaible predictor of specific languages, and the family model upon
>which it is based is not a reliable predictor of the divergence/convergence
>dicotomy of all languistic groups.

The family tree / convergence / sprach bund complex of models is not
designed to predict, but to explain actual individual events, and this it
does very well.

He further said:
>It seems to me, that the theorectival framework of an
>epidemicological model of the spread of disease would easily supply this
>need of the linguistic community.

He means the need to predict.   Yes, models that help us understand
processes and make general predictions have their place.   But he should
understand that this is a different purpose from his original question, and
he should not confuse general statistical prediction with the careful and
detailed explanatory work of historical linguistics.

Peter