The linear vestibulo-ocular reflex (LVOR) was studied in eight normal human subjects of average age 24+/-5 years. Subjects underwent a sudden heave (mediolateral) translation delivered by a pneumatic servo-driven chair with a peak acceleration of 0.5 g while viewing earth-fixed targets at 15, 25, 50, and 200 cm. Stimuli were provided both with targets continuously visible or extinguished just prior to motion. Cancellation was tested using chair-fixed targets at each viewing distance. Eye movements were recorded using binocular magnetic search coils. Head translation was measured using a linear accelerometer attached to the upper teeth, to which also was attached a magnetic search coil verifying absence of head rotation. Vergence angles achieved by all subjects were appropriate to interpupillary distance and target distance. Heave translations evoked horizontal ocular rotations in the opposite direction following a brief latency. Latency of the LVOR was determined by automated algorithms based on identification of times when eye position and head acceleration exceeded three standard deviations (SDs) of baseline noise, and was corrected for differing transducer delays. Mean LVOR latency was 30+/-16 ms (mean +/- SD), range 12-53 ms. Slow phase LVOR amplitude was greater for near and less for more distant targets, although all observed responses were suboptimal. Measured 100 ms after head translation onset, mean response was 20% of ideal for the target at 15 cm, 22% at 25 cm, 31% at 50 cm, and 53% at 200 cm. Mean latency was significantly longer than the previously reported values for both the human angular VOR and the monkey LVOR, and had significant inverse correlation with response magnitude. The relatively longer latency of the human LVOR than angular VOR may be tailored to match human head movement dynamics.