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

[Ray Hendon]

>Ray talked of models for predicting language development.

>I think Ray has missed the point. Statistical models remain only that -
>statistical.

 The model for predicting language development that I had in mind is a
mathematical model-a theoretical model that needs the help of those in the
linguistic community to specify.  A statistical model is quite different: it
attempts to test a mathematical model once it has been specified.
Statistical processes are great for getting an idea of the accuracy of your
idea and how confident you can be that the relationship between things you
tested are not attributed to chance.  You are entirely correct when you warn
about the inability of statistics alone to prove anything one way or
another.  Therefore, if statistical work is to mean anything to anyone, it
must first be preceeded by thoroughly professional analysis and observation.
Long before any statistical procedures can be applied or even thought about,
the model itself-the cause-effect theory must first be fully explored.
Here, in supplying possibel cause-effect relationships, a linguist is the
only person who could be of help. But in this theoretical realm I see what I
believe to be a propery analogous situation from medical research: their
use of  a mathematical probability modeling technique to predict the spread
of infectious disease among a community.

 The draw-back of the model is its seeming simplicity.  Every scholar in the
field of linguistics, especially historical linguistics, knows how complex
the process of language adoption and spread is.  A two-variable model
because of its simplicity immediately repels them.  This, I believe, is what
the main objection of the professionals to using this potential tool would
be.

 What I failed to emphasize in my earlier suggestion is each variable is
itself a function, dependent upon many things.  But, all the miriad other
things that influence what langage a given population is likely to speak are
accounted for with the "exposure" and "susceptibility" variables.  Just as
there are many reasons why some people are vulnerable to influenza, and
there are many reasons why a person is exposed to the virus, in the end it
doesn't matter.  They have either been exposed to it or not, and if there is
no exposure there will be no influenza. If they are exposed, they are either
susceptible to the infection or they are not susceptible to the infection.
The why's and wherefore's are a different matter.

In adopting a language, a child or adult is either exposed or not exposed to
another language.  And if they are exposed, some are more likely to begin
using it than others.  Some people seek out exposure and actively pursue
learning a language, others attempt to isolate themselves from a specific
language, showing a strong prererence not to learn another one.  But for an
objective observer, looking at a linguistic community, all these variations
in preference, interests and exposure, will average out, and there will be a
community-wide exposure level and an average susceptibility of some
magnitude.  If these magnitudes are known or can be reasonably well
estimated, then predictions can be made with regard to which language will
become dominate.

 Keep in mind also that when we say prediction, the time-direction of the
prediction can be forward or backward.  We can predict what percentage of
the residents of Rome spoke Latin in 250 BCE using the same model as for
predicting what language will be spoken in San Antonia, Texas in the year AD
2300 or London in 1067.  Prediction relates to estimating the unknown, not
only to future events.

A limiting factor in this model, however, is its time/place specificity.  I
live in a linguistic soup of English, South Texas English, Border Spanish,
Spanglish, and all points in between.  I am curious that in the year 2500,
which language will dominate.  What will be the balance between Spanish and
Texas English.  It seems to me, that if I had an exposure/susceptibility
model, I cold make that prediction with some degree of precision.  But the
prediction is limited by the certainty of our knowledge that what we define
as English in 1999 will not be what English will be in 2500, if it exists at
all.  Certainly the effects of time can be accomodated by the
epidemiological model, but there must be other models that account for
changes within the language itself.  These are more metaphysical in nature,
and need separate treatment.  The epidemiological model accounts for
adoption of a language within an area or population-in other words, the
time/space relationship of language adoption.  Why we speak at all and what
accounts for changes within a language, are separate issues.

The strengths of the epidemiological model is in its robustness and
effectiveness.  It is robust enough to handle convergence or disbursement,
invasion, snob-appeal, migration, technological change, trade, education and
religion.  It is effective because it allows you to dicotomize any
population of language-speakers.  Every person is either susceptible or not,
and every person is or is not exposed to the language.  At a practical level
we know that there are many variants as to how susceptible one person is
over another, and the quantity and quality of the exposure will vary greatly
between individuals.  But I am not sure that these fine-tuning features
could not be accounted for once the model was fully developed.

Whether this is of any use to the community of linguists, however, is
something that I do not know.  I am most interested in the responses I have
received so far, and look forward to the discussion that may arise.

Ray Hendon