Cathodal galvanic currents activate primary vestibular afferents while anodal currents inhibit them. Pulsed galvanic vestibular stimulation (GVS) was used to determine the latency and initiation of the human vestibuloocular reflex. Three-dimensional galvanic vestibuloocular reflex (g-VOR) was recorded with binocular dual-search coils in response to bilateral bipolar 100ms-rectangular-pulse of current at 0.9mA (near-threshold), 2.5mA, 5.0mA, 7.5mA and 10.0mA in 11 normal subjects. The g-VOR comprised three components: conjugate torsional eye rotation away from cathode towards anode; vertical divergence (skew deviation) with hypertropia of the eye on the cathodal and hypotropia of the eye on the anodal sides; conjugate horizontal eye rotation away from cathode towards anode. The g-VOR was repeatable across all subjects, its magnitude a linear function of the current intensity, its latency about 9.0ms with GVS of 2.5mA or greater, and was not suppressed by visual fixation. At 10mA stimulation, the g-VOR [x,y,z] on the cathodal side was: [0.77±0.10°, -0.05±0.05°, -0.18±0.06°] (mean±95% confidence intervals) and on the anodal side was: [0.79±0.10°, 0.16±0.05°, -0.19±0.06°], with a vertical divergence of 0.20°. While the horizontal g-VOR could be due to activation of the horizontal semicircular canal afferents, the vertical-torsional g-VOR resembled the vestibuloocular reflex in response to roll-plane head rotation about an earth-horizontal axis and might be due to both vertical semicircular canal and otolith afferent activations. Pulsed GVS is a promising technique to investigate latency and initiation of the human vestibuloocular reflex, as it does not require a large mechanical apparatus nor have problems of head inertia or slippage.