The present paper describes the propagation of ocular potentials across the scalp on a biophysical basis. It is concluded that 3 EOG derivations (two for EEG records along the midline) are generally necessary to account for ocular disturbances in the EEG. The inadequacy of many methods suggested for EOG artifact control may be due to the false assumption that just one EOG derivation provides enough information to remove ocular potentials from any EEG recording along the mid(-sagittal) line. A comparison of compensation with one or with two EOG derivations is described for a data set of slow brain potentials. A frequency dependence of the ocular influence cannot be neglected, if fast and slow EOG activities have to be removed. The present considerations should allow a more theoretically based decision of the EOG correction method necessary for a certain data set.