The vestibulo-ocular reflex (VOR) was studied in three squirrel monkeys subjected to rotations with the head either centered over, or displaced eccentrically from, the axis of rotation. This was done for several different head orientations relative to gravity in order to determine how canal-mediated angular (aVOR) and otolith-mediated linear (IVOR) components of the VOR are combined to generate eye movement responses in three-dimensional space. The aVOR was stimulated in isolation by rotating the head about the axis of rotation in the upright (UP), right-side down (RD), or nose-up (NU) orientations. Horizontal and vertical aVOR responses were compensatory for head rotation over the frequency range 0.25-4.0 Hz, with mean gains near 0.9. The horizontal aVOR was relatively constant across the frequency range, while vertical aVOR gains increased with increasing stimulation frequency. In the NU orientation, compensatory torsional aVOR responses were of relatively low gain (0.54) compared with horizontal and vertical responses, and gains remained constant over the frequency range. When the head was displaced eccentrically, rotation provided the same angular stimuli but added linear stimulus components, due to the centripetal and tangential accelerations acting on the head. By manipulating the orientation of the head relative to gravity and relative to the axis of rotation, the IVOR response could be combined with, or isolated from, the aVOR response. Eccentric rotation in the UP and RD orientations generated aVOR and IVOR responses which acted in the same head plane. Horizontal aVOR-IVOR interactions were recorded when the head was in the UP orientation and facing toward ("nose-in") or away from ("nose-out") the rotation axis. Similarly, vertical responses were recorded with the head RD and in the nose-out or nose-in positions. For both horizontal and vertical responses, gains were dependent on both the frequency of stimulation and the directions and relative amplitudes of the angular and linear motion components. When subjects were positioned nose-out, the angular and linear stimuli produced synergistic interactions, with the IVOR driving the eyes in the same direction as the aVOR. Gains increased with increasing frequency, consistent with an addition of broad-band aVOR and high-pass IVOR components. When subjects were nose-in, angular and linear stimuli generated eye movements in opposing directions, and gains declined with increasing frequency, consistent with a subtraction of the IVOR from the aVOR. This response pattern was identical for horizontal and vertical eye movements. aVOR and IVOR interactions were also assessed when the two components acted in orthogonal response planes. By rotating the monkeys into the NU orientation, the aVOR acted primarily in the roll plane, generating torsional ocular responses, while the translational (IVOR) component generated horizontal or vertical ocular responses, depending on whether the head was oriented such that linear accelerations acted along the interaural or dorsoventral axes, respectively. Horizontal and vertical IVOR responses were negligible at 0.25 Hz and increased dramatically with increasing frequency. Comparison of the combined responses (UP and RD orientations) with the isolated aVOR (head-centered) and IVOR (NU orientation) responses, indicates that these VOR components sum in a linear fashion during complex head motion.