Constructions versus words and rules
In their influential 1988 paper, Fillmore, Kay and O’Connor challenge the ‘words and rules’ approach assumed by the standard generative model. According to this model, the properties of language can be accounted for by a system of ‘words and rules’, where the words are the individual lexical items in the speaker’s lexicon, and these words are subject to rules of different types within the language system. Phonological rules govern the assembly of complex strings of sounds. Syntactic rules govern the assembly of words into grammatical structures such as phrases and sentences, while semantic rules assign a semantic interpretation to the clause according to the principle of compositionality. As we saw in PartII of the book, this principle holds that the meaning of a sentence arises from the meanings of the words it contains, together with the way in which these words are syntactically arranged. This gives rise to propositional meaning, a ‘purely semantic’ meaning that is independent of context. In addition to syntactic and semantic rules, speakers also have knowledge of pragmatic principles that map propositional meaning onto context and guide the hearer in drawing the relevant inferences. Crucially, as we saw in Part I of the book, this approach is modular in the sense that syntax, semantics (and phonology) are encapsulated subsystems that only communicate with one another via linking rules. This type of model can be represented by the diagram in Figure 19.1.
Observe that there is no ‘pragmatics box’ in this model. As we saw in Part II of the book, this is because the standard generative model views pragmatic knowledge as peripheral to linguistic knowledge ‘proper’ in the sense that pragmatic knowledge involves the interface between language and other systems of knowledge and information processing. This model of speaker knowledge only accounts for what is regular in language, and leaves aside idiomatic units, which, according to (Fillmore et al. 1988: 504), have the status of an ‘appendix to the grammar’. In other words, in the standard generative model, the only the complex units that are ‘stored whole’ are those whose properties cannot be predicted on the basis of the regular rules of the grammar. As we saw in Chapter 1, idiomatic expressions like ‘kick the bucket’ fall into this category.
According to Fillmore et al., this appendix is not only very large, but also has the potential to reveal much about how language works. For this reason, as we will see in the next two sections, they propose a model of language that accounts for idiomatic constructions not as an exception to the norm, but as a central feature of human language. Furthermore, Fillmore et al. propose that the same theoretical machinery should be held to account for both regular and idiomatic grammatical units.
