Daniel Everett dan.everett at MAN.AC.UK
Tue Sep 30 19:39:55 UTC 2003

With regard to the discussion between Tom Payne and Matthew Dryer, the
distinction between a 'statistical cross-linguistic generalization' and
its 'explanation' could be less clear-cut than Tom and Matthew suppose.

I agree with Matthew that proving that we have an explanation is hard.
In fact, I think it is impossible to do so. To 'prove an explanation'
would be to find Truth. But if what you have found is Truth, then
(Rorty makes these points with incomparably more originality and
eloquence than I) it would be in no need of revision. But this is not
what we find. All proposed explanations are revised. But if a statement
is revised, what part of it was ever true? How could we know this in
principle? We cannot. Hence I agree with Rorty that 'Truth is just a
compliment we pay to ourselves' when our accounts are useful to our
goals. Any explanation will be useful or not useful according to our
present goals and it is that usefulness, not an inherent property of
'Truthfulness' that most of us seem to be after in practice. But I
cannot prove that, nor do I say it is Truth.

With regard to the generalizations, these also have no guarantee of
long-term usefulness. If I propose generalizations, say, regarding SOV
vs. VSO languages, these are based on the idea that languages have
Subjects and Objects. But if I do accept this idea (for example, if my
last name is Van Valin), then there may not be a generalization. The
same could be said for any generalization we make. They are based on
assumptions, ideas, theories, etc. that we find useful at present. But
they could all go the way of phlogiston or the ether.

Once again, we do what we do based on what interests us. We call this
or that an 'explanation' or a 'generalization' based on who we are,
what we read, what we believe, when we live, where we live, etc. This
is all natural and fine. Much of this seems to be useful to our species
(the subset of the species called 'linguists') across temporal and
spatial boundaries. But that hardly entitles it to the claim that it is
Truth - i.e. always to be accepted without revision by all future

Matthew is right that the generalizations are easier to get. That is
likely because they are based on socially shared assumptions and
require less originality to make. The explanations also involve
cultural knowledge and particular members of specific societies, but
also, perhaps, greater reflection and originality and thus are harder
because they receive less (transparent) help from social knowledge. But
no more true. And no more valuable either. Thinking that explanations
are somehow on a higher plane than discoveries seems to be  to result
from being enamoured with the idea of Truth - that it can be found or
approached asymptotically. I reject this. But, on the other hand, if
one says one likes explanations because they involve more originality
and thus are more like art, then I can agree.

More and more, I see languages themselves as works of art, the results
of diachronic accident, cultural values, general cognitive constraints,
arbitrary structures, and structures and structural principles that
result from other sorts of functional considerations. But irreplaceable
like the finest works of art and no more reduceable to science that any
other work of art.

-- Dan


Daniel L. Everett
Professor of Phonology
Postgraduate Programme Director
Postgraduate Admissions Officer
Department of Linguistics
The University of Manchester
Oxford Road
Manchester, UK M13 9PL
Fax: 44-161-275-3187
Office: 44-161-275-3158

More information about the Lingtyp mailing list