Beyond literal similarity.

It is argued that even those extant theories of similarity (e.g.Tversky, 1977) that are sensitive to the fact that many similarity statements are asymmetrical are unable to deal with a number of important symmetry related issues.In particular, it is claimed that an entire class of similarity statements remains largely unaccounted for.These statements comprise nonliteral similarity statements such as similes.It is suggested that what is needed is some way of relating similarity to nonliteralness, or metaphoricity.A proposal for doing this based on a modification of Tversky's contrast model, and on comparisons of the relative degree of salience of attributes of the two terms that are shared or shareable is offered.The ramifications of this proposal are reviewed and the central issues that a marriage between a theory of similarity and of metaphoricity ought to address are identified.