Page 128 - Encyclopedia of Philosophy of Language

Truth and Meaning
and Ladusaw (1979), and Janssen (1986a) for a detailed description of the mathematical model.
6. Objections Against Compositionality
Several objections have been raised against com- positionality. Some authors, investigating phenomena that seem to call for a noncompositional solution, argue that natural language is not compositional at all. But often a non-compositional proposal can be replaced by a compositional one, which then not only turns out to solve the original problem but also has a wider area of application. Partee (1984) considers many phenomena where context seems to play a role, e.g., the different interpretations of the subjects in 'the horse is widespread' versus 'the horse is in the barn.' She refers in some cases to proposals in the literature where a satisfactory and compositional solution is given, but other phenomena remain a challenge.
Not all phenomena of natural language have (yet) found a compositional solution. And those solutions that have been or can be proposed are of course always open to further debate. In all cases, however, the right question to ask is not whether a compositional treat- ment is in principle possible (it always is), but what price has to be paid, or what the reward of a com- positional solution is.
A different objection, of a more methodological nature, concerns the various kinds of meanings used in Montague Grammar. If meanings are built from meanings of structural parts, then meanings of parts tend to become highly abstract. One kind of meaning which has proved to be of great use is the type of term phrases, which led to the theory of generalized quantifiers. But other kinds seem to be motivated only by the wish to keep the system compositional. Montague Grammar has been criticized for its appar- ent willingness to accept any kind of abstract entity in the semantic domain. And indeed, if one expects of a semantic theory that it not only accounts for the semantic phenomena of language but, in addition, does so under certain philosophical constraints, such as maximal simplicity of the used elements, then there can be a problem.
Compositionality is not an empirically verifiable property of natural language. Compositional rules, powerful as they are, can seem quite counter intuitive. There are authors who reject Compositionality for this reason, or replace it by a more restrictive version. One may propose to restrict it to grammars in which the derivations are closer to the surface form. However, if restricted too much, the principle may no longer be applicable to natural language, and its advantages will be lost. Several proposals for restrictions have been put forward. Partee (1979) proposed to use only a few basic syntactic operations. Generalized phrase struc- ture grammar (Gazdar, et al. 1985) can be seen as a form of Montague Grammar in which only context- free rules are used, and categorial grammar can be
seen as an even more restricted form. Hausser (1984) aims at what he calls 'surface compositional grammars.'
An argument against Compositionality sometimes raised by grammarians in the Chomsky tradition has to do with the position of syntax in a compositional grammar. They defend the principle of the autonomy of syntax, which is meant to specify the well-form- edness conditions of sentences on syntactic grounds only. As has been shown in Sect. 2.5, a compositional grammar, being homomorphically connected with the semantics (so that every non-lexical ambiguity is syn- tax-based), allows for semantic considerations to have relevance for the selection and formulation of the rules of syntax. The very notion of autonomous syntax thus implies a rejection of such semantic considerations in syntax. But, as Gamut (1991:148) says, 'It remainsan open question whether this potential transgression of the autonomy of syntax by semantics will be enco- untered in reality, that is, in the actual description of some natural language.'
7. WhyCompositionality?
An interpretation of the principle of Compositionality of meaning has been discussed here, and somewhat implicitly, the argument has been in favor of it. A theoretical argument is that Compositionality makes for an attractive framework in which semantic prob- lems are dealt with as locally as possible, the various solutions being combined into larger wholes. Com- positionality thus provides a strategy for dealing with the complexities of language. A practical argument is that experience has taught us that observance of Compositionality usually leads to better solutions, which makes it a useful and attractive heuristic device.
See also: MontagueGrammar.
Bibliography
Gamut L T F 1991 Logic. Language and Meaning. Vol. II: Intensional Logicand LogicalGrammar.University of Chicago Press, Chicago
Gazdar G, Klein E, Pullum G, Sag 11985 Generalized Phrase Structure Grammar.Basil Blackwell, Oxford
Groenendijk J, Stokhof M 1991 Dynamic predicate logic. Linguistics and Philosophy 14:39-100
Halvorsen P K, Ladusaw W A 1979 Montague's 'Universal Grammar.' An introduction for the linguist. Linguistics and Philosophy 3:185-223
Hausser R 1984 Surface Compositional Grammar.W. Fink, Munich
Janssen T M V 1986a Foundations and Applications of Mon- tague Grammar. Pan I: Foundations. Logic, Computer Science. CWI Tract 19. Center for Mathematics and Com- puter Science, Amsterdam
Janssen T M V 1986b Foundations and Applications of Mon- tague Grammar. Part 2: Applications to Natural Language. CWI Tract 28. Center for Mathematics and Computer Science, Amsterdam
106