Contents Introduction Similarity versus contiguity?

4.3. (Some) metaphor as double metonymy

One type of theoretical move that considerably muddies the figurative water is to claim that metaphor in general, or some types of metaphor, are actually composed of metonymy in some way. For instance, the analyses of Riemer (2001, 2002) and Barcelona (2000b) indicate that any metaphor whose meaning is construed as transferring features identically from source to target can be viewed as metonymy, because those features can be viewed as metonymically linked to both the source item and the target item. (See also Haser [2005: 25].) In this vein Riemer mentions the view of Group (1981) that metaphor is double synecdoche (therefore double metonymy). That is — with some deviation from Group ’s own terminology — there is a metonymy from the source concept to a feature also possessed by the target concept, and from that feature to the target concept. Alternatively, this can be viewed as a metonymy from the source concept up to a shared superordinate category (the category of things that possess the feature in question) and then down to the target concept.

A theoretician is free to regard metaphor as composed of metonymy in some such way, but there is a sense in which it does not really matter — it is a labelling move that leaves undisturbed the important point, namely that certain types of language involve certain types of conceptual link, or collections of links, where individual links may have qualities of similarity, contiguity or both, or may have other qualities; where the assembly of links used may be some 24 strange mixture of types; and where different instances of language can structurally and procedurally arrange the links involved in a large number of different ways.

Notice also that casting metaphor as metonymy does not of itself mean that the importance of, say, similarity in metaphor is downgraded. For example, in a double-metonymy view we still have the issue of what particular shared attribute (or set of attributes) or covering abstraction is being used, i.e., what the similarity is. It is just that this similarity is being theoretically structured as being via metonymies involving shared properties or via a covering abstraction. And of course we cannot couple just any two metonymies together to get a metaphor.