Electroencephalography (EEG) source localization is a canonical ill-posed inverse problem: reconstructing thousands of cortical dipole moments from few scalp electrodes requires regularization. Standard methods, like Tikhonov regularization and LORETA, impose spatial constraints (minimum energy and/or spatial smoothness) but treat each time point independently. This assumption regarding temporal independence is too simplistic to describe a complex phenomenon. Neural populations exhibit intrinsic smoothness due to the neuron membranes time constants (10–50 ms), synaptic filtering, and recurrent network dynamics. Since EEG measures emitted field potentials, and not spike trains, at its typical timescales (1–100 Hz) it is safe to assume that neural activations are temporally correlated, not white.