In this paper we introduce a biologically and psychologically plausible neuronal model of hierarchical categorization. The knowledge in our model is represented by a taxonomical arrangement of verbal categories. This caregorical representation is psychologically motivated and also offers an explanation of how to deal with uncertain knowledge. It is an alternative to other well known uncertainty calculi. An observer specifies the known features before the hierarchical categorization begins. During the categorization the model learns to favor those categories which often lead to a successful goal. This may help to speed up the search. A computer simulation of a system for the diagnosis of the problem with a car is presented.