The Evolutionary Dichotomy of Human Visual TiltThe Evolutionary Dichotomy of Human Visual Tilt

The goal of scientific inquiry is to explain what happens, how it happens, and why it happens. The first level corresponds to observation, the second to analysis, and the third to understanding. Without understanding why something happens, we have knowledge without wisdom and cannot formulate a philosophy of science.

In 1866, Javal inferred the existence of human ocular torsion by noting that he was no longer able to see clearly through his cylindrical lenses when he tilted his head.¹ Since that time, innumerable studies have documented the existence of a static ocular counterroll in humans.^{2 - 7} What is remarkable is that, despite the robust dynamic ocular counterroll in humans,² its static counterpart is relatively miniscule (in the range of 5%-10% of the head tilt), leading many to conclude that it no longer serves a compensatory function and some to argue that it is virtually nonexistent.⁸ Jampel has recognized that one of the major functions of the vestibular system is to counteract the effects of gravity on the human body⁸ and that the oblique muscles function to actively constrain torsion under static conditions.⁹ Accordingly, our retinas are oriented gyroscopically to the brain and not the horizon.⁸

In 1989, Leigh et al¹⁰ found that the human dynamic ocular counterroll was maximal with fixation on a stationary earth-fixed visual target (0.72%), but that it decreased (to 0.46%) when viewing a visual display that tilted in line with the head. In this issue of Archives, Chandrakumar et al¹¹ use 3-dimensional video-oculography to measure ocular torsion induced by head tilt while their patients viewed a large-angle grid of horizontal and vertical lines that either remained upright or tilted in lockstep with the head. They document a 40% reduction in the static gain of the ocular counterroll when the visual target moves along with the head. These findings demonstrate that a portion of the human static ocular counterroll is visually driven rather than utricular in origin. More specifically, they show that the human brain uses directional visual information on a moment-to-moment basis to modulate ocular torsion. Although the measured differences in countertorsion were small (1° for a 25° head tilt), testing was performed at a relatively close viewing distance, so the influence of a tilted visual percept may be even greater at distance, where the inhibitory effects of convergence are rendered inoperative.¹²

As with most studies of human ocular torsion, these findings are important more for their evolutionary implications than for their clinical relevance. In lower lateral-eyed animals, utricular input from the 2 labyrinths summates with luminance input from the 2 eyes to determine postural orientation in the roll plane.^{13 - 14} In nature, luminance comes from the sky, so verticality is associated with equal luminance to the 2 eyes.¹⁵ Consequently, unequal luminance to the 2 eyes evokes a compensatory body tilt toward the brighter side (a dorsal light reflex). In lateral-eyed animals, the eyes and ears evolved as complementary balance organs through which visual and vestibular input are yoked together within the central vestibular system to determine postural orientation in the roll plane.^{13 - 15}

At the cortical level, frontal-eyed animals retain their evolutionary programming that light comes from above, as evidenced by the fact that humans perceive circles that are lighter on the upper half as convex spheres and those that are lighter on the lower half as concave spheres (Ramachandran's illusion).¹⁶ However, subcortical luminance reflexes are not completely discarded. Dissociated vertical divergence is a cyclovertical divergence that is seen almost exclusively in humans with infantile strabismus. This human counterpart to the dorsal light reflex is driven by unequal input from the 2 eyes rather than the 2 labyrinths.¹⁵ Although dissociated vertical divergence is not visible in humans with normal binocular vision, van Rijn et al¹⁷ detected small degrees of dissociated vertical divergence using scleral search coil recordings. As with its labyrinthine counterpart (the cyclovertical divergence that is composed of the static ocular counterroll and physiologic skew deviation), the amplitude-dissociated vertical divergence is measurable but negligible under normal circumstances. As with its labyrinthine counterpart (the cyclovertical divergence that is composed of the static ocular counterroll and physiological skew deviation), dissociated vertical divergence is rendered vestigial in normal binocular humans.

Gravity dictates that real-world landmarks are vertical (trees and mountains) and horizontal (lakes and horizons) in orientation, so that natural landmarks become visual anchors for vertical orientation in frontal-eyed animals. Evolution has incorporated these visual cues to preserve accurate multisensorial input to vertical orientation. In 1861, Aubert¹⁸ recognized that a vertical line appeared tilted when he was lying down in an otherwise dark room. Mittelstaedt¹⁹ attributed what became known as the Aubert phenomenon to an idiotropic vector that skewed the perceptual vertical toward the body in the absence of visual cues. The current video-oculography study¹¹ provides corollary evidence that we rely on visual information not only to accurately judge vertical orientation, but to control torsional eye position. A tilted visual contour alone has been previously shown to induce small torsional eye movements.^{20 - 23} The perception of verticality is similarly altered by a tilted visual stimulus.²⁴ As with unequal luminance, a tilted visual percept transiently resets the internal gyroscope that calibrates the baseline torsional eye position.²⁵ Primitive luminance reflexes are therefore supplanted by visual orientational cues in frontal-eyed animals.²⁶

The development of stereopsis has been invoked to explain why static countertorsion diminishes with near fixation. According to this argument, static countertorsion around the visual axes would dislodge retinal correspondence when the eyes assume a convergent position.²⁷ Consequently, the evolution of stereopsis has created a bottleneck wherein the brain permits just enough countertorsion to preserve binocular vision and stereopsis.^{27 - 29} The negligible skew deviation evoked by static head tilt³⁰ may be similarly constrained by the needs of stereoscopic vision.³¹ This control mechanism need not imply that stereopsis (a cortical function) actively inhibits the static torsional counterroll (a subcortical function) on a moment-to-moment basis (in which case, preexisting infantile strabismus or monocular occlusion should amplify countertorsion). Rather, it indicates that the evolution of stereopsis has selected for the nonexpression of the ocular counterroll and other torsional eye movements. Because the observed visual target in the present study was nonstereoscopic, one can conclude that it is visual tilt (rather than a stereoscopic percept) that actively down-modulates the ocular torsion.

If stereopsis were the single driving force for suppression of the static counterroll, one would expect to see robust static countertorsion during distance viewing, where convergence and stereoacuity are minimal. This is not the case. In theory, the brain could execute an eye-head coordinate transformation to cope with a large compensatory static counterroll (as it does during saccadic eye movements by erasing perception of the movement).²³ By constraining static countertorsion in a rapidly moving world, however, the brain can more quickly and accurately deconstruct perceived tilt into visual tilt and body tilt. Maintenance of torsional eye position near 0 preserves a gyroscopic stability that serves to promote balance and to poise the system for rapid visual interpretation of any subsequent roll movement.

The “antivestibular” forces by which these primitive reflexes are suppressed are probably nonvisual in origin. Ocular centration seems to be a fundamental prerequisite for visual orientation.³² If baseline eye centration is not instantaneously restored after eye-head movements, visual computation of postural orientation becomes subservient to transitory eye position. Jampel and Shi³² have postulated that the presence of an eye-centering reflex corrects eccentric positioning of the eyes and recenters them near primary position. This eye-centering reflex can be extrapolated to explain the fast phases of vestibular and optokinetic nystagmus, which momentarily decenter the eyes into the upcoming visual field to restore centration at the end of the slow phase.³³ This reflex is operative in all 3 planes of physical space, so that the eyes lead the head where the body wants to go. Even volitional head tilt is preceded by large conjugate torsional saccades that reposition the eyes in close alignment with the final position of the head.³⁴ These torsional saccades anticipate the effects of the dynamic counterroll and thereby minimize static countertorsion. The dynamic counterroll adapts during sustained head roll and approaches the initial reference position (ie, eye centering).³⁵ While stereopsis is cortical in origin, eye-centering movements are almost certainly subcortical, as they do not rely on visual input for their execution.

One can conclude that the detection of visual tilt is evolutionarily preserved and highly protected and that ocular torsion operates in the service of vertical orientation. In all its elements, our torsional control system has evolved in a dialectical fashion to meet the conflicting needs of positional orientation, binocular alignment, and stereoscopic vision.

Correspondence: Dr Brodsky, Departments of Ophthalmology and Neurology, Mayo Clinic, 200 First St SW, Rochester, MN 55905 (brodsky.michael@mayo.edu).

Financial Disclosure: None reported.

Funding/Support: This work was supported in part by a grant from Reseach to Prevent Blindness.