- Reply to Kac
In our article ‘Compositional idioms’ (Language 76.409–32), Jerrold Katz and I identified a large and productive class of expressions, typified by plastic flower, stuffed animal, rubber chicken, and kosher bacon, which, we claimed, have a distinctive semantics involving both idiomatic and compositional components.1 Our account of these expressions was motivated by a puzzle we encountered in attempting to accommodate certain intuitive semantic facts about them. For example, while the analyticity (and truth)2 of plastic flowers are not flowers, the contradictoriness of plastic flowers are flowers, and the antonymy of plastic flower and flower suggest some kind of incompatibility between the meanings of plastic and flower, any such incompatibility would deprive plastic flower of an extension, and prevent plastic flowers are not flowers, plastic flowers are plastic, plastic flowers are fakes, and so forth, from being true. We thus seemed to be faced with two conflicting sets of facts.
We resolved the apparent conflict by supposing that plastic flower is ambiguous.3 It has a conjunctive reading, on which it means ‘flower made of plastic’, and a nonconjunctive reading, on which it means ‘imitation flower made of plastic’. Since, in general, imitation Fs are by definition not Fs, expressions of the forms imitation F and F, for any simple predicate F, will be antonymous. Plastic flower on its nonconjunctive reading is antonymous with flower because imitation flower is antonymous with flower. For the same reason, plastic flowers are flowers is contradictory and plastic flowers are not flowers is analytic on their nonconjunctive readings (imitation flowers made of plastic are flowers is contradictory, and imitation flowers made of plastic are not flowers is analytic).4 Since, however, plastic and flower are not themselves semantically incompatible, plastic flower can have an extension, and plastic flowers are not flowers, plastic flowers are plastic, plastic flowers are fakes, and so on can be true.
Our account is further supported by the case of stuffed animal, where the ambiguity is especially obvious.5 There are stuffed animals in a straightforwardly conjunctive sense—the things produced by taxidermists—as well as stuffed animals in a nonconjunctive [End Page 197] sense—the things produced by toy companies. Stuffed animals in the first sense are animals and in the second sense are not. Hence, for example, taxidermy is the art of producing stuffed animals is analytic (and true) on the conjunctive reading of stuffed animals, and false on the nonconjunctive reading.
Given that the concept of an imitation is to be found in neither the sense of the head nor the sense of the modifier in these constructions, it follows that the semantics of plastic flower, et al. on their nonconjunctive readings is both noncompositional and decompositional. It is noncompositional because the meanings of these expressions are not determined by the meanings and syntactic relations of their syntactic constituents. It is decompositional because the meanings of these expressions contain elements that are not the meanings of any of their syntactic constituents. The underived element imitation is assigned to a nonterminal node in the syntactic structure of the phrase, in the manner of an idiomatic interpretation. In order to accommodate the productivity of these constructions (paper flower, stuffed African elephant, kosher Canadian bacon), we proposed that the underived element take the form of an ‘idiom schema’, for example, imitation X made of [by] Y[ing]. Productivity is accounted for by the combination of readings for heads and modifiers with the idiom schema in the derivation of a phrasal reading.
We argued that extensionalist semantic theories cannot accommodate the nonconjunctive readings of these expressions, since there are no distinctions among the extensions of their modifiers and heads, or among the intensional functions determining such extensions across possible worlds, that can capture them in a non–ad hoc way. Michael Kac (‘The semantics and pragmatics of appearance’) has suggested that, after all, there is a way for a purely extensional semantics to account for the intuitive facts about these constructions (which Kac calls oxymoroids). All one need do is introduce a principle by which the extension of a common noun N is generalized to include not only Ns, but also...