In this paper, the recently introduced concept of Nearest Centroid Neighbourhood is used to select an appropriate subset of prototypes for the Nearest Neighbour classifier. The approach consists of designating for deletion prototypes that have been misclassified by their surrounding neighbours (instead of nearest neighbours) and then discarding them from the input data set. The performance of the method has been evaluated on synthetic and real databases. The comparison results show that the proposed algorithm outperforms other well-known editing algorithms in most of the experiments.