In this paper I am proposing a cognitive theory and description of the semantic structure of figurative language and rhyme. My assumption is, that an adequate theory of metaphor must satisfy four requirements: 1. to give a structural description of metaphors; 2. to explain how human beings understand novel metaphors; 3. to explain the relationship between this process and the process by which human beings produce and understand novel pieces of literal discourse; 4. to explain the relationship between these processes and the perceived effects of metaphors. The semantic information processing model underlying this paper is a hierarchic model of “meaning components’*. The same model can explain word acquisition by infants, word association, the understanding of new metaphors, the perceived effects of rhymes and other cognitive phenomena. A theory of minimum features cannot do justice to the richness of real-life categories. So the model is further developed in three directions: Mervis’ work on “good examples”; Collins and Quillian’s work on the structure of semantic memory; Rumelhart’s work on “cognitive schemata”.