Reply to R. Jones, "Dale and Compositional Semantics"

Monday, February 28, 2011
New York, NY

This post is in response to Roger Jones's "Dale and Compositional Semantics" post of Monday, February 28, 2011.

Hi Roger. I didn't and don't take you to be rude at all, but merely critical. And, I thank you for that. Philosophical work would not be possible without open and vigorous opposition. So, thank you.

Chapter 4 of my dissertation (The Theory of Meaning (1996); go to russelldale.com for a copy) is in many ways the most radical chapter. I defend Schiffer's idea that our physicalist sensibilities can be satisfied short of a compositional meaning theory. Most philosophers, and notably Fodor--who responded directly to Schiffer's suggestion in the book on Fodor in which Schiffer first really published this possibility (Lower and Rey, Meaning in Mind: Fodor and his Critics, 1991)--reject this perspective.

My views have changed somewhat, but not in the direction of compositional meaning theories. For there to be a compositional meaning theory for a natural language three things are required: propositions, meanings for sub-sentential expressions (particularly words), and a finitely statable mapping from the latter (along with a characterization of the syntax of sentences) to the former. I don't think there are either clearly defined contents that can function as propositions, nor that there are clearly defined entities that can function as the meanings of sub-sentential linguistic entities--and, most specifically, words. My reasoning is very dialectical here, and my dissertation is a first statement of that dialectic. Schiffer's work is also fundamentally helpful in this regard. (I always liked, the form of his paper "The Mode-of-Presentation Problem" (1990) in Anderson and Owen, Propositional Attitudes: The Role of Content in Logic, Language, and Mind (1990). I don't mean to say that this paper is the key to the issue in substance, but only that the form of the argument there is crucial (even if that paper is dealing with "modes of presentation" and not propositions): nothing is available to possibly play the role, therefore, there's no such thing.)

The process for finding such a thing as a proposition is completely tied up with the process of finding such a thing as a sub-sentential expression meaning as well as the mapping from the latter to the former. You can't do with one without the others. But, the three things cannot be found. It is like trying to solve a system of simultaneous linear equations in three variables, where there is no solution. Each individual idea seems to make sense on its own, but they are really not independent in the relevant sense, and contradiction follows from any attempt to find a solution. Therefore, there is no solution. But, of course, the problem is not a simple set of three linear equations in three variables where this can be shown by straightforward techniques. To see the point requires going through many, many specific attempts to solve the problem and seeing that in principle, none of them works, and there really aren't any other possibilities. The process is made even more complex, obviously, by the possibility of pragmatic solutions to counterexamples to theories that employ specific conceptions of meanings and mappings between them. The process of coming to these arguments is highly complex. In any event, I have become convinced that nothing can play the roles of meanings in the sense required by the existence of a compositional meaning theory, and therefore, there are no compositional meaning theories.

None of this entails scuttling the obvious appearance of compositionality or of what Chomsky calls the productivity and systematicity of language. All that is possible (and I think obviously must be) without meanings in the technical sense that has been desired by those philosophers who have believed in compositional meaning theories.

Does that mean that computer scientists (I am one myself) must give up formalizing languages and making computer models of language? No. Engineering problems don't have anything necessarily to have anything to do with what is actually going on inside of human heads.

So, have I changed my views in the past 15 years? Yes. Do I now believe in compositional meaning theories for natural languages? No. Does it matter to any practical engineering project? No. Does it matter in other ways? Yes. (But, I will leave that for now.)

Mind you, I am not pretending to have given my arguments here. As I say above, the start of my arguments are in my dissertation, and in Schiffer's work. The end of my arguments are just in my private noodle for the time being, my not having the leisure to write all this stuff out and publish it. So, I apologize for that. I would be curious to see why you disagree, if you do, with my assessment of the possibility of their being a compositional meaning theory.

--Russell Dale