Mesial temporal sources are presumed to escape detection in scalp electroencephalographic recordings. This is attributed to the deep localization and infolded geometry of mesial temporal structures that leads to a cancellation of electrical potentials, and to the blurring effect of the superimposed neocortical background activity. In this study, we analyzed simultaneous scalp and intracerebral electroencephalographic recordings to delineate the contribution of mesial temporal sources to scalp electroencephalogram. Interictal intracerebral spike networks were classified in three distinct categories: solely mesial, mesial as well as neocortical, and solely neocortical. The highest and earliest intracerebral spikes generated by the leader source of each network were marked and the corresponding simultaneous intracerebral and scalp electroencephalograms were averaged and then characterized both in terms of amplitude and spatial distribution. In seven drug-resistant epileptic patients, 21 interictal intracerebral networks were identified: nine mesial, five mesial plus neocortical and seven neocortical. Averaged scalp spikes arising respectively from mesial, mesial plus neocortical and neocortical networks had a 7.1 (n = 1,949), 36.1 (n = 628) and 10 (n = 1,471) µV average amplitude. Their scalp electroencephalogram electrical field presented a negativity in the ipsilateral anterior and basal temporal electrodes in all networks and a significant positivity in the fronto-centro-parietal electrodes solely in the mesial plus neocortical and neocortical networks. Topographic consistency test proved the consistency of these different scalp electroencephalogram maps and hierarchical clustering clearly differentiated them. In our study, we have thus shown for the first time that mesial temporal sources (1) cannot be spontaneously visible (mean signal-to-noise ratio -2.1 dB) on the scalp at the single trial level and (2) contribute to scalp electroencephalogram despite their curved geometry and deep localization.