PIE vs. Proto-World (Proto-Language)

[Larry Trask]

On Thu, 29 Jul 1999, Rick Mc Callister wrote:

> 	Unless you're saying that language arose in the last 100,000
> years [i.e. before humans left Africa], then I don't think you can
> make a serious claim that there are any true isolates.

Er -- what is a "true isolate"?  A language which arose independently of
all other languages?  If so, I'm afraid I can't see that as a useful
construct.  Since we have no way of investigatng the ultimate origins of
languages, we have no way of determining whether any language has such
an independent origin.  I prefer to use `isolate' in its ordinary sense
of a language which cannot be shown to be related to any other known
language.

> Basque and Burushaski are surely related to other languages but the
> lumpers are going to have to work a lot harder to prove it.

Quite possibly Basque and Burushaski really are very remotely related to
other known languages, but no one has ever succeeded in finding any
persuasive evidence that either is related to anything else at all.
At least for Basque, which I know better, I believe that the chances are
now vanishingly small that any relatives will *ever* be discovered.
There's almost nothing left to look at.

> 	My understanding is that language arose before humans left Africa,
> so any claims of polygenesis would have to be examined among African
> languages. Given that the only existing language families in Africa are
> Niger-Kordofanian, Nilo-Saharan, Afro-Asiatic and the Khoisan languages
> [which may be between 1and 5 families],

Well.  Niger-Kordofanian and Afro-Asiatic are widely accepted as valid,
even though the published evidence in support of each is sparse.
However, not only Khoisan but also Nilo-Saharan are at present little
more than hopeful areal groupings.  Neither is supported by any
significant body of evidence, and both are doubted by some specialists.

> it seems that the onus of proof is on the polygenesists.

Why?

In comparative linguistics, the null hypothesis is always this:

	No languages are related.

This is the hypothesis we seek to falsify when we try to show that some
languages really are related -- as we have managed to do in many cases,
of course.

But putting the burden of proof onto the polygeneticists strikes me as a
serious methodological error: it is equivalent to changing the null
hypothesis to this:

	All languages are related.

And I cannot see that this is a wise move, or even a possible move.
Doing so would render demonstrations of relatedness otiose, and require
instead demonstrations of *unrelatedness*.  And it is a logical
impossibility to demonstrate that two languages are unrelated.  The best
we can ever hope for in this direction is to show that there exists no
evidence to relate two languages -- but such an outcome clearly does not
prove that those languages are unrelated.

If we had no IE languages but Welsh and Albanian, I very much doubt that
we could make a persuasive case that they were related, and we would
have to conclude that there was no evidence to relate them, yet this
would clearly not prove unrelatedness.

I'm afraid that the burden of proof is *always* on the shoulders of
those who wish to argue for relatedness, not of those who want to argue
for unrelatedness.

Larry Trask
COGS
University of Sussex
Brighton BN1 9QH
UK

larryt at cogs.susx.ac.uk