Abstract. An electronic tongue system was developed based on 20 all-solid-state potentiometric sensors and chemometric data processing, with polymeric membranes applied on solid conducting silver-epoxy supports and a Ag=AgCl reference electrode. The sensor array was applied to 52 commercial honey samples obtained randomly from different regions of Portugal. These samples were analysed independently for their pollen profiles by biological techniques and the data collected with the tongue were evaluated for discrimination of the samples with multivariate statistical methods (principal component analysis and linear discriminant analysis), to investigate whether the device may provide an analytical alternative for classification of honey samples with respect to pollen type, a task which is time consuming and requires skilled labour when performed by biological techniques. It was found that the tongue has a reasonable efficiency for classification of honey samples of the most common three types (with Erica, Echium and Lavandula as predominant pollens). With linear discriminant analysis, the honey samples yielded about 84% classification accuracy and 72% for crossed validation. In this study, the honey samples correctly classified for the different types of the dominant pollen were: 53% for Lavandula, 83% for Erica and 78% for Echium pollen.Keywords: Honey; pollen; electronic tongue; multivariate analysis Multi-sensor arrays that provide global information on complex samples have deserved much interest recently. Instead of measuring specific parameters, these devices acquire global information which, after treatment by appropriate chemometric methods, can be used for multicomponent classification analysis, taste evaluation, etc. Electrochemical sensor arrays or electronic tongues built with non-specific, low-selectivity, chemical sensors with high stability and cross sensitivity to different species in solution, are suitable for analysing complex liquid samples [1]. Electronic tongues or taste sensors based on different electrochemical principles, such as potentiometry [2][3][4][5][6] or voltammetry [7,8], have been described. Several array types have been tested for potentiometric devices, namely chalcogenide glass sensors [3][4][5], lipid=poly-meric membranes [2,6] and ion selective membranes [9]. The signal profiles generated by such devices vary with the characteristics of different samples and upon data treatment with multivariate statistical methods for pattern recognition (identification, classification