universality of grammatical functions

[Christopher D. Manning]

Dear all,

I accept the arguments of Joan and the Mohanan's that the absence of
evidence for grammatical relations in some languages does not
necessarily mean that one shouldn't use a framework with GRs, since they
are well motivated elsewhere and we want a universalist theory of grammar.

Nevertheless, I think this misses the essence of the claim about some
languages lacking GRs (or in particular a SUBJ/pivot) that you find in
the tradition of Dixon, Foley & van Valin, etc. (and which hence tends
to influence Australianist thinking).  The essence of the claim is that
some languages lack GRs (or fail to gramaticize a pivot), and
_because_of_that_ these languages are typologically very different.
They lack valence changing operations like passive and applicative, do
not possess controlled clauses, etc.  Phenomena in general appear more
"semantic" because of the lack of a grammaticized pivot.  To the extent
that this is true (and it is a position that I have some sympathy with),
then it isn't sufficient to just dismiss the absence of evidence for
GRs.  We would be missing the explanation for why these languages are
typologically distinctive and just saying that the language happens to
lack a passive, an applicative, controlled complements, etc.  The
question, then, it seems to me, is whether there really is a
well-defined typological cluster of languages where these properties can
be well-explained by the notion of them failing to project a pivot, or
whether the attempt to create such a cluster attributable to a single
parameter would tend to fall apart on closer observation (as "the
non-configurationality parameter" has).

Chris Manning