1. The vestibuloocular reflex (VOR) stabilizes images on the retina against movements of the head in space. Viewing distance, target eccentricity, and location of the axis of rotation may influence VOR responses because rotation of the head about most axes in space rotates and translates the eyes relative to visual targets. To study the VOR response to combined rotation and translation, monkeys were placed on a rate table and rotated briefly in the dark about a vertical axis that was located in front of or behind the eyes. The monkeys fixated a near or far visual target that was extinguished before the rotation. Eye movements were recorded from both eyes by the use of the search coil technique. 2. Peak eye velocity evoked by the VOR was linearly related to vergence angle for any axis of rotation. The percent change in the VOR with near target viewing relative to far target viewing at a vergence angle of 20 degrees was linearly related to the location of the axis of rotation. Axes located behind the eyes produced positive changes in VOR amplitude, and axes located in front of the eyes produced negative changes in VOR amplitude. An axis of rotation located in the coronal plane containing the centers of rotation of the eyes produced no modification of VOR amplitude. For any axis, the VOR compensated for approximately 90% of the translation of the eye relative to near targets. 3. The initial VOR response was not correct in magnitude but was refined by a series of three temporally delayed corrections of increasing complexity. The earliest VOR-evoked eye movement (10-20 ms after rotation onset) was independent of viewing distance and rotational axis location. In the next 100 ms, eye speed appeared to be sequentially modified three times: within 20 ms by viewing distance; within 30 ms by otolith translation; and within 100 ms by eye translation relative to the visual target. 4. These data suggest a formal model of the VOR consisting of four channels. Channel 1 conveys an unmodified head rotation signal with a pure delay of 10 ms. Channel 2 conveys an angular head velocity signal, modified by viewing distance with a pure delay of 20 ms, but invariant with respect to the location of the axis of rotation. Channel 3 conveys a linear head velocity signal, dependent on the location of the axis of rotation, that is modified by viewing distance with a pure delay of 30 ms.(ABSTRACT TRUNCATED AT 400 WORDS)