INTRODUCTION

1. Cranial Nerves: III, IV, & VI and Motor Nuclei

2. Motor Neuron Response

FUNCTIONAL CLASSIFICATION INTO THREE GENERAL CATEGORIES

B. Dynamic Control of Eye Alignment (Smooth Tracking Responses to Open and Closed-Loop Stimuli):

1. Conjugate Smooth Pursuit Tracking

2. Disconjugate Smooth Vergence Tracking

3. Adaptable Interactions between Smooth Pursuit and Smooth Vergence

C. Neuro-Control of Smooth Foveal Tracking

1. Saccadic Gaze Shifting System

2. Vergence Gaze Shifting System

1. Strabismus

2. Nystagmus

INTRODUCTION

Three Fundamental Visual Sensory-Motor Tasks

The neural control of eye movements is organized to optimize performance of three general perceptual tasks. One task is to resolve the visual field while we move either by translation or rotation through space (self motion). Our body motion causes the image of the visual field to flow across the retina and reflexive eye movements reduce or stabilize this image motion to improve visual performance. The second task is to resolve objects whose position or motion is independent of the background field (object motion). Eye movements improve visual resolution of individual objects by maintaining alignment of the two foveas with both stationary and moving targets over a broad range of directions and distances of gaze. The third task is to explore space and shift attention from one target location to another. Rapid eye movements place corresponding images on the two foveas as we shift gaze between targets lying in different directions and distances of gaze.

Binocular Constraints on Eye Position Control

Binocular alignment of retinal images with corresponding retinal points places additional constraints on the oculomotor system. Because the two eyes view the world from slightly different vantage points, the retinal image locations of points subtended by near objects differ slightly in the two eyes. This disparity can be described with three degrees of freedom (horizontal, vertical and torsional components) that are analogous to the angular rotations of the eye shown in Figure 1. The main task of binocular eye alignment is to minimize horizontal, vertical and cyclodisparities subtended by near targets on the two foveas. This requires a conjugate system that rotates the two eyes in the same direction and amount, and a disconjugate system that rotates the visual axes in opposite directions. As described by Hering (1868), a common gaze direction for the two eyes is achieved by a combination of conjugate and disconjugate movements that are controlled by separate systems. The version system controls conjugate movements and the vergence system controls disconjugate movements. Pure version and vergence movements are described respectively by the isovergence and isoversion contours shown in Figure 2. The isovergence circle describes the locus of points that stimulate the same vergence angle in all directions of gaze (see Ono, 1983 for a review). A different isovergence circle exists at each viewing distance. The isoversion lines describe the locus of points that stimulate the same version angle over a range of viewing distances in a common direction of gaze relative to the head. Pure vergence movements occur along any of the isoversion lines and not just along the central or midsagittal plane. Fixation changes along any other contour result from a combination of version and vergence movements. Both version and vergence movements are described as combinations of horizontal, vertical and torsional rotations. For example there can be horizontal and vertical version and vergence movements. Torsional rotations are usually referred to as cyclo rotations (e.g. cycloversion or cyclovergence). Hering’s law implies that there is equal innervation of yoked muscle pairs. "… one and the same impulse of will directs both eyes simultaneously as one can direct a pair of horses with single reins." The law should not be taken literally, because common gaze commands from higher levels are eventually parceled into separate innervation sources in the brainstem that control individual muscles in the two eyes.

4. Increasing the frequency of spike potentials for a given neuron increases contractile force. Once their threshold is exceeded all cells increase their firing rate as the eye moves further along in the on direction of the muscle until they saturate. Cells increase their firing rate linearly as the eye moves into their on field.

FUNCTIONAL CLASSIFICATION INTO THREE GENERAL CATEGORIES

1. Extra-Retinal Signals:

The vestibular system contains two types of organs that transduce angular and linear acceleration of the head into velocity signals (Figure 5) (see reviews by Simpson and Graf, 1995; Melvill Jones 1989). Three semicircular canals lie on each side of the head in three orthogonal planes that are approximately parallel to a mirror image set of planes on the contralateral side of the head. These canals are stimulated by brief angular rotations of the head and the resulting reflexive ocular rotation is referred to as the vestibulo-ocular reflex (angular VOR). In addition, two otoliths (the utricle and sacculus) transduce linear acceleration caused by head translation as well as head pitch (tilt about the interaural axis) and roll (tilt about the nasal-occipital axis) into translation velocity and head orientation signals (linear VOR). These signals stimulate eye rotations that are approximately equal and opposite to the motion of the head. This stabilization reflex has a short 7-15-msec latency because it is mediated by only three synapses (Lisberger 1984) and is accurate for head turns at velocities in excess of 300 deg/s (Keller 1978). Hair cells in the canals can be stimulated by irrigation of one ear with cold water. This produces a caloric-vestibular nystagmus that causes the eyes to rotate slowly to the side of the irrigated ear (see Cogan, 1956 for a review). These slow-phase movements are interrupted by fast saccadic eye movements that reset eye position in the reverse direction (fast phase). A sequence of slow and fast phases is referred to as jerk nystagmus (Figure 6). Head rotations about horizontal, vertical, and nasal-occipital axes produce VOR responses with horizontal, vertical and torsional counter-rotations of the slow phase of the nystagmus (Seidman and Leigh, 1989).

Fig.  5 Vestibular end-organs in the human temporal bone. Three canals transduce angular head acceleration and two otoliths, the sacculus and utricle, transduce linear acceleration and head orientation. Right labyrinth and chochlea are viewed from horizontal aspect. (Drawings by Ernest W. Beck: courtesy Beltone Electonics Corp., Chicago, Ill.) The canals are in three orthogonal planes that are approximately parallel to a mirror image set of planes on the contralateral side of the head that lie roughly in the pulling direction of the three muscle planes A, Lateral Canals. B, Anterior and posterior vertical canals. LC, lateral canal; AC anterior vertical canal: PC, posterior vertical canal. (From Barber HO, Stickwell CW: Manual of electronystagmography, St Louis, CV Mosby, 1976). Adler’ sfigures 1992 page 148 & 149)



Fig.  6 OKN and VOR are composed of a slow phase (Beta) that rotates the eye in a direction that stabilizes the retinal image and a fast phase (alpha) that resets the eye’s position. The figure illustrates the vestibulo-ocular response to sustained rotation. The slow phase is in the direction opposite to head rotation. Horizontal position is plotted against time. The reflex gradually habituates and has disappeared by about 30 seconds. (Miller, 1982)

To be effective, these reflex eye rotations must stabilize the retinal image. If the axis of angular head rotation coincided with the center of eye rotation, perfect compensation would occur if angular eye velocity equaled angular head velocity. However, the axis of head rotation is the neck and not the center of eye rotation such that when the head rotates, the eyes both rotate and translate with respect to the visual field. This is exacerbated during near viewing conditions. To stabilize the retinal image motion of a nearby target caused by angular head movements, the eye must rotate more than the head. Indeed, the gain of the VOR increases with convergence (Snyder and King, 1992). Mismatches between eye and head velocity also occur with prescription spectacles that magnify or minify retinal image motion. Because the VOR responds directly to vestibular and not visual stimuli, its response is classified as open loop. The VOR compensates for visual errors by adapting its gain in response to retinal image slip to produce a stabilized retinal image (Melvill Jones, 1989). Perfect compensation would occur if angular eye velocity were equal and opposite to angular head velocity while viewing a distant scene. However empirical measures show that at high oscillation frequencies (2Hz) compensation by the VOR is far from perfect and yet the world appears stable and single during rapid head shaking without any perceptual instability or oscillopsis (Collewijn, Steinman, and Erkelens, 1991). Thus, to perceive a stable world, the visual system must be aware of both the amount of head rotation and the inaccuracy of compensatory eye movements so that it can anticipate any residual retinal image motion during the head rotation.

B. Retinal Signals:

Head motion also produces whole-field retinal image motion of the visual field (optic flow) (see Collewijn 1989 for a review). These retinal signals stimulate reflexive compensatory eye rotations that stabilize the retinal image during slow or long-lasting head movements. The eyes follow the moving field with a slow phase that is interrupted by resetting saccades (fast phase) 1 to 3 times per second (Cheng and Outerbridge, 1975). This jerk nystagmus is referred to as optokinetic nystagmus (OKN) and it complements the VOR by responding to low-velocity sustained head movements such as those that occur during walking and posture instability. Like the VOR, OKN also responds with horizontal, vertical and cyclo eye rotations to optic flow about vertical, horizontal and nasal-occipital axes (Cheung and Howard, 1991).

The optokinetic response to large fields has two components, including an early and delayed segment (OKNe and OKNd). OKNe is a short-latency ocular following response (< 50 msec.) that constitutes the rapid component of OKN (Miles Kawano and Optican, 1986), and OKNd builds up slowly after 7 seconds of stimulation (Raphan and Cohen, 1985). OKNe is likely to be mediated by the pursuit pathway (Miles, 1993). The delayed component is revealed by the continuation of OKNd in darkness (optokinetic after nystagmus, OKAN). OKNd results from a velocity memory or storage mechanism (see Collewijn 1985 and Raphan and Cohen, 1985 for reviews). The time constant of the development of the OKAN matches the time constant of decay of the cupula in the semicircular canals (Robinson, 1981). Thus OKAN builds up so that vision can compensate for loss of the vestibular inputs during prolonged angular rotation that might occur in a circular flight path. OKN can be used clinically to evaluate visual acuity objectively by measuring the smallest texture size and separation in a moving field that elicits the reflex.

C. Neuro-Control Of Stabilization Reflexes

1. Vestibulo-Ocular Reflex

The transducer that converts head rotation into a neural code for driving the VOR consists of a set of three semicircular canals paired on each side of the head (See Melvill Jones 1989 for a review). The horizontal canals are paired and the anterior canal on one side is paired with the posterior canal on the contralateral side. (Figure 5) These are opponent pairs so that when one canal is stimulated by a given head rotation its paired member on the contralateral side is inhibited. For example, downward and forward head motion to the left causes increased firing of the vestibular nerve for the left anterior canal and decreased firing of the vestibular nerve projections from the right posterior canal. The three canals lie roughly in the pulling directions of the three muscle planes (Melvill Jones 1989). Thus the left anterior canal and right posterior canal are parallel to the muscle planes of the left eye vertical recti and the right eye obliques. Pathways for the horizontal VOR are illustrated in Figure 7 for a leftward head rotation (Leigh and Zee, 1999). Excitatory innervation projects from the left medial vestibular nucleus to the right abducens nucleus to activate the right lateral rectus, and an interneuron from the right abducens nucleus projects to the left oculomotor nucleus to activate the left medial rectus. The abducens serves as a premotor nucleus to coordinate conjugate horizontal movements to the ipsilateral side in accordance with Hering’s law.



Fig. 7  Pathways of the horizontal VOR in the brain stem for leftward head rotation. Inhibitory connections are shown as filled neurons, excitatory connections as unfilled neurons. Leftward head rotation stimulates the left horizontal canal and inhibits the right horizontal canal. This results in an increased discharge rate in the right lateral and left medial rectus and decreased discharge rate in the left lateral and right medial rectus (From Goldberg, Eggers and Gouras, 1991).

The cerebellar flocculus is essential for adaptation of the VOR to optical distortions such as magnification. The flocculus receives excitatory inputs from retinal image motion (retinal slip) and head velocity information (canal signals) and inhibitory inputs from neural correlates of eye movements that provide a negative feedback signal (Lisberger, 1988). Adaptation only occurs if retinal image motion and head turns occur together (Ito, 1985). The gain of the VOR will be adapted to decrease whenever retinal slip and head turns are in the same direction, and to increase whenever they are in opposite directions. Following adaptation, an error correction signal is projected from the flocculus by Purkinje cells to floccular target neurons (FTN) in the vestibular nucleus to make appropriate changes in VOR gain (Lisberger, 1988).

2. Optokinetic Nystagmus

The visual stimulus for OKN is derived from optic flow of the retinal image (for reviews see Fuchs and Mustari, 1993; Wallman, 1993; Albright, 1993). The retina contains ganglion cells that respond exclusively to motion in certain directions or orientations. This information passes along the optic nerve, decussates at the chiasm and projects to the cortex via the geniculate body (LGN) or to the midbrain via the accessory optic tract (Hoffmann Kistler and Ilg, 1992) (Figure 8). This tract has several nuclei in the pretectal area. One pair of these nuclei, the nucleus of the optic tract (NOT), is tuned to horizontal target motion to the ipsilateral side (i.e. nasal to temporal motion). The lateral and medial terminal nuclei (LTN and MTN) are tuned for vertical target motion (Pasik and Pasik, 1964). Neurons in these nuclei only receive subcortical inputs from the contralateral eye. They have large receptive fields and respond to large textured stimuli moving in specific directions. Stimulation of the right NOT with rightward motion causes following movements of both eyes to the right or ipsilateral side, and similarly stimulation of the left NOT with leftward motion causes leftward conjugate following movements. Each NOT projects signals via the inferior olive to the vestibular nuclei and possibly to the flocculus via the climbing fibers of the cerebellum (Fuchs and Mustari, 1993). The NOT provides a visual signal to the vestibular nucleus, and the motor response is the same as for velocity signals originating from the semicircular canals.



Fig.  8 Simplified schematic illustrating the inputs to the NOT from subcortical crossed retinal projections and cortical-tectal projections. Each NOT gets direct retinal input form the contralateral nasal retina, which is excited by temporal-to-nasal motion. It also receives indirect cortical input (CTX) from the temporal retina of the ipsilateral eye, which tends to be excited more by nasal-to-temporal motion as well as input via the accessory optic system (AAOS), and input from the contralateral visual cortex via the corpus callosum (adapted from Hoffman et al 1992)

The cortical region that organizes motion signals is the medial superior temporal lobe (MST). This region is important for generating motion signals for both pursuit and OKN (Albright, 1993). Binocular cortical cells receive projections from both eyes and code ipsilateral motion from the contralateral visual field at higher velocities than the subcortical system (Wallman, 1993). The cortical cells project to the ipsilateral NOT.

Until the age of 3-4 months, the monocular subcortical projections predominate because the cortical projection has not yet developed (see Schor 1993 for a review). As a result, OKN in infancy is driven mainly by the crossed subcortical input. The consequence is that monocular stimulation only evokes OKN with temporal-to-nasal motion but not with nasal-to-temporal motion. After 3-4 months of development, the infant’s cortical projections predominate and horizontal OKN responds to both monocular temporalward and nasalward image motion. The cortical projections to the NOT fail to develop in infantile esotropia and as adults these patients exhibit the same asymmetric OKN pattern as observed in immature infants (Schor, Wilson, Fusaro, and McKee 1997). This anomalous projection is responsible for a disorder known as latent nystagmus in which a jerk nystagmus occurs when one eye is occluded with the slow phase directed toward the side of the covered eye. During monocular fixation, the stimulated retina increases the activity of neurons in the contralateral NOT via subcortical crossed projections, but it is unable to innervate the ipsilateral NOT via the ineffective cortical-tectal projection. The result is that both eye’s positions are drawn to the side of the stimulated NOT (i.e. the side of the covered eye). The fixation error is corrected with a saccade and a repeated sequence is described as latent or occlusion nystagmus (Dell’Osso, Schmidt and Daroff 1979).

II. FOVEAL GAZE LOCK (MAINTENANCE OF FOVEAL ALIGNMENT WITH STATIONARY AND SLOWLY MOVING TARGETS)

A. Static Control of Eye Alignment (Fixation):

The oculomotor system enhances visual resolution by maintaining alignment of the fovea with attended stationary and moving targets (Westheimer and McKee, 1975) (See Kowler 1989 for a review). During fixation of stationary targets, the eyes sustain foveal alignment over a wide range of target locations in the visual field. Gaze direction is controlled by a combination of eye position in the orbit and head position (Land 1992). Gaze is mainly controlled by eye position for targets lying at eccentricities of less than 15 degrees from primary position (Bahill, Adler and Stark, 1975). Steady fixation at larger gaze eccentricities is accomplished with a combination of head and eye position. Holding eye fixation at large eccentricities (>30 degrees) without head movements is difficult to sustain and the eye drifts intermittently toward primary position in gaze evoked nystagmus (Able, Parker, Daroff and Dell’Osso 1977). This drift is exacerbated by alcohol (Baloh, Sharma, Moskowitz and Griffith, 1979). Even within the 15-degree range, the fixating eye is not completely stationary. It exhibits physiological nystagmus that is composed of slow horizontal, vertical and torsional drifts (0.1 deg/sec), micro-saccades (< 0.25 deg), and a small amplitude (< 0.01 deg) high frequency tremor (40-80Hz) (Carpenter, 1988). Some of the drifts and saccades are error producing, while others are error correcting when they serve to continually minimize motion and adjust the alignment of the target of regard with the fovea (Kowler, 1989; Van Rijn Van der Steen and Collewijn, 1994). The high frequency tremor reflects the noise possibly originating from asynchronous firing of individual motoneurons that is filtered by the mechanical properties of the eye.

Stereoscopic depth perception is enhanced by accurate binocular fixation (see Schor 1983 and 2000 for reviews). Precise bifoveal alignment requires that the eyes maintain accurate convergence at the distance of the attended target. Small constant errors of convergence during attempted binocular fixation (< 15 arc min) are referred to as fixation disparity (see Ogle Martens, Dyer 1967 for a review) and these can impair stereo performance (Badcock and Schor 1985). Fixation disparity is a closed-loop error because it occurs in the presence of retinal image feedback from binocular disparity. Fixation disparity results from incomplete nullification of an open-loop error of convergence known as heterophoria. Heterophoria equals the difference between the convergence stimulus and the convergence response measured under open-loop conditions (e.g. monocular occlusion). The magnitude of fixation disparity increases monotonically with the disparity stimulus to convergence when it is varied with horizontal prisms (Ogle et al 1967). The slope of this function is referred to as the forced-duction fixation disparity curve (Ogle et al, 1967). Shallow slopes of these curves indicate that the heterophoria is reduced by adaptation during binocular or closed-loop conditions (Schor, 1983). Vergence adaptation is very rapid. For example, when convergence is stimulated for only one minute with a convergent disparity and then one eye is occluded, the convergence response persists in the open-loop state (Schor 1983). Typically the convergent disparity is produced by prisms that deflect the perceived direction of both eye’s images inward, or in the nasalward direction, (i.e. the base of the prism before each eye is temporalward or "base-out"). This adapted change in heterophoria is constant in all directions of gaze and is referred to as concomitant adaptation. Prism adaptation is in response to efforts by closed-loop disparity vergence to nullify the open-loop vergence error (heterophoria). Prism adaptation improves the accuracy of binocular alignment and stereo-depth performance.

Horizontal vergence equals the horizontal component of the angle formed by the intersection of the two visual axes and it is described with three different units of measurement or scales that include the degree, prism diopter (PD) and meter angle (MA). Prism diopters equal 100 times the tangent of the angle. Meter angles equal the reciprocal of the viewing distance (specified in meters) at which the visual axes intersect as measured from the center of eye rotation. A target at one meter stimulates one meter angle of convergence (approximately 3.4 degrees) and one diopter of accommodation. Prism diopters can be computed from meter angles by the product of MA and the interpupillary distance (IPD), measured in cm. For example, convergence at a viewing distance of 50 cm by two eyes with a 6 cm IPD equals either 2 MA or 12 PD. The advantage to units in MA is that the magnitude of the stimulus to accommodation in diopters is approximately equal to the magnitude of the stimulus to convergence, assuming that both are measured from a common point such as the center of eye rotation. This assumption produces large errors for viewing distances less than 20 cm because accommodation is usually measured in reference to the corneal apex that is 13 mm anterior to the center of eye rotation. The advantage to units in prism diopters is that they are easily computed from the product of MA and IPD. The advantage to units in degrees is that they accurately quantify asymmetric convergence where targets lie at different distances from the two eyes.

The Maddox classification (Maddox, 1893) describes open- and closed-loop components that make up the horizontal vergence response. The classification includes three open-loop components that influence heterophoria. These include an adaptable intrinsic bias (tonic vergence), a horizontal vergence response to monocular depth cues (proximal vergence), and a horizontal vergence response that is coupled with accommodation by a neurological cross-link known as accommodative-convergence. Tonic vergence has an innate divergence posture (5 deg) known as the anatomical position of rest (Robinson, 1975b). Tonic vergence adapts rapidly to compensate for vergence errors during the first 6 weeks of life. At birth the eyes diverge during sleep, but after 6 weeks the visual axes are nearly parallel during sleep (Rethy, 1969). The adapted vergence, measured in an alert state, in the absence of binocular stimulation and accommodative effort, is referred to as the physiological position of rest. The physiological position of rest equals the sum of tonic vergence and the anatomical position of rest. Tonic vergence adaptation ability continues throughout life to compensate for trauma, disease, and optical distortions from spectacles and aging factors.

Proximal vergence describes the open-loop vergence response to distance percepts stimulated by monocular depth cues such as size, overlap, linear perspective, texture gradients and motion parallax (McLin, Schor and Kruger, 1988; Wick and Bedell, 1989). This proximal response accounts for a large portion of the convergence response to changing distance (Judge, 1991; Schor et al 1992). The open-loop convergence response is also increased by efforts of accommodation (Muller 1826; Alpern and Ellen 1956). The eyes maintain approximately 2/3 of a meter angle of convergence (4 prism diopters) for every diopter of accommodation (Ogle et al 1967). A target at one meter stimulates one meter angle of convergence (approximately 3.4 degrees) and one diopter of accommodation. Thus when the eyes accommodate 1 diopter the accommodative-convergence response increases by approximately 2.3 degrees or 4 prism diopters which is only 68 percent of the convergence stimulus. The sum of the three open-loop components of convergence typically lag behind the convergence stimulus and produce an open-loop divergence error (exophoria) that is usually less than two degrees under far viewing conditions and 4 degrees at near viewing distances (Borish, 1970). Esophoria describes open-loop vergence errors caused by excessive convergence that lead the stimulus. The distribution of heterophoria in the general population is not gaussian or normally distributed. It is narrowly peaked about a mean close to zero (Tait, 1933), indicating that binocular errors of eye alignment are minimized by an adaptive calibration process. During closed-loop stimulation, vergence error is reduced to less than 1/10 degree by the component of the Maddox classification controlled by visual-feedback, disparity vergence, which is stimulated by retinal image disparity. Variation in any of the three open-loop components of convergence will influence the magnitude of heterophoria and the resulting closed-loop fixation disparity. Fixation disparity acts as a stimulus to maintain activity of the disparity vergence so that it continues to sustain its nulling response of the underlying heterophoria during attempted steady binocular fixation (Schor, 1983). Other dimensions of vergence also adapt to disparity stimuli. Vertical vergence adapts to vertical prism before one eye (Schor, Gleason, Maxwell, and Lunn 1993) and cyclovergence adapts to optically produced cyclodisparity stimuli (Maxwell and Schor, 1999; Taylor, Roberts and Zee, 2000; Schor Maxwell and Graf, 2001)

 

B. Dynamic Control of Eye Alignment (Smooth Tracking Responses To Open- and Closed-Loop Stimuli):

1. Conjugate Smooth Pursuit Tracking:

Smooth following pursuit movements allow the eyes to maintain foveal alignment with a moving target that is voluntarily selected (Kowler et al 1984;see Pola and Wyatt, 1991 for a review). Pursuit is defined as the conjugate component of smooth following eye movement responses to target motion. As described below, disconjugate following responses to changes in target distance are referred to as smooth vergence. Conjugate pursuit differs from the delay portion of OKN, which is a reflex response to optic flow of the entire visual field. A conflict occurs between the pursuit and optokinetic systems when the eyes pursue an object that is moving across a stationary background. Pursuit stabilizes the moving target on or near the fovea but it causes optic flow or retinal slip of the stationary background scene. Conflicts can also occur between pursuit and the VOR if gaze is controlled with head tracking movements. Stabilizing the retinal image of a moving object with head movements produces a vestibular signal from the canals. Consequently, pursuit of a small moving target against a stationary background with eye or head movements requires that OKN and the VOR be ignored or suppressed. This is accomplished most effectively when the background lies at a different distance than the pursuit target (Howard and Marton, 1997).

Pursuit responds to target velocities ranging from several minutes of arc/sec to over 175 deg/sec (Pola and Wyatt, 1989). The gain or accuracy of pursuit is reduced as target velocity increases above 100 deg/sec (Myer, Lasker and Robinson, 1985). The VOR has a much higher velocity range than pursuit. This can be demonstrated by comparing two views of your index finger. Either keep your finger stationary while you shake your head rapidly from side to side at 2-3 Hz, or shake your finger at the same frequency without moving your head. The eye cannot follow the moving finger, but it can follow the stationary finger while shaking your head even though the head-relative motion is identical in these two examples. The pursuit response is more accurate when the target motion is predictable such as with pendular motion (Stark et al. 1962). Pursuit errors are reduced by modifying pursuit velocity and with small catch-up saccades. The combination of pursuit and catch-up saccades that appear at low stimulus velocities in patients with pursuit deficits is referred to as cogwheel pursuit. The ratio of eye velocity over target velocity (gain) or accuracy of pursuit is normally affected by target visibility (contrast) as well as drugs and fatigue (O'Mullane and Knox, 1999; Rashbass 1961).

The pursuit response to sudden changes in target velocity has a short latency (80-130 msec) (Lisberger and Westbrook, 1985) and is composed of two general phases referred to as open-loop and closed-loop (Lisberger et al 1985) (Figure 9). Pursuit is initiated during the open-loop phase and it is maintained during the closed-loop phase. The open-loop response is divided into an early and late component. The early component is a feed forward phase that lasts for only 20 msec. During this early phase, there is a rapid acceleration of the eye (40- 100 deg/sec/sec) that is in the correct direction but is independent of the stimulus velocity and initial retinal image position (Lisberger and Westbrook 1985). During the late open-loop component that lasts 80 msec, the initiation of pursuit depends strongly on target velocity and retinal image position (Carl and Gellman, 1987). Eye acceleration is highest in response to targets imaged near the fovea and decreases sharply with increasing eccentricity up to 21 degrees. These open-loop components are calibrated by adaptation (Carl and Gellman, 1986).



Fig.  9  Eye velocity during the onset of pursuit to a 15 deg/sec ramp target motion, the ramp of motion beginning at different eccentricities as indicated at the right portion of each trace. The velocity of the early component (indicated by arrows) was the same for all starting positions, but the velocity of the late component varied (from Lisberger and Westbrook, (1985)

Pursuit is maintained during the closed-loop phase, in response to negative feedback from retinal image velocity (retinal slip) and position as well as an internal estimate of target velocity relative to the head. If the eye lags behind the stimulus, the retinal image velocity is not nulled leaving a residual retinal image slip and position error away from the fovea. The pursuit system accelerates to correct both retinal position and velocity errors (Rashbass 1961). When pursuit is very accurate and there is no retinal error, the eye continues to pursue the target without the eye-referenced signals. Pursuit is maintained by an internal estimate of target velocity or a head referenced motion signal that is computed from a combination of retinal slip and an internal representation of eye velocity (Pola and Wyatt, 1989). This can be demonstrated by attempting to fixate an ocular floater or a retinal afterimage that is located near the fovea. Attempts to foveate the stabilized retinal image lead to smooth following eye movements even though the retinal image is always motionless. The eye is tracking an internal correlate of its own motion that causes the target to appear to move with respect to the head.

2. Disconjugate Smooth Vergence Tracking:

Foveal alignment of targets that move slowly in depth is maintained by smooth vergence following eye movements (for reviews see Judge, 1991; Collewijn and Erkelens 1990). In addition to improving stereoscopic depth perception, smooth vergence provides information about changing target distance that affects size and depth perception (Hollins and Bunn, 1977; Foley 1978; Bradshaw, Glennerster and Rogers, 1996). Slow changes in smooth vergence respond to body sway and posture instability. While smooth vergence responses can be very inaccurate during natural rapid head movements, when the head is stationary they are very accurate at temporal frequencies up to 1 Hz (Rashbass and Westheimer, 1961). At higher frequencies accuracy is reduced but it can improve with small disparity stimuli (Collewijn and Erkelens, 1990). It is likely that accuracy of smooth vergence is task dependent. Smooth vergence accuracy is more demanding for spatial localization tasks that lack depth cues other than disparity, compared to tasks that have ample monocular cues for direction and distance. The accuracy of smooth vergence responses to changing disparity improves with predictable target motion (Rashbass and Westheimer, 1961). Stimuli for smooth vergence tracking include magnitude and velocity of retinal image disparity (Rashbass and Westheimer, 1961) and perceptual cues to motion in depth, including size looming (Collewijn and Erkelens 1990, McLin et. al, 1988). Smooth vergence tracking is very susceptible to fatigue and central nervous system suppressants (Rashbass and Westheimer, 1961).

3. Adaptable Interactions between Smooth Pursuit and Smooth Vergence:

In natural scenes, motion of a target in the fronto-parallel plane is tracked binocularly with conjugate smooth pursuit. However in conditions of anisometropia corrected with spectacle lenses, the image motion of the two eyes is magnified unequally and this produces variations of binocular disparity that increase with target eccentricity from the optical centers of the lenses. Tracking motion of targets in the fronto-parallel plane then requires both smooth pursuit and smooth vergence eye movements such that one eye moves more than the other does. The oculomotor system can adapt to the binocular disparity that changes predictably with eye position. Adaptation produces open-loop non-conjugate variations of heterophoria that compensate for the horizontal and vertical disparities produced by the unequal magnifiers during smooth vergence tracking responses (Schor Gleason and Horner 1990). After only one hour of pursuit tracking experience with anisometropic spectacles, one eye can be occluded and the two eyes continue to move unequally during monocular tracking (Schor et al 1990). The adapted heterophoria is coupled to vary with both eye position and direction of eye movement (Gleason Schor Lunn and Maxwell 1993).

C. Neuro-Control of Smooth Foveal Tracking

1. Smooth Pursuit Tracking System

Smooth following eye movements result from cortical motion signals in extrastriate cortex in areas MT and MST that lie in the superior temporal sulcus (for reviews see Keller, 1991; Albright, 1989). Area MT encodes speed and direction of visual stimuli in three dimensions relative to the eye. MT receives inputs from the primary visual cortex and projects visual inputs to area MST and the frontal eye fields (FEF). Cells in MST fire in concert with head-centric target movement; i.e. they combine retinal and efference copy signals (Newsome, Wurtz and Komatsu, 1988). Each hemisphere of the MST codes motion to the ipsilateral side. Cells have two types of visual motion sensitivity; they respond to motion of large-field patterns and small spots but the direction preferences for the two stimulus types are in opposite directions. The anti-directional large-field responses could facilitate pursuit of small targets moving against a far stationary field, and motion parallax stimuli. Efference from MST and FEF projects ipsilaterally to the NOT to generate OKN and to the dorsal lateral pontine premotor nuclei (DLPN) for pursuit tracking (Fuchs and Mustari 1993). Velocity signals are projected from DLPN to the floccular region and to the vermis lobules VI and VII of the cerebellum (Fuchs and Mustari 1993). The flocculus is thought to maintain pursuit eye movements during steady constant tracking while the vermis is important when the target velocity changes or when initiating pursuit. The role of the cerebellum is to sort out eye and head rotations in the tracking process and to sort out the ocular pursuit signal from visual and eye-head motor inputs (Noda and Warabi 1987). From here, activity passes via parts of the vestibular nuclei, which perform the necessary neural integration of the velocity signal to a position signal that is sent to the eye muscle motor neurons.

2. Smooth Vergence Tracking System.

Vergence results from the combined activity of intrinsic tonic activity, accommodative vergence, and responses to binocular disparity and perceived distance (see Leigh and Zee, 1999 for a review). The sensory afferent signals for vergence (binocular disparity and blur) are coded in the primary visual cortex (area V1) (Poggio, 1995; Busettini ; Miles, 1997). Some cells in V1 incorporate vergence to code egocentric (head-referenced) distance (Trotter, Clelebrini, Stricanne, Thorpe, and Imbert 1996). Cells in area MT and MST respond to retinal disparity and changing size (DeAngelis and Newsome.1999; Maunsell and Van Essen 1983; Roy, Komatsu and Wurtz, 1992). Cells in the parietal cortex respond to motion in depth (Colby, Duhamel and Goldberg, 1993). Efferent commands for vergence appear in cells in the frontal eye fields (Gamlin and Yoon 2000). In the midbrain the premotor nucleus reticularis tegmenti pontis (NRTP), located just ventral to the rostral portion of the paramedian pontine reticular formation (PPRF), receives projections from the frontal eye fields and the superior colliculus (Gamlin and Clark 1995). The NRTP projects to the cerebellum, and appears to be associated with vergence and accommodation (Ohtsuka, Maekawa and Sawa, 1993). The posterior interposed nucleus (PIN) of the cerebellum projects to supraoculomotor regions that contain near response cells in the mesencephalic reticular formation (MRF) (May, Porter and Gamlin 1992; Zhang and Gamlin 1998). The supraoculomotor nucleus contains both burst and tonic neurons (Mays and Porter, 1984). The burst cells code velocity signals for smooth vergence, and the tonic neurons code position signals to maintain static vergence angle. Excitatory connections of the supraoculomotor nucleus project to the oculomotor nucleus, driving the medial recti (Judge and Cumming, 1986; Zang, Mays and Gamlin, 1992). Inhibitory connections project to the abducens nucleus to inhibit the lateral rectus (Mays and Porter 1984). The supraoculomotor nucleus relays control of both accommodative and disparity vergence (Judge and Cumming, 1986).

III.FOVEAL GAZE SHIFTS: TARGET SELECTION AND FOVEAL ACQUISITION.

A. Rapid Conjugate Shifts of Gaze Direction (Saccadic Eye Movements)

Saccades are very fast, yoked eye movements that have a variety of functions (for reviews see Becker 1989 and Van Gisbergen and Van Opstal, 1989). They produce the quick phase of the VOR and OKN to avoid turning the eyes to their mechanical limits. They reflexively shift gaze in response to novel stimuli that appear unexpectedly away from the point of fixation. Saccades shift gaze during reading from one group of words to another. Saccades search novel scenes to assist us in acquiring information. They also return gaze to remembered spatial locations. Two primary functions in all of these tasks are to move the eye rapidly from one position to another and then maintain the new eye position. The rapid movement is controlled by a pulse and slide innervation pattern and the position is maintained by a step innervation.

The separate components of innervation for the saccade match the characteristics of the plant (i.e. the globe, muscles, fat and suspensory tissues). The rapid changes in orbital position are made by the saccade at a cost of considerable energy. Saccade velocities can approach 1000 deg/sec (Becker 1989). In order to achieve these high velocities a phasic level of torque is needed to overcome the viscosity of orbital tissues, most of which is in the muscles (Robinson, 1975). The phasic-torque is generated by a large brief force resulting from a pulse or burst of innervation. The torque is dissipated or absorbed by the muscles, so that the force developed by the pulse of innervation does not reach the tendon (see chapter on Three-Dimensional Rotations of the Eye). At the end of the saccade, a lower constant force resulting from a step innervation generates a tonic level of torque that is needed to hold the eye still against the elastic restoring forces of the orbital tissues (van Gisbergen and Van Opstal, 1989). The eye positions resulting from the pulse and step forces must be equal to produce rapid gaze shifts. Pulse-step mismatches will result in rapid and slow components of gaze shifts. For example, if the pulse is too small, the saccade will slide (post-saccadic drift, called a glissade) to the new eye position at the end of the rapid phase of the saccade. The slide component is adaptable as has been shown by long-term exposure to artificially imposed retinal image slip immediately after each saccade (Optican and Miles, 1985). The adapted post-saccadic drift cannot be explained by an adjustment of the pulse-step ratio, suggesting that the slide innervation is an independent third component of saccadic control (pulse-slide-step). Slide innervation produces a phasic-torque in addition to the pulse component that adjusts the duration and velocity of the saccade so that its amplitude matches the position maintained by the step component (Optican and Miles, 1985). The pulse, slide and step are all under independent, cerebellar control, with the primary goal of protecting vision by preventing retinal slip, and a secondary goal of making accurate saccades.

The amplitude of the saccade determines its dynamic properties (e.g., its peak velocity and duration). The main sequence diagram plots these two dynamic parameters as a function of amplitude (Bahill, Clark and Stark 1975) (Figure 10). As saccade amplitude increases from 0.1 to 10 degrees, duration increases from 20 to 40 msec and peak velocity increases from 10 to 400 deg/sec. Peak velocity saturates for saccades larger than 20 degrees, such that the amplitude of larger saccades increases primarily with duration. Abnormal saccade amplitudes (dysmetria) can either be too small (hypometric) or too large (hypermetric). Large gaze shifts are normally accomplished with a sequence of hypometric saccades that are composed of a series of short-latency corrective saccades in the same direction (Becker 1989). The normal latency of a saccade to an unpredictable stimulus is 180-200 msec (Carpenter, 1988). However, corrective saccades occur with shorter latencies (100-150 msec). Saccade latency can be reduced by a blank or gap interval before the saccade, resulting in an express saccade with latencies less than 100 msec (Fischer and Ramsperger 1984). Saccade latencies to predictable target changes, such as occur in a tennis match, can be reduced to zero (Stark, Vossius and Young, 1962).



Fig.  10  Main sequence diagram. Peak velocity and saccade duration are plotted against magnitude of human saccadic eye movements. (from Bahill AT, Clark MR and Stark, L: Math Biosci 24:191, copyright 1975 by Elsevier Science Publishing Co, Inc.) Adler’s figure 1992 p 158

Although saccades are too brief to utilize visual feedback during their response, they do use fast internal feedback based upon an internal representation of eye position (efference copy signal) that helps control the position of the eye on a moment-to-moment basis (Robinson, 1975). Thus saccadic eye movements are not ballistic in that they are guided by extra-retinal information during their flight. The goal of the saccade is to reach a specified direction in head-centric space. Normally the perceived head-centric direction of a target does not change when the eye changes position. However spectacle refractive corrections that magnify or minify the retinal image produce changes in perceived direction with eye position. Because the entrance pupil of the eye translates when the eye rotates, prior to the saccade the eye views a non-foveal eccentric target through a different part of the lens than after the saccade when the target is viewed directly along the line of sight. The prismatic power of the lens increases with distance from the optical center of the lens such that when viewing a target through a magnifier, the saccade amplitude needed to fixate an eccentric target is larger than the gaze eccentricity sensed prior to the saccade. Saccades are controlled by a feedforward system that does not utilize visual feedback during the motor response. Consequently, initial saccadic responses to visual distortion produced by the magnifier are hypometric. However, using position errors after the saccade, the system adapts rapidly (within 70 trials) to minimize its errors (Deuble, 1995). In cases of anisometropia in which the retinal images are magnified unequally by the spectacle refractive correction, the saccadic system adapts to produce unequal or disconjugate saccades that align both visual axes with common fixation targets (Erkelens, Collewijn and Steinman, 1989; Oohira, Zee and Guyton 1991; Lemij and Collewijn, 1992). The same adaptive process is likely to calibrate the conjugate saccades and maintain that calibration throughout life in spite of developmental growth factors and injury.

2. Disconjugate Shifts of Gaze Distance (The Near Response in Symmetrical Convergence)

Large abrupt shifts in viewing distance stimulate adjustments of several motor systems including accommodation, convergence and pupil constriction (see Judge, 1989 for a review). Separate control systems initiate and complete these responses (Semmlow, Hung, Horng and Ciuffreda, 1994). Initially the abrupt adjustments are controlled by feedforward systems that do not use visual feedback until the responses are nearly completed. However they are guided by fast feedback from efference copy signals to monitor both the starting point of the near response and the accuracy of its end point (Zee and Levi, 1989; Schor, Alexander, Cormack and Stevenson, 1992). Visual feedback is unavailable during the response because the blur and disparity cues are too large at the beginning of the gaze shift to be sensed accurately. The motor responses are initiated by high-level cues for perceived distance and by voluntary shifts of attention. Retinal cues from blur and disparity are only used as feedback to refine the responses once the stimuli are reduced to amplitudes that lie within the range of visual sensitivity, i.e., as the eyes approach alignment at their new destination. The three motor systems are synchronized or coordinated with one another by cross-links. When they approach their new target destination, accommodation and convergence use visual feedback to refine their response.

The cross-couplings are demonstrated by opening the feedback loop for one motor system while stimulating a coupled motor system that is under closed-loop control. For example, accommodative convergence, measured during monocular occlusion, increases linearly with changes in accommodation stimulated by blur (Muller, 1826 Alpern and Ellen 1956). Similarly, convergence accommodation, measured during binocular viewing through pinhole pupils, increases linearly with changes in convergence stimulated by disparity (Fincham and Walton, 1957). The pupil constricts with changes in either accommodation or convergence to improve the clarity of near objects (Wilhelm; Schaeffel; Wilhelm 1993). These interactions are greater during the dynamic changes in the near response than during the static endpoint of the response (Schor and Kotulak, 1986).

The cross-couplings between accommodation and convergence are illustrated with a heuristic model (Figure 11) (Schor and Kotulak 1986). Three of the Maddox components are represented in the model by an adaptable slow tonic component, the cross-links between accommodation and convergence, disparity driven vergence and blur driven accommodation (fast phasic components). The enhanced gain of the cross-link interactions associated with dynamic stimulation of vergence and accommodation results from the stimulation of the cross-links by the phasic but not the tonic components. When accommodation or convergence is stimulated, the faster transient-phasic system responds first, but it does not sustain its response. The slower and more sustained adaptable tonic system gradually takes over the load of keeping the eyes aligned and focused by resetting the level of tonic activity. Because the cross-links are mainly stimulated by the phasic component, accommodative vergence and vergence accommodation are stimulated more during the dynamic response than during steady fixation when the tonic components control eye alignment and focus.



Fig.  11  Model of cross-link interactions between vergence and accommodation. A fast phasic system drives the cross-links from accommodation to convergence (AC/A) and from convergence to accommodation (CA/C). The slow-tonic system adapts to the faster phasic system and gradually replaces it. Cross-link innervation is reduced when the tonic system reduces the load on the fast phasic system. (From Kotulak and Schor, 1986).

Traditionally, only the horizontal component of vergence has been considered as part of the near response. However, vertical and cyclovergence must also be adjusted during the near response to optimize the sensory stimulus for binocular vision (Allen and Carter, 1967; Tweed, 1997; Schor, Maxwell and Stevenson, 1994; Van Rijn and Van den Berg 1993; Yegge and Zee, 1995). The primary goal of the near response is to minimize large changes in horizontal, vertical and cyclodisparities at the fovea that normally accompany large shifts in viewing distance. Horizontal disparities arise from targets that are nearer or farther than the convergence distance, assuming that the horopter equals the Vieth-Muller circle. The isovergence circle describes the locus of points that stimulates a constant vergence angle in all directions of gaze (see Ono, 1983 for a review) and it is equivalent to the Veith-Muller circle that was described in the chapter on Binocular Vision (Figure 2). Convergence and divergence stimuli lie closer or farther, respectively, from the isovergence circle. Vertical disparities arise from targets in tertiary gaze directions at finite viewing distances, because these targets lie closer to one eye than the other and their retinal images have unequal size and vertical eccentricity (vertical disparity) (Figure 12). Torsional disparities arise from elevated targets at finite viewing distances, because during convergence, Listing’s law predicts that the horizontal meridians of the two eyes will be extorted in upward gaze and intorted in downward gaze (Tweed, 1997). These torsional eye postures would produce incyclodisparity in upward gaze and excyclodisparity in downward gaze. With large shifts of viewing distance, horizontal, vertical and cyclodispaities can exceed the stimulus operating range for continuous feedback control of disparity vergence. Initially the near response is stimulated by perceived distance and voluntary changes in horizontal vergence respond without disparity feedback in feedforward control. Unlike horizontal vergence, neither vertical vergence nor cyclovergence is normally under voluntary control. They participate in the near response through cross-couplings with motor responses that are under voluntary control. This allows potential vertical and torsional eye alignment errors to be reduced during the near response without feedback from retinal image disparity.



Fig. 12  For convergence at a finite viewing distance, points in tertiary directions subtend unequal vertical visual angles at the two eyes, which produce vertical disparities.

Empirical measures demonstrate that the gains for the three coupling relationships for horizontal, vertical and cyclovergence are optimal for reducing disparity at the fovea to zero during the open-loop phase of the near response (Schor and Kotulak, 1986; Schor, Maxwell Stevenson 1994; Yegge and Zee, 1995; Somani et al., 1998). The tuned coupling gains for all three vergence components of the near response are the product of neural plasticity that adapts each of these cross-couplings to optimize binocular sensory functions (Miles, Judge & Optican 1987; McCandless and Schor, 1997; Schor Maxwell and Graf, 2000). Plasticity also exists for other couplings. For example, vertical vergence and cyclovergence can both be adapted to vary with head roll (Maxwell and Schor, 1997; 1999). The coupling of pupil constriction with accommodation and convergence may also be under adaptive control. The pupil constriction component of the near response does not appear until the end of the second decade of life (Wilhelm, Schaeffel, and Wilhelm 1993) suggesting that it responds to accommodative errors resulting from the aging loss of accommodative amplitude. The pupil constriction component of the near response is an attempt to restore clarity of the near retinal image.

3. Interactions Between Conjugate and Disconjugate Eye Movements (Asymmetric Vergence).

The frontal eye fields (FEF) mediate voluntary control of contralateral saccades. The FEF is active whether saccades occur or not (see Goldberg and Segraves, 1989 for a review). The activity is related to visual attention, and when saccades occur the related activity in the FEF precedes them by 50 msec. The surface of the FEF has a coarse retinotopic organization. Stimulation of a particular area causes a saccade to change eye position in a specific direction and amplitude. These cells are active before saccades to certain regions of visual space. These regions are called the movement field of the cell, and they are analogous to the receptive fields of sensory neurons in the visual cortex. Stimulation of FEF cells in one hemisphere causes conjugate saccades to the contralateral side. Vertical saccades require stimulation of both hemispheres of the FEF. Modalities that can stimulate movement fields include vision, audition, and touch. The FEF project two main efferent pathways for the control of saccades. One projection is to the superior colliculus (SC). The other projection is to the midbrain, to the paramedian reticular formation (PPRF) and the rostral interstitial medial longitudinal fasciculus (riMLF) for the control of horizontal and vertical saccades, respectively (for reviews see Hepp, Henn, Vilis and Cohen 1989; Sparks and Hartwich-Young, 1989). The fibers from the frontal eye fields descend to the ipsilateral superior colliculus and cross the midline to the contralateral PPRF. Neither the superior colliculus nor FEF are required exclusively to generate saccades. Either one of them can be ablated without abolishing saccades, however if both are ablated, saccades are no longer possible. The function of the SC is to represent intended gaze direction resulting from combinations of head and eye position. Stimulation of a specific region in the intermediate layers of the colliculus can result in several combinations of head and eye position that achieve the same gaze direction relative to the body (Freedman and Sparks, 1997). Cells in the superior colliculus respond to all sensory modalities including vision, audition, and touch. The spatial locations of all of these sensory stimuli are mapped in the colliculus relative to the fovea. Like the FEF, stimulation of one SC causes a conjugate saccade to the contralateral side; stimulation of both sides is necessary to evoke purely vertical saccades. 

The output of the SC and FEF project to two premotor nuclei, the PPRF and the riMLF, that shape the velocity and amplitude of horizontal and vertical components of saccades, respectively (Hepp et al 1989 Keller, 1991). The PPRF projects to the ipsilateral abducens nucleus, which contains motoneurons that innervate the ipsilateral lateral rectus and interneurons that project to the contralateral oculomotor nucleus to innervate the medial rectus. The PPRF also projects inhibitory connections to the contralateral PPRF and vestibular nucleus to reduce innervation of the antagonist during a saccade. The riMLF projects to the ipsilateral trochlear nucleus (IV) and to both oculomotor nuclei (III). Four types of neurons including burst cells, tonic cells, burst-tonic cells and pause cells control saccades in several premotor sites. The pulse component of the saccade is controlled by medium-lead burst neurons. Long-lead burst neurons discharge up to 200 msec before the saccade and receive input from the SC and FEF. They drive medium-lead burst neurons (MLB) that begin discharging at a high frequency (300-400 spikes/sec) immediately at the beginning of the saccade and throughout its duration. Duration of MLB activity ranges from 10 to 80 msec. They project to the motor nuclei and control pulse duration and firing frequency, which determine saccade duration and velocity. Inhibitory burst neurons inhibit antagonist muscles by suppressing neurons in the contralateral abducens nucleus.

Saccades enhance the velocity of both vergence and accommodation. Current models of saccade-vergence interactions suggest that OPNs in the midbrain gate the activity of both saccade bursters and vergence bursters (Zee et al. 1992). Accelerated vergence caused by saccadic facilitation results from a release from inhibition of vergence bursters by reduced firing of the OPN's, which share their inhibition with saccadic and near response bursters. This facilitation is correlated with an augmentation of firing rate of a sub-set of convergence burst neurons of the near response cells (Mays and Gamlin 1995). This model has been developed further to include the potential for saccadic facilitation of accommodation (Mays and Gamlin (1995 a, b). Because near response cells provide innervation for both accommodation and vergence, release of inhibition from OPNs augments activity of both accommodation and vergence when associated with saccades (Schor, et al 1999). This augmentation would not only enhance velocity of the near response, but it would also gate or synchronize innervation for vergence and accommodation with saccades.

NEUROLOGICAL DISORDERS OF THE OCULOMOTOR SYSTEM

Neurological disorders have greater variability across individuals than is found for the normal oculomotor system. The sources of variability come from the location and extent of lesions, associated anomalies or syndromes, different sources or causes of the problem, history of prior treatment interventions, and adaptive responses that attempt to compensate for the disorder. A multi-perspective classification can reflect this individual variation. It views the anomaly in different ways and highlights very different types of information about the condition. Its versatility describes the range of anomalous behaviors that may develop as a function of various combinations of causal factors. The pattern of associated conditions or syndromes facilitates estimating the prognosis for cosmetic and functional correction. Oculomotor disorders are classified in terms of behavioral descriptions, etiology, and neuro-anatomical sites of involvement. Behavioral categories can be descriptive such as the magnitude and direction of a strabismus (an eye turn), gaze restrictions, saccadic disorders and nystagmus. They can also be described by an associated cluster of anomalies that collectively characterize a disease (syndrome). Categories of etiology include congenital, developmental and acquired. Congenital disorders appear in early infancy (< 18 months). Developmental anomalies can result from interference of normal sensory-motor interactions during the first 6 years of life that constitute the critical period for visual development. Both congenital and developmental disorders are associated with anomalies throughout the oculomotor pathways, including sensory or afferent components of the visual system and they are classified in terms of syndromes. Acquired disorders can result from trauma or disease and they are classified in terms of specific anatomical sites of involvement and by syndromes. (see Leigh and Zee 1999 for a comprehensive review).

IV. NYSTAGMUS.

Nystagmus refers to a regular pattern of to and fro movements of the eyes, usually with alternating slow and fast phases (Figure 6). The direction of the nystagmus is specified according to the direction of the fast phase (e.g. "jerk right"), because the fast phase is more visible than the slow phase. When the velocity of the eye oscillations is equal in both directions the waveform is classified as pendular.

Nystagmus can also have vertical and torsional components. For example, see-saw nystagmus is an acquired pendular from of nystagmus in which there is a combination of vertical and torsional oscillation of eye orientation. The eyes bob up and down by several degrees in opposite directions (skew movements). As one eye elevates and intorts, the other eye depresses and extorts. The exact cause is unknown, but see-saw nystagmus is associated with visual loss in bitemporal hemianopia and optic chiasm disorders. The nystagmus is thought to result from an inappropriate ocular tilt response (ocular counter-roll during head tilt) orchestrated by the cerebellum in association with the otoliths.

All of the oculomotor anomalies described above show some evidence of attempts by the oculomotor system to correct them using the same adaptive mechanisms that normally calibrate the various classes of eye movement systems. However the anomalies are beyond the corrective range of the adaptive processes. If it were not for these adaptive processes, these anomalies would be far more prevalent and the oculomotor system would be extremely susceptible to permanent injury resulting from disease and trauma. The consequences would be very dramatic. For example, about 50 years ago a physician named John Crawford lost function of his vestibular apparatus as a result of an overdose of streptomycin used to treat tuberculosis located in his knee. He reported that every movement of his head caused vertigo and nausea, even when his eyes were open. If his eyes were shut, the symptoms intensified. He attempted to steady his head by lying on his back and gripping the bars at the head of the bed. However even in this position the pulse beat in his head became a perceptible motion, disturbing his equilibrium. It is difficult to imagine how we would survive without the adaptive processes that continually calibrate our oculomotor system and allow us to distinguish between motion of our head and eyes from motion of objects in the world.

Badcock DR, Schor CM: Depth-increment detection function for individual spatial channels. J Opt Soc Am, 2A, 1985; 1211-1215

Bahill AT, Clark MR, Stark L : The main sequence, a tool for studying human eye movements. Mathematical Biosciences 1975; 24: 191-204

Bahill AT; Adler D; Stark L.: Most naturally occurring human saccades have magnitudes of 15 degrees or less. Investigative Ophthalmology, 1975 14(6):468-9.

Borish IM : Clinical refraction. 3^rd ed. p 910 Chicago: Professional Press: 1970

Bradshaw MF, Glennerster A, Rogers BJ : The effect of display size on disparity scaling from differential perspective and vergence cues. Vision Research1996; 36: 1255-1264

Cheung BSK, Howard, IP :Optokinetic torsion: Dynamics and relation to circular vection. Vision Research1991; 31, 1327-1335

Cogan DG: Neurology of the Ocular Muscles, Ed 2. Charles C Thomas, Springfield, Illinois 1956

Zhang H; Gamlin PD.: Neurons in the posterior interposed nucleus of the cerebellum related to vergence and accommodation. I. Steady-state characteristics. Journal of Neurophysiology, 1998 ; 79(3):1255-69.