The brain circuitry underlying image stabilization is organized around a vestibular apparatus in the temporal bone that can sense head rotations in any direction. The basic VOR is augmented by the OKR and several other specialized circuits, located in the brain stem and midline cerebellum, to become a very sophisticated system for minimizing retinal slip. These additional circuits are schematized in Fig. 8 and described next.
1. Circuit 1: The Three-Neuron Vestibuloocular Arc
How the brain fashions image-stabilization eye movements based on the vestibular sense has been extensively studied. Of prime importance is the vestibular sensory organ and the three-neuron arc composed of neurons in the vestibular ganglion, the vestibular nuclei, and the oculomotor nuclei.
On each side of the head within the labyrinth of the inner ear lie three semicircular canals filled with endolymph. Sensory receptor cells project hairs into a gelatinous mass, the cupula, that extends across the canal. When the head moves the bony canals move with it, but the endolymph lags behind, bending the cupula and the hairs embedded in it. Deforming the hairs hyperpolarizes their receptor cells for one direction of head rotation and depolarizes them for the opposite direction, and the resting discharge rate of bipolar neurons of the vestibular ganglion that innervate the receptor cells is modulated up or down by the receptor polarization (Fig. 6).
The central course of the bipolar cells' axons is the vestibular component of nerve VIII. They synapse in the vestibular nuclei, located in the medulla. Each canal projects to both the medial and the superior vestibular nucleus, and some vestibular afferents go directly to the cerebellum.
The three semicircular canals (horizontal, anterior, and posterior) lie in three different planes at approximately right angles to each other so that rotations of the head about any axis can be sensed. The planes of the three semicircular canals approximately correspond to the planes of action of the three pairs of extraocular muscles (Fig. 7), and the three-neuron pathway connects each canal to the extraocular muscles that move the eye in the canal's plane. This pathway was illustrated for the horizontal plane in Fig. 6. Horizontal head rotation excites the three-neuron pathway leading from the horizontal canal to the medial and lateral rectus, thus effecting an appropriate horizontal counterrotation ofthe eyes. Each canal also has a three-neuron inhibitory pathway to the corresponding antagonist muscles via inhibitory relay neurons in the vestibular nucleus. There are left and right sets of canals, and the symmetry of the VOR circuitry effectively uses the difference of the opposing signals from the left-right canal pair in each plane to increase the sensitivity of the VOR response, as shown in Fig. 6 for the horizontal canals.
The same principles hold for the anterior and posterior canals and the remaining four extraocular muscles, but in a less straightforward manner. The anterior and posterior canals lie in approximately vertical planes that are midway between the sagittal and coronal planes (Fig. 7). Thus, they do not correspond to the x-y-z axes of eye rotations described previously. Each anterior canal is paired with the contralateral posterior canal because they are in approximately the same plane and have the opposite responses to rotations in that plane. Also, each anterior-posterior pair controls a yoked pair of extraocular muscles that comes closest to its plane. For example, as can be deduced from Fig. 7, the excitatory three-neuron pathway of each anterior canal leads to the ipsilateral superior rectus muscle and the contralateral inferior oblique.
2. Circuit 2. Position Holding Mechanism (the ''Neural Integrator'')
The first problem for the basic three-neuron VOR circuit occurs when the head stops moving after a brief
Figure 8 Neural circuits for image stabilization. (A) Basic three-neuron VOR circuit. Head rotations transduced by the semicircular canal (SCC) activate a three-neuron circuit that rotates the eye in the opposite direction. The circuit is as follows: neurons in the vestibular ganglion (VG) to neurons of the vestibular nuclei (VN) to oculomotor neurons (OMN) in one of the three oculomotor nuclei. (B) The VOR with the addition of a neural integrator. (Top) The neural integrator for horizontal eye movements is located in the n. prepositus hypoglossi (NPH) (and in the medial vestibular nucleus, which is not illustrated). Its integration function is symbolized by the recurrent connection of the NPH onto itself. (Middle) The plots show idealized spike rate functions for the components of this circuit, with idealized spikes above the plot boxes. With the neural integrator, retinal slip is minimal as gaze is held steady during and after the head movement. (Bottom) Without the integrator, the OMN does not cancel the elastic forces of the eye; thus, the eye slides back to the center of its orbit, both during and after the head movement, resulting in considerable retinal slip. (C) The quick-phase addition to the VOR. The quick-phase addition is symbolized by the inclusion of the saccade generator circuit with its excitatory (EBN) and inhibitory burst neurons (IBN) that enable the eye to be rapidly "reset" to the central position. Again, the plot boxes and spikes are hypothetical neural responses of the components of this circuit. (D) Addition of the optokinetic reflex (OKR) and a circuit for VOR gain adaptation. Visual motion signals from the nucleus of the optic tract (NOT) and the accessory optic system (AOS) go directly to the VN to add an OKR to the image-stabilization system. This OK signal is also sent to the inferior olivary nucleus (IO), and then to the cerebellar Purkinje cells (Pc) via their climbing fibers (cf). Likewise, the canal signal is relayed to the cerebellum's floccular Pc on mossy fibers (mf) that are axon collaterals of vestibular neurons via granule cells (gc) and their parallel T fibers. These inputs enable the flocculus to adaptively adjust VOR gain via its direct projections to the VN; the VN cells that receive this inhibitory projection are termed floccular target neurons (FTN). Pcs are also thought to receive a copy of the eye velocity signal, hence "E?" (adapted with permission from A. E. Luebke and D. A. Robinson, Exp. Brain Res. 98, 379-390, 1994. Copyright © 1994 by Springer-Verlag).
turn. The canal signal quickly returns to baseline when the head stops moving because the cupula is no longer distended. However, if the innervation of the extraocular muscles also returns to its baseline level, then the elastic forces of the orbit would pull the eye back to the central position, thus negating the benefits of the VOR because vision would be blurred throughout this prolonged return.
Thus, for stable vision, the eye should not only stop rotating when the head stops turning but also must be held stationary at its current position and not slip back toward the central position. This is elegantly accomplished by the brain stem's neural integrator, a hypothesis of David Robinson, who reasoned that a command for eye position could be obtained by integrating eye velocity commands. Robinson's integrator now has a specific location, a neural signature (tonic activity coding static eye position), and distinctive neurological consequences of injury to it (e.g., gaze-evoked nystagmus).
The stabilization system, based on the VOR with the addition of the neural integrator, is shown in Fig. 8B, with the neural integrator represented by a local recurrent connection. Such a local recurrent connection was perhaps the earliest neural mechanism proposed for a short-term memory by Rafael Lorente de No and others. As the VOR "eye velocity signal'' moves the eye, the integrator output grows, continually signaling where the eye should be held both during the VOR movement and after it. Thus, the neural integrator serves to exactly overcome the elastic force that accompanies increasingly eccentric eye rotations. Referring back to the equation ROMN=A • (0—0O) + B • (d0/dt), the head velocity signal from the vestibular sensory nucleus is fed to the oculomotor nuclei as a command for eye velocity (B • (d0/dt)) [where B is the synaptic strength] and the neural integrator obtains its eye position signal by integrating the eye velocity command, d0/dt, to obtain the A • (0-0O) term for the OMN. This scheme is used for all types of eye movements: The command signal is formulated as a desired eye velocity (the velocity command hypothesis), and a common neural integrator integrates velocity commands of all types to continuously maintain the eye position signal.
What if a large head rotation is made, or there are successive turns in the same direction? The eye can move only about 40-60° in any direction from its central position in the orbit. Thus, the VOR/neural integrator circuit will quickly drive the eye to the extreme of the orbit and the retinal image will begin to slip. Furthermore, in extreme orbital positions the eye's visual field is partially occluded.
Nature's solution is to periodically reset the eye back to the central position (usually slightly past it) during prolonged or large head rotations. Moving back to center results in considerable blur; however, the vertebrate solution is for the reset mechanism to recenter the eye as fast as possible in order to minimize the duration of the temporary blindness due to the high-velocity smearing of the retinal image. This requires near-maximal contraction of the agonist muscles coupled with complete relaxation of the antagonist muscles in order to produce eye movements approaching 1000°/sec and directed opposite the compensatory vestibuloocular movement.
The resets are obvious when prolonged rotation occurs, the resulting to-and-fro pattern of movement is called nystagmus (Fig. 9). In vestibular nystagmus, fast movements alternate with slow ones, yielding a characteristic sawtooth waveform. The slow phase is the compensatory reflex, and the rapid resets are called the quick phase. The quick-phase circuit is indicated by the addition of "burster" neurons in Fig. 8C. The excitatory burst neurons (EBNs), with their high spiking rate, project to the OMNs to realize the quick-phase movements. Actually, these EBNs are the output neurons of the "saccade generator'' circuit, so labeled because it is appropriated by the foveation system to effect voluntary saccadic eye movements that are completely independent of the VOR. Moreover, the saccade generator is usually studied in the context of voluntary saccadic eye movements, in part because rotating subjects about different axes to achieve vestibular nystagmus can be a daunting experiment. Therefore, the saccade generator is further elaborated later in conjunction with voluntary saccades.
4. Circuit 4. The Optokinetic Reflex Complements the VOR
The VOR handles briefhead movements very well, but it has trouble with prolonged rotation. This is because, with continued rotation of the head, the endolymph in the semicircular canals begins to catch up with the movement of the head, the cupula returns to its resting position, and hence the vestibular afferents return to their baseline rate, falsely indicating that the head is no longer rotating. Consequently, for continued rotation in the dark (or with the eyes closed) vestibular nystagmus will gradually slow down and completely stop after 30-60 sec. One solution to this problem of vestibular transducer adaptation is to let retinal slip help in driving this compensatory reflex. This is OKR, which is added to the basic VOR circuit in Fig. 8D. The principal sources of this retinal slip signal are two lesser known targets of the optic nerve: the nucleus of the optic tract (NOT) and the accessory optic system (AOS). Cells in these brain stem nuclei are tonically driven by the movement of large patterned visual stimuli, the best stimulus for eliciting OKR, with different neurons tuned to different directions of motion. In primates, the responses of these cells depends not only on their direct retinal inputs but also on pathways that involve neocortex, especially primary visual cortex and extrastriate motion areas. The direct and indirect projections of AOS/NOT to the vestibular nuclei complete the OKR circuit, which means that neurons in the vestibular nuclei also respond to visual (OK) stimuli. Moreover, despite
Figure 9 The optokinetic reflex and nystagmus. (A) A depiction of a rotating chair and drum (with striped inner walls) used for testing the vestibuloocular (VOR) and optokinetic reflexes (OKRs), together and separately. Rotation of the drum (surround) alone, with illumination, elicits the OKR in a stationary subject. Rotation of the chair in the dark elicits the VOR. Rotation of the chair in the light activates both reflexes. (B) Idealized nystagmus during chair rotation with full illumination. (Top) The chair velocity plot indicates leftward rotation. (Middle) The horizontal (HOR) eye position record shows slow-phase movements, directed opposite the chair rotation, which compensate for the chair rotation, combined with quick-phase movements in the direction of rotation that periodically reset the eyes toward their central positions in the orbit. (Bottom) The "envelope" of the velocity trace also shows the rightward direction of slow-phase motion, which is quite distinct from the brief, high-velocity leftward-moving quick phase of the nystagmus. (C) Experimental nystagmus data (1) (Top trace) Horizontal eye movement record showing vestibuloocular nystagmus elicited by chair rotation in the dark. (Bottom trace) Eye velocity record from the same trial (see Fig. 9B). The velocity envelope shows that slow-phase rightward movements are elicited with a very short latency after rotation begins but taper off as rotation continues (adaptation). Note that the postrotatory response is in the opposite direction. The exponential dark lines, in both the adaptation and the postrotatory epochs plot the theoretical strength of the vestibular canal signal, which adapts much faster than the slow-phase velocity adapts. (2) Eye velocity during drum rotation in the light. Optokinetic nystagmus (OKN) refers to the combination of the slow-phase OKR response and the quick-phase resets. The OKN velocity envelope during rotation is similar to that of the VOR except that (i) there is no diminution in the response to ongoing rotation in the light and (ii) there is a more gradual rise time to the maximal slow-phase velocity. Also notice that the optokinetic after nystagmus (OKAN) is in the same direction as the OKN, whereas the postrotatory response was opposite the VOR. (3) Chair rotation in the light elicits both VOR and OKN responses. There is a quick rise in slow-phase velocity because of the VOR, no adaptation because of the OKN, and no postrotatory nystagmus because the OKAN cancels the postrotatory nystagmus from the canals (Fig. 9C adapted with permission from T. Raphan, V. Matsuo, andB. Cohen, Exp. Brain Res. 35,229,1979. Copyright © 1979 by Springer-Verlag).
the addition of many refinements to the VOR, and the addition of new types of eye movements, the vestibular nuclei remain the principal gateway to the oculomotor nuclei.
Since the OKR is indefatigable, why not dispense with the VOR and base image-stabilization eye movements solely on the OKR? The reason is that the OKR is much slower to react than the VOR. OKR latency is 50-100 msec (reflecting the slow pace of visual processing), whereas VOR latency is ~ 10 msec. To demonstrate that OKR alone is poor, rotate the head left and right while reading. The text is legible because image stability is provided by the VOR. Now move the book left and right with the head still; the page will be blurred despite the OKR. This test hints at another reason the VOR is needed: Once a visual image is moving very fast across the retina, then it is blurred and hence is a poor stimulus for engaging the OKR in order to decrease its retinal slip and blur.
VOR adaptation during prolonged rotation in the dark reflects the adaptation of the vestibular transducer in the semicircular canals. However, closer inspection of this phenomenon reveals that the vestibular apparatus adapts rapidly, with a time constant of ~ 6 sec, but the reflexive eye movement adapts with a much longer time constant (15-30 sec) (Fig. 9C.1). The reason is that a central neural process stores up the vestibular signal, thereby providing a more veridical representation of head motion than is provided by the raw VIIIth nerve signal.
This velocity-storage circuit stores optokinetic as well as the vestibular velocity (because they are merged in the vestibular nucleus). Indeed, the OKR does not stop immediately with the cessation of OK stimulation but continues, in the same direction, for several seconds after the lights go off (Fig. 9C.2). This continuation is called optokinetic afternystagmus (OKAN). Thus, the OKR is like charging a battery: It takes several seconds of stimulation for the OKR to reach its asymptote velocity (Fig. 9C.2) and to ensure robust OKAN after the optokinetic stimulation ceases.
This velocity storage underlies another aspect of the symbiosis between the VOR and the OKR regarding cessation of prolonged head rotations. If the VOR is adapted when rotation stops, the canal endolymph continues to move and produces a strong VOR in the opposite direction, called "postrotatory nystagmus'' (Fig. 9C. 1). It is accompanied by a loss of balance and vertigo and lasts several seconds. However, postrota-tory nystagmus is best demonstrated by rotation in the dark so there can be no optokinetic stimulation. Indeed, there is usually no problem after rotation in the light (Fig. 9C.3) because OKAN serves to cancel postrotatory nystagmus.
Whereas the OKR is a closed-loop system in that the response cancels its own input signal, the VOR is an open-loop system because the counterrotation of the eye has no direct effect on the semicircular canal head-rotation signal. Also, because it is an open-loop system, the VOR reflex strength is critical. Ideally, VOR gain is exactly 1.0; however, it would seem too much for the genome to so completely specify the VOR circuitry so as to yield this ideal gain. Instead, an adaptation circuit continually adjusts VOR reflex strength so as to minimize the retinal image slippage that accompanies head movements. For example, wearing 2 x magnifying goggles for a few days will drastically increase VOR gain because 2 x goggles require that the VOR gain be 2.0 to cancel retinal slip when the head moves. In real life, less drastic but nevertheless critical adjustments of VOR strength need to be made—for example, when the head changes size during development, when people don spectacles (which more modestly magnify or minify the visual world), or whenever vestibular hair cells die or extraocular muscles weaken because of old age or other factors.
The crucial brain structure for VOR gain adaptation is the cerebellum, specifically the flocculonodular lobe that is interconnected with the vestibular nuclei that mediate the VOR (Fig. 8D). Monkeys without a cerebellum still have a robust VOR; however, VOR gain does not adapt in response to experimental goggles and other manipulations that induce adaptation in normal subjects. Other structures critical for VOR gain adaptation include the NOT and AOS, which provide an optokinetic signal not only directly to the vestibular nuclei but also to cerebellar cortex via the dorsal cap of the inferior olive. The overall goal of the gain-adjustment circuit (Fig. 8D) is to minimize the optokinetic signal that the cerebellum ''sees''; the better the VOR works, the less the OKR is needed. Thus, the optokinetic signal serves two important purposes: It effects the OKR to assist the VOR and it provides the error signal for continually fine-tuning VOR gain.
Was this article helpful?