semantics

[Joan Bresnan]

Why is the latest discussion on this list about semantics making me
cringe?  I have the feeling that people are shooting from the hip
without having had a lot of target practice.  Why, for example, do we
get generalities about "the LFG semanticists" but no reference to
specific works by actual researchers of semantics in lfg
(e.g. Fenstad, Halvorsen, Langholm, and van Benthem 1987 or Dalrymple,
Lamping, Pereira, and Saraswat 1996)?  In these works, the issues of
compositionality and the relation between syntax and semantics 
are explicitly addressed within a carefully defined context.   
This work belies some of Miriam's pronouncements about
constituent structure determining semantic projections, which Alex has so
quickly accepted.  

	Concerning Hudson's points:

>dh: Surely we'd all agree that semantics is compositional in the sense that
>the total meaning of the whole sentence is influenced by the sentence's
>syntactic structure? If semantics was `independent' of syntax that wouldn't
>be true, so no-one really believes in independence in that sense. Presumably
>the debate is about two other things:

I'd certainly agree with this.

> Q1. How simple is the mapping relation? (GB: simple, LFG: complex?)
> 

No, I think the mapping between syntax and semantics in lfg does *not*
require greater complexity just on account of the abandonment of
compositionality in the phrase structure.  For example, in the
linear logic framework of semantics being developed by Dalrymple et
al., scope ambiguities just fall out of the means of logically inferring
the possible meanings of a sentence from its lexical items and their
simple argument structure relations.  No special representations equivalent
to QR (or Cooper storage) are needed in the mapping.

> Q2. Does the semantic structure include lexical semantic structures (e.g.
> showing a result state for resultative verbs)? I know Chomsky used to assign
> lexical semantics to the semantic structure (rather than LF, of course), but
> I assume that people like Higginbotham would exclude it even from semantic
> structure, now that this is identified with LF. I don't know what happens in
> LFG (though I'd be interested). 

This intrigues me, too.  Through Hale and Keyser's work on syntactic
decomposition of lexical meanings (building on Baker's incorporation
theory), the latest versions of LF now uncannily resemble old-time Generative
Semantics (when they are fully explicit).  The LFG semantic works
cited above do not employ lexical decomposition, and in fact,
Dalrymple et al. show that an advantage of their analysis of
quantifiers over current categorial approaches is that it can capture
the interactions with intentional contexts without requiring
"otherwise unnecessary semantic decompositions of lexical entries":
e.g. analyzing "seek" as "try to find" in order to get the semantics
of quantified objects right.

I believe that Dalrymple et al. would keep lexical semantics separate
from the compositional semantic structures, on the assumption that
there is little interaction between, say, quantifier scope and
anaphora, on the one hand, and the internal lexical semantics of verbs, on the
other.  This used to be a major issue between generative
semantics/Abstract Syntax  and lexicalist theories.  LFG remains
lexicalist.  (But see recent work on *relativized* lexical integrity
in lfg--e.g. Tara Mohanan's "Lexicality and Wordhood in Hindi" and
Bresnan and Mchombo's "The Lexical Integrity Principle: Evidence form
Bantu"  both in NLLT 1995 .)

Joan Bresnan

References:

Fenstad, Jens Erik, Per-Kristian Halvorsen, Tore Langholm, and Johan
van Benthem.  1987. SITUATIONS, LANGUAGE AND LOGIC. D. Reidel,
Dordrecht. 

Dalrymple, Mary, John Lamping, Fernando Pereira, and Vijay Saraswat.
1996. Quantifiers, Anaphora, and Intensionality. MS., Xerox PARC, Palo Alto
CA 94304.  (dalrymple at parc.xerox.com)  [see LFG biblio for other
related works]