A new ''electronic tongue'' has been developed and used for the qualitative analysis of natural waters. Both wires and conducting surfaces in thick-film technology were used as potentiometric electrodes and the emf of each electrode in contact with a certain aqueous solution was used as input signal for a PCA analysis. An array containing the electrodes; RuO 2 (with resistivities of 10Ω/sq and 1MΩ/sq), C, Ag, Ni, Cu, Au, Pt, Al, Sn, Pb and C (graphite) were used in a first approach. To test this ''electronic tongue'' design, a family of different natural waters were studied. These include seven mineral waters, tap water and osmotised water. Additionally, a qualitative analysis of the different waters was performed using fuzzy ARTMAP neural networks. The final system was successfully employed for the differentiation of the above mentioned waters with a success rate higher than 93%.