> > As far as I am concerned
> > (and I don't knoe if Dick agrees with me on this), the only really
stable
> > distinction between (named) valents and other dependencies (ie adjuncts)
> > is in their semantic behaviour
>
> What does 'stable' mean here?
>
Good question. I've had a think about it, and I find it means 'accepted by
me as the definition'.
A long time ago I tried collecting together all the properties taken as
distinguishing valents and adjuncts. I found that very few of them ever
coincided exactly. I will have been trying to give a prototypy sort of
account, and even that failed. I can only conclude that the properties don't
define a gategory of grammatical relation.
What I say now is that some dependents (to some extent) refer to (semantic)
arguments of the parent, and these are pretty much the same ones we usually
call complements or valents. Since this class is defined by only one
property it means that you don't need to represent it in the grammar.
Jasp
> --And.
>
|