Ординатура / Офтальмология / Английские материалы / The Neurology of Eye Movements_Leigh, Zee_2006

ated, based upon the visual information that was acquired during the initial snapshot. Once the saccade is completed, the visual world is again sampled to determine if another saccade is still needed to bring the target of interest onto the fovea. Westheimer's results could then be interpreted by assuming that the return of the target to its initial position was not actually "seen" by the saccadic system until after the first saccade was made. Therefore, a normal saccadic latency, determined by the interval between snapshots, was required before making a second saccade to bring the eyes back to the target.

Although the sampled-data model accounts for many aspects of saccadic eye tracking, such a scheme does not explain all of the responses that normal individuals make. If Westheimer's experimental paradigm is expanded to include target jumps of different sizes and directions, and if large numbers of responses are analyzed, it can be shown that visualinformation can be continuously acquired and used to modify the initial saccadic response until about 70 msec before the movement begins.13'20'27 This is approximately the time it takes visualinformation to traverse the retina and central visual

pathways and reach the brain stem ocular motor mechanisms.

Furthermore, the saccadic system does not have an obligatory refractory period; two saccades may occur with virtually no intersaccadic interval in response to the appropriate sequence of double-step motion of the stimulus.29'153 Slow saccades, which occur in certain neurologic diseases, can be interrupted in mid-flight when the target position is changed, even after the eye has already begun to move.461 When presented with twodimensional, double-step stimuli, normal subjects make a single curved saccade rather than two successive straight saccades, indicating that the saccade trajectory has been modified in flight.426 The earliest responses to such two-dimensional, double-step stimuli suggest that direction and amplitude may be programed separately.147 Thus, the central nervous system appears able to change saccades at any stage, and even to overlap some presac-

cadic planning. Normal saccades only appear to be ballistic because of their high velocities and brief durations.

Saccades during Visual Search and Reading

The idiosyncratic pattern of eye move-

ments made when viewing a pictorial display is called a scan path.1283 It has been

shown that during manual tasks such as copying a design, frequent eye movements are used to scan the display for information, rather than committing that information to working memory.21'114 Such patterns may be severely disrupted in patients with neurologic lesions that create visual neglect or simultagnosia, but on occasion may be surprisingly nor- mal.239-331'332 Saccades made during visual search for targets embedded in an array of stimuli are not random and such behavior can be quantified and used to study visual attention.2653

Interpretation of ocular motor behavior during reading remains a controversial issue and is certainly difficult to interpret in the context of the known control of sac-

cades. For example, there is disagreement as to whether the spaces between words,

or the words themselves, serve to guide saccades.111"113'327'438 It is suggested that

when subjects read music, the pattern of saccades reflects not the visual stimulus or the manual response but the flow of infor-

mation from the musical score to performance.220

There has been substantial research effort to determine whether developmental dyslexia is due to abnormal control of saccadic eye movements. The lack of consensus may, in part, reflect the heterogeneity of patients with dyslexia. In some patients, the underlying cause of the reading disability may be auditory-linguistic defects, and in others, visuospatial defects. Thus, shifts of attention are important in reading,201 and they may be disturbed in dyslexia. Cytoarchitectonic abnormalities in the cerebral hemispheres and in the

thalamus of dyslexic patients have been reported,142-143 and EEC and evoked po-

tential asymmetries have been noted.102-103 Defects in interhemispheric transfer of information have also been implicated in dyslexia.178

One reproducible finding is that dyslexic children often show impairment of steady fixation,34'107 with excessive numbers of square-wave jerks.80 The presence of this fixation abnormality in dyslexics during non-reading tasks indicates that the underlying abnormality is not caused by language problems alone.107 Some individuals with developmental dyslexia show an ability to generate express saccades, even during the overlap stimulus (see Saccadic Initiation Time (Latency), above).62 It is suggested that their excessive distractibility for visual stimuli may reflect impaired "fixation" behavior, perhaps involving cortical projections to the rostral pole of the superior colliculus.35

Although the relationship between eye movement abnormalities and childhood dyslexia is not clear, patients with certain types of acquired ocular motor abnormalities do have difficulty reading.67 This effect is seen in patients with slow saccades, ocular motor apraxia, saccadicoscillations, and various forms of nystagmus.166 Furthermore, homonymous hemianopia may disrupt reading eye movements,especially when it is due to damage of the occipital white matter.467

Visual Consequences of Saccades

An important perceptual problem is how the brain can correctly interpret motion of images on the retina as being due to movement of the eyes rather than of the visual scene. It is possible to identify two components of this problem: the perception of motion during the eye movement, and correct localization of an object in space following a gaze shift.

VISUAL STABILITY

DURING SACCADES

We appear not to see during saccadic eye movements. Even though the seen world is rapidly sweeping across the retina, there

is no sense of motion or a blurred image. One proposed explanation for this is that the clear perceptions before and after a saccade would mask the gray-out due to image motion on the retina at speeds up to 500°/sec, and both behavioral and elec-

trophysiological findings suggest that this may be part of the explanation.58'2583

However, if subjects attempt to view a moving stimulus, it is still possible to see the lower spatial frequencies in an image when it moves across the retina at speeds of up to 800°/sec.50 The situation is different if vision is tested during saccades: there is a selective suppression of vision for lower spatial frequencies.49 It has been suggested that this saccadic suppression selectively affects the magnocellular pathway, which is mainly responsible for "motion vision." The mechanism of this suppression is presently unknown, but might occur as early as the lateral geniculate nucleus.344 Functional imaging studies have demonstrated decreased regional cerebral blood flow in striate cortex during repetitive saccades made in darkness, which is consistent with the notion of saccadic suppression of visual inputs.306

SPATIAL CONSTANCY FOLLOWING SACCADIC GAZE SHIFTS

While every saccadic eye movement causes the entire visual world to be shifted upon the retina, we are still able to maintain an appropriate sense of straight-ahead. How do we ensure such spatial constancy? The classic explanation is that our central nervous system monitors the effort of will (efference copy) and then sends this motor information (as a corollary discharge) to sensory systems.172 In this way, our perceptual sense knows and adjusts for the shift of images upon the retina using an egocentric frame of reference. It is also possible that extraocular proprioception could serve this function, but available evidence suggests that such inputs are more important for long-term adap-

tive changes in the ocular motor sys-

tems.145-160'236

Other evidence suggests that the brain estimates the location of objects in space with reference to other objects in the

visual scene; these are called egocentric cues.83'308 For example, if visual targets are flashed just prior to or during a saccade, they are incorrectly localized asjudged by

subsequent saccades or finger pointing.57'83'193'256 If efference copy were the

mechanism by which spatial constancywas maintained, then there should be no difference in spatial localization of targets presented just before or after a saccade.

The changes in visual responses to targets flashed just before a saccade may reflect changes in apparent visual direction and a compression of visual space that antici-

pates the consequences of the upcoming saccade.259'345

What relative reliance does the brain place on visual and extraretinal estimates of the direction of gaze? It seems that visual estimates are more important, if they are available. Thus, a classic line of evidence to support the role of efference copy in spatial localization is that, in darkness, normal subjects perceive a

small afterimage, induced by a photoflash, as moving with the eye.244 The af-

terimage is stationary on the retina, and its apparent movement in space is probably due to efference copy signaling movement of the eye. However, if a large

afterimage of a complex scene is induced, it does not seem to move as the

eye drifts in darkness.308 Thus, a large visual afterimage appears to override nonvisual cues about eye movements. Other experiments have shown that visual estimates of the direction of gaze are given preference over efference copy, even if the visual information is corrupted by illusory stimuli.470 During saccades, visual inputs become less reliable, but the brain still puts more reliance on visual than on extraretinal information. Thus, if a visual target is displaced during a saccadic eye movement, its movement may go unnoticed.89 If, however, the target is only shown in its new position 100 msec after the saccade ends, then its displacement to a new position is detected. Thus, it seems that the brain weighs visual and extraretinal estimates of the direction of gaze,213 putting more reliance on the visual estimates except when visual factors are not available.

NEUROPHYSIOLOGY OF SACCADIC EYE MOVEMENTS

In this section we review the neural machinery by which saccades shift the line of sight so that an image detected in the retinal periphery is brought to the fovea, where it can be seen best. In primaryvisual cortex (Brodmann area 17, VI) the location of a visual stimulus is represented by the distribution of activity across a "place" map. Thus, different parts of this map correspond to different locations on the retina. The neural representation of the motor command for the saccadic response is quite different. The ocular motoneurons encode the characteristics of the saccade in terms of their temporal discharge; the size of the saccade is proportional to the total number of discharge spikes. The ocular motoneurons lie in the third, fourth, and sixth cranial nerves and cause the extraocular muscles to move the eyes with respect to the head, in craniotopic coordinates.

This means that the brain must transform the stimulus, which is encoded in terms of the location of active neurons withinvisual cortex (i.e., place-coded), into the saccadic command on ocular motoneurons, which is encoded in terms of discharge frequency and duration (i.e., temporally-coded). Furthermore, a transformation from retinal coordinates into craniotopic coordinates is necessary. The retinal coordinates are twodimensional, whereas the eye rotates about three axes.79 We will return to these issues in the following section, Brain Stem Pathways for Saccades, and in our discussion of the cortical and subcortical structures that project to thesepathways.

Brain Stem Pathways for Saccades

FINAL SACCADIC COMMAND FROM

OCULAR MOTONEURONS

Electrophysiologic studies of extraocular muscle activity in humans, and of ocular motoneurons in animals, have delineated the innervational changes that accompany saccadic eye movements (see Fig. 1-3,

Chap. 1). During a saccade, a highfrequency burst of phasic activity can be

recorded from the agonist ocular muscle and, as shown in experimental animals, from the corresponding ocular motoneurons. This burst of activity, the saccadic pulse of innervation, starts about 8 msec before the eye movement425 and generates the forces necessary to overcome orbital viscous drag so that the eye will quickly move from one position to another. Following a saccade, the agonist eye muscle and its ocular motoneurons assume a new, higher level of tonic innervation, the saccadic step of innervation, which holds the eye in its new position against orbital elastic restoring forces. The transition between the end of the pulse of innervation and the beginning of the step of innervation is not abrupt but gradual, taking up to several hundred msec. This is the slide of innervation. Hence the change in innerva-

tion accounting for saccades is actually a pulse-slide-step (Fig. 3-5).294'335 If one

records from the antagonist muscle or its motoneurons, one finds the reciprocal innervational changes.390 The antagonist muscle is silenced during the saccade by an inhibitory, off-pulse of innervation; at the end of the saccade, the antagonist assumes a new, lower level of tonic innervation, the off-step. Measurement of muscle forces generated by extraocular muscles indicates that the eye comes to rest at the end of a saccade owing to the viscous forces of the orbital tissues rather to than any "active braking" by the antagonist muscle.257

BRAIN STEM SACCADIC

PULSE GENERATOR

Two types of neurons are critically impor-

tant in the generation of saccades: burst cells* and omnipause cells.52'173'215'399 Fol-

*In burst cell nomenclature, cells in the brain stem reticular formation that burst (discharge) for saccades are of two main types: medium-lead units that begin their discharge about 12 msec before saccades, and long-lead units that begin their discharge over 40 msec before saccades. Hereafter, we will call medium-lead cells premotor burst neurons or simply burst neurons. Long-lead burst neurons will be referred to by their initials, LLBN. There are also cells that burst for saccades in the superior colliculus, called collicular burst neurons. Burst neurons are of two types: excitatory and inhibitory.

Figure 3-5. Pulse-step-slide of innervation during a 20° leftward saccade in the rhesus monkey. (Top trace) Eye movementrecording from left eye. (Bottom trace) Single unit activity from a neuron in the left abducens nucleus showing the pulse, slide (horizontal bar), and step change in innervation. (Courtesy of H. P. Goldstein.)

lowing the saccade, the eye is held in position by a tonic, step command that is generated by the neural integrator (see Chap. 5 and Fig. 1-3, Chap. 1).A century of clinical experience has demonstrated that the caudal pons is important for horizontal saccades and the rostral mesencephalon for vertical saccades.132'403 For horizontal saccades, burst neurons within the paramedian pontine reticular formation (PPRF) are essential (see Display 6-3, Fig. 6-1,

and Fig. 6-2, Chap. 6). For vertical and torsional saccades, burst neurons lying in the rostral interstitial nucleus of the medial longitudinal fasciculus (riMLF) play the equivalent role (see Display 6-5, Fig. 6-3, and Fig. 6-4). Omnipause neurons lie in the nucleus raphe interpositus, in the midline of the pons (see Display 6-3 and Fig. 6-2).

PREMOTOR BURST NEURONS

Pontomedullary Burst Cells

Excitatory burst neurons (EBN) in the TPRF lie rostral to the abducens nucleus, corresponding to dorsomedial nucleus reticularis pontis caudalis.195 These EBN begin discharging at a high frequency, about 12 msec prior to, and time-locked with, the horizontal component of all types of rapid eye movements, including quick phases.176-407'425 Recent evidence suggests that some EBN in the PPRF en-

code saccades monocularly (i.e., for movements of one eye or the other).2473'466 The

EBN discharge preferentially for ipsilateral saccades and they appear to create the immediate premotor command that generates the pulse of activity for horizon-

tal saccades. Three pieces of evidence support this hypothesis. First, during sac-

cades, the instantaneous burst cell firing

rate of EBN is closely correlated with instantaneous eye velocity,175'425 and the to-

tal number of spikes in the burst of activity (the integral of the discharge rate) is pro-

portional to the amplitude of the ipsilateral, horizontal component of the saccades. Second, stimulation of the PPRF elicits ipsilateral saccades.69 Third, a unilateral lesion within the PPRF abolishes the ability to generate ipsilateral saccades.174 It should be noted, however, that

EBN in the PPRF also discharge during vertical and oblique saccades,323'425 and bi-

lateral PPRF lesions not only abolish horizontal saccades but also cause slowing of vertical saccades.165'174

EBN project directly to the ipsilateral abducens nucleus to contact both abducens motoneurons and internuclear neurons. The latter project up the con-

tralateral MLF to contact the medial rectus subgroup of the contralateral oculomotor nucleus. EBN also project to ipsilateral inhibitory burst neurons, to the perihypoglossal and vestibular nuclei, to the reticular formation adjacent to the abducens nucleus, and to the cell groups of the paramedian tracts (see Display 6-4, Chap. 6). Thus, for horizontal saccades, the excitatory pulse reaches the ocular motoneurons from the EBN in the ipsilateral PPRF. As discussed in Chapter 5 (Fig. 5-3), the step of innervation that is required to hold the eye steady at the end of the saccade arises from the neural integrator, for which the nucleus prepositusmedial vestibular nucleus complex (NPHMVN region) is most important.

Inhibitory burst neurons (IBN) for horizontal saccades have been identified just caudal to the abducens nucleus in the nucleus paragigantocellularis dorsalis of

the dorsomedial portion of the rostral medulla.195'374'408 The IBN send their axons

across the midline to the contralateral abducens nucleus to inhibit contralateral abducens motoneurons and interneurons during ipsilateral saccades. The IBN also project to the vestibular nuclei, nucleus prepositus, and portions of the pontine reticular formation.408 Thus, the role of IBN is to silence activity in the antagonist muscle during horizontal saccades.

Midbrain Burst Cells

The EBN in the riMLF encode the vertical and the torsional components of saccades, just as EBN in the PPRF encode the horizontal component.51'219'435 Excitatory and inhibitory EBN for upward and for downward saccades appear to be intermingled in the riMLF, although their projection pathways show some differences.264'265 The EBN discharge most vigorously for rapid eye movements that rotate the globe in a plane parallel to that of a pair of reciprocally acting vertical semicircular canals (for example, right anterior and left posterior). Hence, EBN in one riMLF increase their discharge when the eye on the same side extorts and the eye on the opposite side intorts. While the direction of torsion

is fixed for EBN on each side, the direction of vertical rotation is upward in some and downward in others. Thus, unilateral lesions have only mild effects on vertical saccades, but abolish ipsilateral torsion. For example, with a lesion of the right riMLF, torsional quick phases, clockwise from the point of view of the subject, (extorsion of the right eye and intorsion of the left eye) are lost.413 Bilateral lesions in riMLF abolish all vertical and torsional saccades.413

Vertical EBN project directly to vertical ocular motoneurons in the CN III and CN IV nuclei and send axon collaterals to the

interstitial nucleus of Cajal (see Fig. 6-4, and Display 6-6).264'265 The latter struc-

ture is an important component of the ve- locity-to-position integrator for vertical and torsional eye movements. Thus, for a vertical or torsional saccade, the pulse reaches the ocular motoneurons from the riMLF, whereas the step of innervation comes mainly from the interstitial nucleus of Cajal. This scheme is supported by the results of pharmacologically inactivating the interstitial nucleus of Cajal; vertical and torsional saccades can still be made, but there is centripetal postsaccadic drift.78

OMNIPAUSE NEURONS

Omnipause cells lie in the nucleus raphe interpositus, which is located in the midline between the rootlets of the abducens nerves (see Fig. 6-2, Chap. 6).53>229 These neurons utilize glycine as their neurotransmitter,196 which is consistent with their inhibitory function. A number of structures project to the omnipause cell region, including the rostral pole ("fixation zone") of the superior colliculus,55'144 the frontal eye fields,404 the supplementary eye fields,386 the central mesencephalic reticular formation, the long-lead burst neurons in the rostral pons and midbrain, and the fastigial nucleus.53'375 Omnipause cells send inhibitory projections

to EBN in the pons, to IBN in the medulla, and to the riMLF.53>196'276'286'406

Omnipause neurons discharge continuously except immediately prior to and

during saccades, when they pause. Omnipause cells cease discharging during saccades in any direction, hence their name. Omnipause cells also cease discharging during blinks.176 When omnipause cells are experimentally stimulated in the monkey, the animal is unable to make saccades or quick phases in any direction, although other types of movements, such asvestibular slow phases, can still be elicited.451 If omnipause cells are stimulated during a

saccade, the eye movement is aborted in mid-flight.216

On the basis of these findings, it has been hypothesized that omnipause cells tonically inhibit all burst cells (Fig. 3-6), and when a saccade is called for, the omnipause cells themselves must be inhibited to permit the burst cells to discharge. By acting as an inhibitory gate, omnipause cells help maintain the necessary synchronization of the activity of premotor saccadic burst neurons to drive the eyes rapidly during the saccade and to keep the eyes still when the saccade is over. Experimental lesions with excitotoxins in the omnipause region have the predominant effect of making horizontal and vertical saccades slow.208 This effect may be because the omnipause neurons normally synchronize the onset and offset of burst neuron discharge and, after such lesions, the activity of the burst neurons is no longer coordinated. However, it also remains possible that these experimental lesions also affected the nearby EBN in the PPRF.

LONG-LEAD BURST NEURONS AND THE CENTRAL MESENCEPHALIC

RETICULAR FORMATION

Neurons that start to discharge 40 msec or more before saccades are found throughout the brain stem. Some long-lead burst neurons (LLBN) lie in the midbrain and receive projections from the superior colliculus.375 They project to pontine EBN, medullary IBN, and omnipause neurons; they also project to the nucleus reticularis tegmenti pontis (NRTP). These mesencephalic LLBN discharge before and dur-

ing saccades to their movement field. The portion of the mesencephalic reticular formation that lies just lateral to the CN III nucleus (central mesencephalic reticular formation, cMRF)70 contains neurons that have reciprocal connections with the superior colliculus,263 and it has been postulated that they may serve in a feedback loop, perhaps acting as a resettable integrator for saccades (see Models for Saccadic Pulse Generation, below).442 However, since these neurons also receive projections from the supplementary eye fields and fastigial nucleus, project to the PPRF, and start to discharge more than 40 msec before saccades,163 they may also serve a long-lead function, perhaps transforming spatially coded to temporally coded commands.175 Experimental lesions of the cMRF cause hypermetria of contralateral and upward saccades and hypometria of ipsilateral and downward saccades.441 More rostral inactivation of the MRF impairs vertical saccades.443

Other LLBN lie in NRTP and project mainly to the cerebellum via the middle peduncle; some of these cells project to the PPRF.376 Thus, it seems that LLBN may serve more than one function. While those LLBN that receive input from the superior colliculus may play a crucial role in a spatial-to-temporal transformation of saccadic commands, other LLBN may synchronize the onset and end of saccades, by

virtue of their projections to omnipause neurons.175'375

Models for Saccadic

Pulse Generation

MODELS FOR HORIZONTAL SACCADES

Early hypotheses for the generation of saccades proposed that the duration of the pulse of activity that creates saccades was predetermined or preprogramed accord-

ing to desired saccadic amplitude. Studies of patients with slow saccades, however, led to an alternative hypothesissuggesting that saccades are generated by a mechanism that drives the eyes to a particular orbital position rather than moving the eyes a specified distance.461 By continuously comparing desired eye position and actual eye position (the latter is probably based on monitoring an internal efference copy of the eye position command), the neurons that generate the saccadic pulse are driven until the eye reaches the target, when they automatically cease discharging. This is the original local-feedback

model for saccades proposed by D.A. Robinson334'461 (Fig. 3-7A). It has the ad-

vantage of automatically generating the main-sequence relationship of saccades. The model also accounts for slow but accurate saccades made both by some patients with neurological disease and by normal subjects taking various sedative or hypnotic medications.206 It can also produce saccadic oscillations such as flutter if the omnipause neurons malfunction or are inhibited.462 Although the notion of local feedback has sometimes been called into question,208 the evidence to support it remains substantial. Thus, inactivation of the PPRF causes slow, usually orthometric saccades, indicating that local feedback sustains the reduced discharge of premotor burst neurons until the error signal is zero.401 Perhaps the most compelling evidence is that if a saccade is arrested in mid-flight by briefly stimulating the omnipause neurons, a new saccade is generated to get the eye on target within 70 msec, which is shorter than could have been achieved by responding to the visual consequences of the arrested movement.216

More recent physiological studies have called for modifications of the original Robinson model, while retaining the notion of local feedback control of saccade generation. One important revision is that the command signal is a desired change in eye position206 (Fig. 3-7B). This signal would be compared continuously with an efference copy of the actual change in eye position in order to determine when to terminate the saccade. Inherent in this modification of the local-feedback model

is the idea of separate integrators—one common neural integrator for conversion of eye velocity to eye position commands (for all types of conjugate eye movements) and a separate resettable neural integrator that operates on saccadic velocity commands in the feedback loop that

controls the duration of the saccadic pulse.1'206'226'278-396'399'400 However, the na-

ture of a separate resettable integrator for saccades is not yet settled.153'359

What presently remains unresolved is the anatomical basis for the local feedback mechanism. One proposal is that the feedback may involve LLBN rather than premotor burst neurons (Fig. 3-7C).371 In this model, LLBN are also the proposed site for the second, saccade-specific integration. However, this model is not consistent with the anatomical projections of the superior colliculus,55 cannot simulate the staircase of saccades that occurs with sustained stimulation of the superior colliculus,42 and does not account for saccadic oscillations. Another suggestion is that feedback control of saccades and the saccadic integrator involve the superior colliculus.440 The electrophysiological evidence to support this concept is discussed below in the section Superior Colliculus. Third, it has been proposed that local feedback occurs via a cerebellar loop.232a'322 This proposal is consistent with the anatomical connections between the premotor burst neurons, dorsal vermis and fastigial nucleus, which are reviewed below (Effects of Total Cerebellectomy on Saccades). Further, hypermetria is a cardinal sign of fastigial nucleus lesions, and it is argued that this structure serves an important role in the feedback control of saccades.2323'322 Inactivation with muscimol of premotor burst neurons to which the fastigial nucleus projects causes both slowing and hypometria of saccades.372 More experimental tests of these hypotheses are needed.

MODELS FOR OBLIQUE SACCADES

The ocular motoneurons innervate extraocular muscles that rotate the eyes approximately in Cartesian coordinates (e.g., the medial and lateral rectus rotate the

globe a specified distance horizontally). However, the premotor burst neurons discharge for oblique movements, and so one aspect of modeling saccades is whether they encode the saccadic command in polar coordinates. In other words, how independent are the two populations in the PPRF and riMLF? This becomes an issue during programing of oblique saccades, when the activity of the two separate populations of neurons need to be coordinated. In one view of the way that oblique saccades are programed, the common source saccadic model, the command from the burst neurons is specified in polar coordinates: an oblique (radial) velocity at angle 0 427 Then, neural circuitry converts this into a signal multiplied by cosine G for the horizontal motoneurons and a signal multiplied by sine 6 for the vertical motoneurons. This model predicts that (7) the duration of the horizontal and vertical components of oblique saccades may differ from those produced when similar-sized movements are made as purely horizontal or vertical saccades; (2) the horizontal and vertical components will have synchronous onset and offset; and (3) the trajectory of the oblique saccade will be straight. An alternative hypothesis, the Cartesian coordinate model, proposes that the central command for the oblique saccade is broken down into horizontal and vertical components before being sent to the horizontal and vertical burst neurons.30'157 The critical predictions of this model are the following: (1) the horizontal and vertical saccadic components of oblique saccades will have the same duration as when made as purely horizontal or vertical saccades; (2) although the horizontal and vertical components of oblique saccades will have a synchronous onset, they may end at different times; (3) the trajectory of oblique saccades could be curved.

Normal human saccades are brief, and it is often difficult to confirm these predictions. However, oblique saccades that have dissimilar horizontal and vertical components appear curved,18 and have different times of offset of the two components.157 Studies of patients with selective slowing of the vertical or horizontal components provide further evidence to support a Cartesian model.461 Thus, patients with

selective slowing of vertical saccades due to Niemann-Pick type C disease show

markedly curved oblique saccades (Fig. 3-3B).348 The initial movement wasmainly

horizontal and most of the vertical component occurred after the horizontal component ended. After completion of the horizontal component of an oblique saccade, the eyes oscillated horizontally at 10-20 Hz until the vertical component ended (Fig. 3-3C). These horizontal oscillations may occur because the omnipause neurons are silent until the vertical component is complete, and the normal horizontal burst neurons oscillate until the whole saccade is over. Additional support for independence of the horizontal and vertical premotor burst neurons comes from electrically stimulating the superior colliculus during oblique saccades.279 To resolve the issue of how oblique saccades may have a straight trajectory in monkeys, it is proposed that prolongation of the smaller component is achieved down-stream from the colliculus, by "cross-coupling" between the horizontal and vertical burst neurons.30-157'279 However, such models do not take account of the neurophysiological finding that some premotor burst neurons discharge maximally for directions tilted from their cardinal direction (e.g., horizontal for the PPRF). A model that incorporates such variations within a distributed network of neurons is able to simulate several characteristics of oblique saccades, with no cross-coupling.323

MODELS FOR THREEDIMENSIONAL SACCADES

The development of reliable methods to measure 3-D eye rotations has led to the development of models to account for rotations of the eye in three planes during saccades. In fact, eye rotations are essentially restricted to rotation about axes that lie in the frontal plane (Listing's plane; see Fig. 9-3, Chap. 9), and so three degrees of freedom are reduced to two. What remains to be settled is the mechanism that imposes Listing's law, and the relative contributions made by the mechanical and suspensory

properties of the orbital tissues on the one hand,84'295-324-326'370 and by neural factors on the other.420a'421 At present, models as-