In this paper we want to show the use of inverse solutions ‒in particular cortical current density (CCD)‒with respect to the current surface electroencephalogram (EEG) in classifying event related potentials. We construct feature vector from selected channels which had highest discriminant power for a two class problem. We present the results for 6 subjects, where in general, the CCD based method performs significantly better in comparison to the surface EEG methods.