question on Preferred Arg. Str.

Martin Haspelmath haspelmath at
Tue Oct 12 14:43:33 UTC 2004

I have a question about Du Bois's Preferred Argument Structure. I have
read some of the relevant literature, and I wonder why he states this as
FOUR different "constraints", i.e. four different observations about
statistical tendencies in discourse:

(i) Avoid more than two lexical core arguments ("One Lexical Argument
(ii) Avoid lexical A-argument ("Non-lexical A constraint")
(iii) Avoid more than two new core arguments ("One New Argument Constraint")
(iv) Avoid new A-argument ("Given A constraint")

It seems to me that the first and third constraints follow
straightforwardly from the second and fourth, respectively: If the
A-argument is not lexical/new, there cannot be two lexical/new core
arguments, because every clause has at most two core arguments by
definition. Why do we need (i) and (iii) as independent constraints?

Also, it appears that (ii) follows from (iv) and the well-known strong
correlation between given status and non-lexical (pronominal, zero)
expression. Finally, (iv) would seem to follow from the strong tendency
for A-arguments to be human, and the strong tendency for human arguments
to be topical.

So the Preferred Argument Structure constraints seem to be completely
predictable consequences of well-known statistical tendencies in
discourse. Or is there an error in my reasoning? Is there perhaps any
independent evidence that the constraints are (partially) independent of
each other?

I don't mean to say that the constraints are uninteresting because they
are predictable, I just wonder why the literature doesn't seem to
mention this predictability.

Thanks a lot,

More information about the Funknet mailing list