Construction Grammar: a broadly generative model
It is important to reiterate the fact that Kay and Fillmore’s Construction Grammar model is a formal model. In other words, it requires the statement of exhaustive, precise and unambiguous theoretical machinery that is intended to be sufficient in accounting for the properties of human language. As we have observed, and as Kay and Fillmore themselves acknowledge, their model is reminiscent of other broadly generative formal models, particularly models like HPSG that assume a non-transformational monostratal syntax. What ‘broadly generative’ theories have in common is that they assume Universal Grammar as a working hypothesis, and attempt to build a model that represents this knowledge. In other words, these are not usage-based theories.
The differences between non-transformational generative models on the one hand and the transformational generative model on the other are obvious. While the transformational model captures phenomena like the wh-dependency by means of two syntactic representations linked by a trans formation or movement operation, the monostratal generative models assume a single syntactic representation and build into that representation features or ‘tags’ that capture the same linguistic phenomena. We have seen how Kay and Fillmore’s Construction Grammar accounts for the wh-dependency, for example, by means of the left-isolation construction.
The differences between Kay and Fillmore’s Construction Grammar and another monostratal model like HPSG, however, may be less obvious but are no less important. The HPSG model is, like the transformational generative model, a ‘words and rules’ model. In other words, it assumes a lexicon in which items are tagged with a complex and detailed set of features (including cate gory, valence, number and so on) and a set of rules that assemble those lexical items into a syntactic structure. For example, HPSG assumes a head complement rule and a subject-predicate rule, which are ‘structure building rules’ in the same sense as the phrase structure rules that operate within the transformational model. Although meaning and grammar are arguably more closely integrated in HPSG than in the transformational generative model, HPSG can still be described as a modular theory, particularly given that it assumes the autonomy of syntax.
In contrast, as we have seen in this chapter with respect to Construction Grammar, and as we saw in our discussion of Cognitive Grammar in Chapters 16–18, a constructional model does not assume ‘words and rules’ but instead assumes ‘ready-made’ grammatical constructions, some of which are highly detailed and some of which are highly generalised. A further important difference between the HPSG model and Kay and Fillmore’s Construction Grammar is that the latter assumes that non-compositional meaning (such as the incongruity judgement associated with the WXDY construction) is directly linked to the grammatical construction itself. Furthermore, this meaning is linked to the construction as a whole rather than being derived from some subpart of the construction. This is important because it shows that the constructional model is not modular. In other words, constructions contain information about syntax, morphology, semantics and pragmatics (and, in principle, phonology) within a single integrated representation.
