In this paper we present a hybrid approach to handwritten symbol recognition based on two different methods and principles. A fuzzy rules based recogniser and a Self-Organizing Map recogniser are combined to form our hybrid system. These two systems complement each other well, firstly because their feature extraction techniques differ greatly, and secondly because one is a model-based and the other is a discriminative classifier. Each system generates a ranked list of outputs with associated confidence values, and these outputs are combined to produce a single result. The approach has achieved high recognition rates in testing on digits and lowercase characters from the UNIPEN database.