Even though the vestibulo-ocular reflex (VOR) to head rotation has a long tradition of research, several elementary properties of the VOR are not yet well known. Its role in gaze-stabilisation is well recognised, but properties such as latency and gain in response to brief, transient disturbances of head position have been under-investigated. The very short basic pathway of the VOR (only 3 neurones) suggests a role especially in the compensation of such fast, transient head movements, which cannot be adequately dealt with by visually mediated eye movements. Despite this, and to a large extent due to technical limitations, the emphasis in VOR research has been on sustained rotations or oscillations of the head, especially in human subjects. Recently, there has been a trend to study the VOR with transients and for this purpose we have introduced a head-helmet, driven transiently by reactive forces. This work has revealed a number of interesting and partly unexpected properties. Brief, well-controlled head rotations with an initially approximately constant acceleration (~1000 deg/s²) were imposed on healthy subjects. Binocular eye movements were recorded with scleral coils in a magnetic field and analysed for a period of ~100 ms after the start of the head pulse. Averaging procedures were used to reduce noise. In the first period after the start of the head movement (~10 ms), eye movements were, unexpectedly, typically anti-compensatory, with peak velocities of a few deg/s and zero-latency. This component is interpreted as purely mechanical, and probably due to the eccentric rotation of the orbit around a head-centred axis (with linear forces acting on the orbital contents). Active, compensatory eye movement started after a VOR latency of ~8 ms, according to estimates that took into account the anti-compensatory phase. Statistical analysis showed a highly significantly shorter delay (by ~1 ms) of the contralateral vs. the ipsilateral eye (relative to the direction of head rotation), in agreement with the known shortest synaptic pathways, that involve 1 extra neurone (the interneurone in the abducens nucleus). The moment-to-moment VOR gain (G) was calculated as

Veye (t)              
Vhead (t-latency).

The time course of G was very sensitive to the duration of the latency; our surprisingly short latency estimates yielded the most simple, monotonous build-up of gain after the anti-compensatory phase, without implausible undershoots (for too short latencies) or overshoots (for too long latencies). Effects of the distance of a visual target (a larger VOR gain for a near target, as necessitated by the eccentric rotation of the eye) appeared 20 - 30 ms after the start of the active VOR. Responses in darkness and with a distant target were identical.