To investigate the effect of asymmetrical vestibular input on the perceived straight-ahead direction, we compared 7 subjects (age 59 +/- 8 yrs, mean +/- SD) who had chronic (>10 mos) unilateral vestibular deafferentation with 10 age matched controls (age 61+/-6 younger controls (age 28 +/- 7 yrs). Despite the age difference, the two control groups performed similarly and were therefore pooled. Eye and head movements were recorded using search coils as subjects underwent 30 s trials of sinusoidal, whole body oscillation (0.4-2 Hz, peak velocities 0-120 degrees /s) in darkness while attempting to maintain gaze on a remembered target 5 m distant. As a control, most stimulus oscillations were randomly superimposed on an imperceptible, constant velocity of +/-0.5 degrees /s that produced a whole-body offset of 15 degrees by the end of the trial. Following oscillation, subjects remained motionless in darkness and were asked to orient both gaze and a manipulandum to the remembered target location. In control subjects, mean final gaze and manipulandum positions were within 15 degrees of the target for all testing conditions. There was no dependence of final gaze and manipulandum positions on the frequency or velocity of the preceding whole-body oscillations (p > 0.05). In four of seven unilaterally deafferented subjects there was an ipsilesional bias of final eye position of > or =10 degrees. These subjects moved both eye and manipulandum to the ipsilesional side, with the error increasing at higher stimulus velocities. For the 120 degrees /s peak head velocity, mean ipsilesional gaze bias ranged from 10-37 degrees and mean manipulandum bias ranged from 26-108 degrees. Although the errors depended on velocity p < 0.01), errors were independent of frequency (p > 0.1). In the remaining three subjects with vestibular deafferentation, final gaze and manipulandum positions [were not statistically different from controls.] Early gain (eye velocity / head velocity) of the VOR averaged 0.82 +/- 0.01 for the first 10 s of all trials and was similar in all groups (p > 0.1). Gain during the final 10 s gain averaged 0.78 +/- 0.01 for control subjects, but was significantly lower at 0.70 +/- 0.01 for unilaterally deafferented subjects, whose eye positions reached the limit of the ocular motor range. We conclude that many humans with chronic unilateral vestibular deafferentation have a large ipsilesional dynamic bias of eye position and the perceived straight ahead direction reflecting persistent asymmetry of vestibular processing.