Fake Guns and Stone Lions: Conceptual Blending and Privative Adjectives

Kamp (1975) defines a privative adjective as one for which, given adjective A and noun N, the claim 'No AN is a N' is necessarily true.For example, 'fake' is a privative because 'No fake gun is a gun,' is true by definition. In order to explain how 'fake' acts as a modifier, the semanticist's job has been to provide a mechanism for specifying the similarities and differences between real and fake guns. Because, at least at first blush, privatives pose a greater problem for the semanticist than, say 'brown cow,' Franks (1995) has argued that privative constructions should be employed as a test for models of conceptual combination.One such model is Franks' (1995) sense generation model. Another is offered by conceptual blending theory (Coulson, 1997; Fauconnier & Turner, 1996). In this paper we contrast the treatment of privative adjectives in these two models. Specifically, we compare how the examples 'fake gun' and 'stone lion' are dealt with in conceptual blending theory and in the sense generation model. Although both models are motivated by the observation of substantial flexibility and context-sensitivity in actual instances of meaning construction, these phenomena are dealt with differently in the two models. While Franks' approach is an attempt to extend objectivist semantics to explain data which resist a strictly compositional account, our own model is motivated by ideas in cognitive semantics. In contrast to the view of language as a code which combines context-invariant meanings, the cognitive semanticist argues that language provides clues for the language user to flexibly address her knowledge base so as to construct a contextually appropriate message-level representation.