Assessing the subjective visual vertical, SVV, in a static upright position is an easy clinical test in which a deviation of some 10 degrees from true vertical indicates an acute loss of unilateral (otolithic) vestibular function on the side to which the SVV is tilted. Because this deviation of the SVV is compensated during the following months, patients with chronic unilateral vestibular loss do no longer differ from normal subjects. This study presents an experimental set-up that allows for clear detection of compensated chronic loss of unilateral otolithic function by testing the SVV. 21 normals and 17 unilaterally vestibular deafferentiated (UVD) patients (vestibular neurectomies) were first rotated on a human centrifuge about an earth vertical yaw axis through the midsagittal plane of the head (240 degrees/s). This induced tilts of the gravito-inertial force (GIF) vectors, which differed at the two inner ears by 8 degrees. During constant velocity rotation, the subjects were moved in pseudo-randomized steps laterally up to 16 cm apart from the rotation axis, inducing roll tilts of the GIF vectors up to 16 degrees. Normal subjects set their SVV to pre-centrifugation values at positions with the midsagittal plane of their head close to the rotation axis, while chronic UVD patients indicated pre-centrifugation values during positions with the rotation axis 5.9 +/- 2.5 cm paramedian on the side of the intact ear. Tilts of the GIF vectors shifted the SVV with a gain of 0.70 in normals and only 0.32 in UVD patients. Roll gains for laterally directed GIF vectors relative to the intact inner ear did not differ from medially directed roll gains in the UVD patients. The roll gains observed in this experimental set-up were lower than those observed with static body tilts or during eccentric rotation with a larger radius, which might be at least partially due to conflicting stimulation between otolithic and extra-vestibular cues.