The problem of metaphoric interpretation has been a neglected topic in computational systems for natural language interpretation. Part of the reason for this is a presumptfon of semantic well-formedness which excludes metaphors as semantically ill-formed. We present a theory of metaphorical interpretation based on lexical knowledge representation by ontDlogical category attribution. As evidence of support for the theory we found empirical studies of English speakers interpreting classifier/noun pairs predicted as different types of interpretationas predicted. We sketch the design of a computational system based on the theory. Finally, we discuss the implications of the theory for semantic interpretation.