A Reconsideration of English Relative Constructions
A main goal of this article is to lend support to the view that the construction is the proper level of description for accounting for grammatical phenomena. I assume that there is a spectrum of complexity for constructions, along the lines of Culicover (1999) and Culicover and Jackendoff (2005). At one end of the spectrum are very regular correspondences between form and function that in more traditional approaches to grammar have been described as ‘rules’, in the middle there are more complex constructions that inherit many of their properties from the regular correspondences but deviate from them in specified ways, and at the other end are correspondences that are highly idiosyncratic and opaque. The main point, which echoes the perspective of many if not all constructionalist approaches to grammar, is that constructions do not differ from one another in completely arbitrary ways, but cluster around the regular correspondences of a language, while maintaining their special distinguishing properties of form and function. In support of this view I propose a reformulation of Sag’s (1997) constructional account of English relative constructions. I show how it is possible to characterize the various types of relatives in such a way that almost all of their properties follow from more general properties of the language, and in some cases possibly even from universals. Nevertheless there are some special facts about relative constructions that need to be specified explicitly, and these contribute to their grammatical complexity. One goal of this paper is to clearly separate the properties of relative clauses that follow in virtue of their being embedded in NPs from the properties that must be stipulated. It appears, in fact, that it is not necessary to say very much at all about the properties of relative clauses qua relative clauses — they look in virtually all respects like other embedded clauses in English. A second goal is to show how relative clauses and other relative constructions constitute a natural class, and do not require any special stipulations. Moreover, the binding relations that govern gaps in relative clauses follow directly and uniformly from the general interpretation assigned to relatives, as specified in the constructions that express their form-meaning correspondences. In the end, it appears that the only true idiosyncrasy of relatives that must be stated explicitly is that an overt wh-XP in initial position in infinitival relatives must be PP.