Analogy Comprehension as Complexity Reduction

In this paper we try to find an explanation for the interesting phenomenon that competent langugage users have little difficulty understanding non-trivial metaphors in ordinary discours. According to the standard interaction view, in metaphor commonplaces associated with the vehicle are projected on the topic. This account has to be dismissed as too general (when not untrue in several cases). On the basis of an elaborated theory of similarity we propose an alternative, more precise, model for analogy comprehesion. This model involves a systematic reduction of complexity in three stages: in stage one the relevant (manifest, specific, and ordinary) features of the vehicle are selected. In stage two we investigate which of these can be meaningfully applied to the topic. Generally, this will reduce even further the set of possible relevant features. The last stage has an important control function: we check whether the remaining set of features makes sense within the linguistic context in which the metaphor occurs. In the light of new data, we may be forced to delete (some of) the features found in stage two and replace them by new ones. By means of this model we are able to reconstruct logically the functioning of metaphors as well as other types of analogy.