Reply to Andrews: The power of linear logic

[Miriam Butt]

Taking up Ron's very nice discussion of why one might not choose
to encode all of LFG's parallel levels of representation through
the same formalism, I just wanted to add a brief comment with regard
to Mark Johnson's proposal.  

The idea of extending linear logic to representing f-structures leaves
me a little at a loss from a syntactician's point of view. 

Basically Mark's proposal seems to reduce f-structures to little more
than an intermediate state from which to compute semantics, and in
effect, they could just as well disappear. Now, if your ultimate and
only goal in doing LFG is getting a semantics from an input string
(modeled by c-structure), then this might be fine. However, if you are
a syntactician, you kind of feel crowded out of the LFG theory at this
point. At which level of representation can I now talk about control,
binding, or agreement? I think one might arrive at a consensus that
some aspects of phenomena such as control and binding are semantic.
However, some aspects of them are very clearly syntactic in nature and
as LFG has spent a long time arguing against a purely structural point
of view, i.e., has placed quite a bit of syntactic importance on the
level of f-structure, it seems a bit strange to me to try to dispense
with it almost entirely.

Miriam