To improve the spatial resolution of human event-related potentials, we developed a new high resolution EEG method based on the improved estimate of the realistic surface Laplacian (SL). The novelty of this method consisted in the computation of the local scalp resistance that was assumed to be inversely proportional to the local scalp thickness measured from magnetic resonance images of the subject's head. The local scalp thickness was then multiplied by the SL estimate of the potential over a realistic magnetic resonance-constructed model of the subject's scalp surface. The new method was applied on human movement-related and somatosensory-evoked potentials, the SL estimate at a constant scalp thickness being used as a reference. The locally-predicted scalp thickness was significantly (P < 0.05) higher in the temporal areas (9.5 +/- 2.6 mm) than in the parieto-occipital (6.6 +/- 1.3 mm) and frontal (4.8 +/- 1.1 mm) areas. Compared to the SL estimate at constant scalp thickness, the improved SL estimate enhanced the spatial detail of both movement-related and somatosensory-evoked potentials.