semantics

[Miriam Butt]

Hi, 

thanks to Alex Alsina for his further insightful contributions to the
discussion. I readily concede that the proposals for objects in LFG
that I'm trying to formulate seem incompatible with the previous
detailed studies on Bantu object asymmetries. The challenge lies in
trying to bring together two divergent subject matters under a single
point of view. The burden of proof there indeed lies with me. However,
since LFG at the moment has no good way to account for some of the
aspectually conditioned syntactic reflexes of objects, my proposal
cannot be dismissed quite out of hand without the existence of a solid
alternative. In fact, I think that my proposals will not be as
"disastrous" as Alex fears, but it will take some careful examination
and explanation that is probably best left for working out in a paper
and then presented in a summary in this forum.  

So, I shall leave the topic of objects and object thetas aside for the
moment.  In the meantime, however, I'd like to say that the input from
this group has been very helpful in helping me think about work that
is in progress.  

As far as Chris Pinon's further comments go -- I feel answering them
would take us too far afield from LFG. The issues being raised are
interesting, but require detailed semantic argumentation (in
particular, his fight is with Gillian Ramchand, and not me). If this
group is interested, I can respond publicly, but my feeling is that
that particular strand of discussion is becoming too detailed for a
public forum.

What I would like to take up is the thread on the syntax and semantics
interdependence.  Alex differentiates between two views on the
interaction between syntax and semantics:

> the "Reduced Autonomy of Semantics" or RAS, in GB, and the "Full
> Autonomy of Semantics" of FAS, typical of LFG.

While I think these are indeed two very interesting possible
hypotheses to consider, I just wanted to throw some doubt on the
rather clean grouping that is assumed:  FAS=typical of LFG,
RAS=typical of GB.  With regard to the first assumption -- it turns
out that in the bulk of computational literature within LFG (and that
makes up quite a good chunk of the overall literature on LFG), the FAS
might be there in theory, but is definitely not reflected in practice.
The s(emantic)-structure is very clearly a derivative of the syntactic
(f-structure).  S-structure is projected "off of" f-structure.  It
must also make reference to c-structure for scope relations that are
usually lost at f-structure when everything gets stuffed into an
adjunct set.  In both cases, however, the s-structure is very clearly
dependent on the syntactic representations.  

Now, RAS=typical of GB. As far as I understand the way things work, LF
must indeed be considered a SYNTACTIC level of representation, not a
semantic one, as was suggested. The idea is that the LF provides a
basic skeleton for the semantics to relate to, but cannot itself be
considered the level at which semantic interpretation is done. This is
done independently, and for the purposes of "pure" semantic research,
one could dispense with syntactic representations altogether. However,
the challenge, and the interest in LF (when it is not being misused by
syntacticians as the garbage projection where you can fix everything
that you haven't found a reasonable movement for beforehand) is
precisely how a syntactic representation can be related to a semantic
representation. This point is made very nicely (but not exclusively)
in the work by van Geenhoven I mentioned previously. She takes the
semantics as a starting point, and then tries to connect up the
semantics she has determined for a given sentence to the syntax of the
sentence. It is precisely in doing this that she discovers that the
head-movement account cannot be upheld, since it in no case could be
reasonably related to the independently motivated semantic
representation. 

That is, both GB and LFG ultimately believe in an independent level of
semantic representation (I am taking the point of view here of GB
researchers who are indeed working on the syntax/semantics interface,
not of those who use LF as a garbage level, along with morphology, PF,
etc.), but both face the necessity of ultimately connecting up the two
levels of representation.  In GB, LF plays this role.  In LFG, the
projection architecture is currently doing the work, and the
interesting questions arise when one looks at how the levels are
related via this projection architecture.  

In particular, the challenge lies in figuring out exactly which pieces
of information are represented at which level, and how these levels
interact. For example, it has often been assumed that
predicate-argument structure and semantic structure are the same
thing, which leads to problems when one then tries to work out the
exact correspondence between f-structure and s-structure in cases like
that of complex predicates (see Kaplan and Wedekind 1993, for
example).

Alex's point that the derivatory nature of LF might condition you to
look for syntactic reflexes where there are none in order to be able
to come up with the right semantics is well taken. However, being
primarily a syntactician myself, I'd like to reiterate that the reason
I got interested in the interaction between object realization and
aspect at all is precisely because there appear to be syntactic
reflexes of the distinction (that's one of de Hoop's points). So,
there is something in need of explanation, and the issue at hand is
precisely how issues at the syntax-semantic interface can be accounted
for within the architecture of our theories.