Towards a linguistic definition of metonymy

In this work, I attempt to develop a rigorous definition of metonymy as a linguistic phenomenon. This is in response to the currently popular notion of metonymy as a conceptual phenomenon, which I find highly problematic. According to my definition, metonymy can be identified thanks to three criteria: 1) the presence of a reference shift from the normal referent of the metonymically-used noun to its intended referent; 2) the presence of a semantic type shift in the metonymically-employed noun, or of some gap in the information conveyed by the metonymically-employed noun which is filled via free pragmatic enrichment; and 3) the presence of conceptual contiguity between the entities involved in the metonymy (the source and target). I built a corpus of 300 examples taken from the linguistic literature using these criteria. I also address the issue of metonymic productivity, suggesting that it is motivated by principles of cognitive and communicative salience (some of which have been listed by Radden and Kövecses, 2007). The novelty of this proposal lies in the idea that metonymic productivity is not motivated just by one principle, but by many of them interacting in a complex way.