Syntactic mimicry as evidence for prototypes in grammar
Elaine J. Francis
This paper examines categorial mismatch phenomena discussed by McCawley (1987) and explores their implications for a prototype theory of syntactic categories. McCawley (1987) identifies nouns like snap, bitch and breeze as ‘adjectival nouns’ due to their adjective-like semantic function (a). According to McCawley, these words involve syntactic mimicry: “surface configurations that admit nouns are used to simulate syntactic constructions typical of adjectives” (McCawley 1987: 459).
(a) The problem was a snap / bitch / breeze to solve.
Adjectival nouns present some puzzling patterns of grammaticality whereby sentences that appear to be both syntactically and semantically well-formed are nevertheless ungrammatical (McCawley 1987: 464-465):
(b) *Finnegan’s Wake and Ulysses are a bitch of books to read.
(c) *John solved every bitch of a problem that they asked him to solve.
McCawley argues that (2) and (3) are ungrammatical because of the explicitness with which these nouns violate the prototypical syntax-semantics mapping (i.e., by being modifiers rather heads). In (2), the violation is made obvious by the plural morphology (applying to books rather than bitch), and in (3) the violation is made explicit by the quantifier every (applying to problem rather than bitch). Extending McCawley’s analysis, Francis argues that such data crucially support a ‘prototype’ theory of syntactic categorization, whereby syntactic categories are prototypically linked to semantic categories. Contrary to Newmeyer (1998), Francis argues that prototypes are significant for grammatical competence because they constitute expectations that speakers and listeners use in sentence processing. While some violations of these expectations are permissible, as in (a), others are not, as in (b-c). Reasons for this, and further data and analysis, are discussed within a multi-modular, prototype theory of syntactic categories (see Francis 1999).