Data from magnetoencephalography (MEG) andelectroencephalography (EEG) suffer from a rather limitedsignal-to-noise-ratio (SNR) due to cortical backgroundactivities and other artifacts. In order to study the effect ofthe SNR on the size and distribution of dipole clustersreconstructed from interictal epileptic spikes, we performedsimulations using realistically shaped volumeconductor models and extended cortical sources with differentsensor configurations. Head models and corticalsurfaces were derived from an averaged magnetic resonanceimage dataset (Montreal Neurological Institute). Extended sources were simulated by spherical patches withGaussian current distributions on the folded cortical surface. Different patch sizes were used to investigate cancellationeffects from opposing walls of sulcal foldings andto estimate corresponding changes in MEG and EEG sensitivitydistributions. Finally, white noise was added to thesimulated fields and equivalent current dipole reconstructionswere performed to determine size and shape of theresulting dipole clusters. Neuronal currents are orientedperpendicular to the local cortical surface and show cancellationeffects of source components on opposing sulcalwalls. Since these mostly tangential aspects from largecortical patches cancel out, large extended sources exhibitmore radial components in the head geometry. This effecthas a larger impact on MEG data as compared to EEG,because in a spherical head model radial currents do notyield any magnetic field. Confidence volumes of singlereconstructed dipoles from simulated data at differentSNRs show a good correlation with the extension ofclusters from repeated dipole reconstructions. Size andshape of dipole clusters reconstructed from extended corticalsources do not only depend on spike and timepointselection, but also strongly on the SNR of the measuredinterictal MEG or EEG data. In a linear approximation thesize of the clusters is proportional to the inverse SNR.