Gibbs and Steen (eds.): Metaphor in Cognitive Linguistics (1999)

In the introduction to this collection, Raymond Gibbs and Gerard Steen warn their readers that

there may not be a direct mapping between linguistic metaphor and conceptual metaphor, on one hand, and between linguistic metaphor and individual cognition, on the othter. (p. 4)

That is, conceptual metaphors may not explain the actual linguistic data; and even if it does, it may not be good model of what goes on in the head (see also p. 2).

Unfortunately, Gibbs and Steen's solution is to posit a "jigsaw" model in which every individual only has a piece of the puzzle:

A complete conceptual metaphor may only emerge from examination of the communication between, or across, participants in some community. This examination yields a "supra-individual," cultural class of metaphors. (p. 3)

And even when a particular individual has the piece of the puzzle stored in his or her brain, the "pre-stored conceptual metaphors may not always be activated" in the expected situations.

This way, cognitive metaphor theory gets to keep its cake and eat it too: It recognizes that there aren't actually any conceptual mappings in there (as shown by experiments), but they are still "out there" (as posited in our speculative analyses). The consequence is of course that the conceptual mappings only appear to those who believe.

Gibbs and Steen proceed to notice that many metaphors lack experiential basis (presumably they're thinking of something like ARGUMENT IS WAR).