A method using the artificial neural network Fuzzy ARTMAP (FAM) was developed to classify cyclic voltammograms according to underlying reaction mechanisms. Different preprocessing methods for reducing input dimensionality, including principal component analysis (PCA), feature extraction, and wavelet transform (WT), were compared. Results obtained for simulated and experimental voltammograms show that FAM can be applied to the classification into E, E-qr EC and EqrC mechanisms successfully. The efficiency of WT for data compression was also confirmed. Experiments demonstrate a significant correspondence between misclassifications and intersections of class distributions for different reaction mechanisms. It was found by analyzing the error distributions of FAM that the most classification errors arise in the overlapping areas of two reaction mechanisms. The relationship of the resulting class distribution to the mechanistic zones of classical zone diagrams is discussed. (c) 2005 Elsevier B.V. All rights reserved.