Horizontal, vertical, and torsional eye movements were recorded using the magnetic search-coil technique during linear accelerations along the interaural (IA) and dorsoventral (DV) head axes. Four squirrel monkeys were translated sinusoidally over a range of frequencies (0.5-4.0 Hz) and amplitudes (0.1-0.7 g peak acceleration). The linear vestibuloocular reflex (LVOR) was recorded in darkness after brief presentations of visual targets at various distances from the subject. With subjects positioned upright or nose-up relative to gravity, IA translations generated conjugate horizontal (IA horizontal) eye movements, whereas DV translations with the head nose-up or right-side down generated conjugate vertical (DV vertical) responses. Both were compensatory for linear head motion and are thus translational LVOR responses. In concert with geometric requirements, both IA-horizontal and DV-vertical response sensitivities (in deg eye rotation/cm head translation) were related linearly to reciprocal fixation distance as measured by vergence (in m-1, or meter-angles, MA). The relationship was characterized by linear regressions, yielding sensitivity slopes (in deg.cm-1.MA-1) and intercepts (sensitivity at 0 vergence). Sensitivity slopes were greatest at 4.0 Hz, but were only slightly more than half the ideal required to maintain fixation. Slopes declined with decreasing frequency, becoming negligible at 0.5 Hz. Small responses were observed when vergence was zero (intercept), although no response is required. Like sensitivity slope, the intercept was largest at 4.0 Hz and declined with decreasing frequency. Phase lead was near zero (compensatory) at 4.0 Hz, but increased as frequency declined. Changes in head orientation, motion axis (IA vs. DV), and acceleration amplitude produced slight and sporadic changes in LVOR parameters. Translational LVOR response characteristics are consistent with high-pass filtering within LVOR pathways. Along with horizontal eye movements, IA translation generated small torsional responses. In contrast to the translational LVORs, IA-torsional responses were not systematically modulated by vergence angle. The IA-torsional LVOR is not compensatory for translation because it cannot maintain image stability. Rather, it likely compensates for the effective head tilt simulated by translation. When analyzed in terms of effective head tilt, torsional responses were greatest at the lowest frequency and declined as frequency increased, consistent with low-pass filtering of otolith input. It is unlikely that IA-torsional responses compensate for actual head tilt, however, because they were similar for both upright and nose-up head orientations. The IA-torsional and -horizontal LVORs seem to respond only to linear acceleration along the IA head axis, and the DV-vertical LVOR to acceleration along the head's DV axis, regardless of gravity.