The primary aim of the study was to determine the best electrode positions for EOG signals in vigilance studies. Two-channel recordings were conducted in analogy to the Rechtschaffen and Kales (1968) system. Twenty electrodes (10 electrode pairs) were compared. Both EOG amplitudes and amplitude asymmetries within an electrode pair were studied. The amplitude of the EOG signal is sensitive to relatively small differences in electrode position. This concerns especially distance from the eye, the direction of eye movement and the effect of the upper eye lid movement. Larger and more symmetrical EOG amplitudes were obtained for different eye movements by placing the electrodes more medially than in the conventionally used system. EOG asymmetry in different electrode positions was dependent on the eye movement direction and even on the starting and end points of a movement with equal angular degrees. Most of the data could be explained by a simple monopolar model when combined with the effects of the upper eye lid movements. The most unexpected finding was that the EOG amplitudes of the horizontal and oblique eye movements were significantly larger when the eyes were moving towards an electrode than when they were moving to the opposite direction.