Ever since Montague’s seminal work in the late 60’s and early 70’s Formal Semantics has worked something like this: (i) Specify a syntax for a fragment of some language; (ii) specify a class of models: (iii) define, by a recursion that follows the syntax, for each expression E of the fragment and each model M from the model class the semantic value of E in M. In particular, such a definition must assign to each sentence S and model M a truth value of S in M, or else a proposition that gets a truth value at each of the relevant different ‘indices’ of M.
As I have just stated it, this schema is an oversimplification. The semantic values of many NL expressions depend on various aspects of context. Therefore ‘semantic value’ must be treated, minimally, as a function with three arguments: (a) expression, (b) model and (c) context.
The ‘Formal Semantics Approach’ (FSA) has been implemented in various ways, e.g. by defining the semantic values directly from the syntactic structure or indirectly, via an intervening semantic representation or logical form. But common to all those implementations is a user-neutral stance: Semantic values (of expressions in models, given contexts) are treated as properties of the language as autonomous system. (Competent users of the language are those that behave in conformity with the properties it has qua autonomous system.)
We will contrast this perspective with that of the Communication-Theoretic Approach (CTA), in which language is studied as a device for communicating information between language users. According to CTA the semantics of the language resides in the coding principles that allow speakers to express their thoughts and the decoding principles that enable their audiences to recover those thoughts from their words. Context is as indispensable in CTA as it is in FSA, but now contexts have to be treated as part of the mental states of language users.
I argue that some phenomena are better handled within a CTA framework than with the traditional methods of FSA. The examples I will discuss have to do with how speakers choose the noun phrases they need to refer to the things they want to talk about and how their addressees interpret those noun phrases. The central part of the talk will be the sketch of a CTA account of these phenomena. After that I will look at a few questions about the relationship between FSA and CTA. In particular:
1. What are the prospects for a rigorous formalization of CTA?
2. Can FSA be regarded as a kind of approximation to CTA?
3. Can FSA be embedded in some way within CTA, and under what conditions?
[Note from local organization: slides available here.]