Stereopsis, cyclotorsional "noise" and the apparent vertical

In principle, stereopsis can be used to evaluate the subjective vertical in a sagittal plane, but temporal variation in cyclotorsion should degrade that ability. Video recordings of eye orientation during steady fixation were used to evaluate long-term instability in cyclotorsion. Torsion was measured simultaneously in each eye at 1-sec intervals during about 30 sequential fixations (5-sec duration) on the same target. For each eye separately, the standard deviation of torsion around its mean value averaged about 18 min arc. Some of this variation was conjugate, but the variability in torsional difference between the eyes averaged 17 min arc. Most of this second-to-second variation arose between fixations (average SD = 15 min arc). Such low-frequency, inter-fixational variation in torsional difference between the eyes must produce spurious horizontal disparities in the upper and lower visual fields, and should thereby limit the precision with which the vertical horopter can be evaluated. All subjects exceeded those theoretical limits on precision, however, in performance tests requiring that two vertically separated targets be adjusted to apparent equidistance--but only when permitted to shift fixation back and forth between the upper and lower targets. That latter result provides a challenge to current understanding of stereopsis.