Fuzzy logic is highly suitable for detailing with uncertainty and variation. Therefore it is reasonable to apply this technique to the recognition of handwritten symbols. This paper presents an approach to the task in which fuzzy logic is used extensively. We present a three-phase process, the central phase being feature extraction. Firstly a pre-processing phase generates a chord vector for each handwritten stroke, thereby eliminationg noise and greatly reducing the number of sections of the input which need to be assessed as potential features. In the feature extraction phase fuzzy rules are used to determine membership values of chord sequences in fuzzy sets corresponding to feature types, and subsequently the most likely set of features is determined. In the final phase, fuzzy classification rules are used to determine the most likely identity of the symbol according to the feature extraction result. The approach has achieved high recognition rates in experiments on isolated symbols from the UNIPEN database.