The Journal of Neurophysiology Vol. 77 No. 3 March 1997, pp. 1099-1118
Copyright ©1997 by the American Physiological Society

Normal and Adapted Visuooculomotor Reflexes in Goldfish

Eric Marsh and Robert Baker

Department of Physiology and Neuroscience, New York University Medical Center, New York 10016

	ABSTRACT

Abstract Introduction Methods Results Discussion References

Marsh, Eric and Robert Baker. Normal and adapted visuooculomotor reflexes in goldfish. J. Neurophysiol. 77: 1099-1118. Under normal physiological conditions, whole field visual motion generally occurs in response to either active or passive self-motion. In the laboratory, selective movement of the visual surround produces an optokinetic response (OKR) that acts primarily to support the vestibuloocular reflex (VOR). During visual world motion, however, the OKR can be viewed as operating independently over frequency and amplitude ranges insufficient for vestibular activation. The goal of the present study was to characterize this isolated behavior of the OKR in goldfish as an essential step for studying central neuronal correlates of visual-vestibular interactions and the mechanisms underlying oculomotor adaptation. After presentation of either binocular sinusoidal or step visual stimuli, conjugate eye movements were elicited with an amplitude and phase profile similar to that of other vertebrates. An early and a delayed component were measured with different dynamics that could be altered independently by visual training. The ensuing visuomotor plasticity was robust and exhibited five major characteristics. First, the gain of both early and delayed components of the OKR increased >100%. Second, eye velocity decreased 0.5-2.0 s before the change in direction of stimulus velocity. Third, on lengthening the duration of a constant velocity visual stimulus (e.g., from 8 to 16 s), eye velocity decreased toward 0°/s. This behavior was correlated with the direction and period as opposed to the frequency of the visual stimulus ("period tuning"). Fourth, visual stimulus training increased VOR eye velocity with a ratio of 0.6 to 1 to that measured for the OKR. Fifth, after OKR adaptation, eye velocity consistently oscillated in a conjugate, symmetrical fashion at 2.4 Hz in the light, whereas in the dark, a rhythmical low-amplitude eye velocity occurred at the visual training frequency. We conclude that the frequency and amplitude of visual stimuli for eliciting the goldfish OKR are well suited for complementing the VOR. Unlike most mammals, OKR adaptive modifications significantly alter VOR gain, whereas the effects of VOR training are much less on OKR gain. These observations suggest that both distributed circuits and discrete neuronal populations control visuo- and vestibulomotor performance. Finally, the existence of a rhythmic, "period tuned" visuomotor behavior provides a unique opportunity to examine the neuronal mechanisms of adaptive plasticity.

	INTRODUCTION

Abstract Introduction Methods Results Discussion References

The principal phylogenetic role of eye movement in vertebrates is to stabilize images on the retina. Two individual oculomotor reflexes, vestibular and optokinetic, have been studied extensively and the brain stem pathways responsible for both behaviors appear to be widely interconnected, if not overlapping, at critical junctions like the vestibular nuclei (Cheron 1991; Keller and Precht 1979a,b; Precht 1979; Scudder and Fuchs 1992). The vestibuloocular reflex (VOR) produces appropriately directed eye movements in response to constant velocity head motion, but the amplitude is not generally either accurate or sustained largely due to the absence of maintained vestibular signals. Hence, in all species, the optokinetic response (OKR) is necessary to close the vestibular sensory-motor loop by continuously adjusting eye movements to be exactly compensatory to the motion of the visual field. Because in most animals, visual stimuli can, by and large, independently initiate and sustain eye movement in response to whole-field visual motion, the OKR should not only be thought of as a system well designed to complement, calibrate and sustain the VOR, but also as a self-contained visuooculomotor reflex.

The OKR has been investigated mostly in mammals, and notably in species with quite disparate oculomotor and visual functions [e.g., rabbit, cat, rat, and primate (Carpenter 1977; Cohen et al. 1977; Collewijn 1971; Evinger and Fuchs 1978; Fuchs 1967; Hess et al. 1985). These diverse studies have agreed on the observation that the OKR works well (i.e., autonomously) at low stimulus velocity and frequency in a range that the VOR is largely lacking in precision. In addition, unique ocular following properties have been identified in primates such as the ability to track small targets at high stimulus velocities independent of the visual surround (Fuchs 1967; Miles and Busettini 1992). However, establishing the interrelationship between the neurons and circuits sufficient for generating ocular following (i.e., pursuit) and the more common, but less demanding, optokinetic response is unclear in all vertebrates. Moreover, few criteria exist to distinguish between either separate or shared pathways responsible for the neuronal organization of the OKR and VOR reflex system in the brain stem and cerebellum.

The role of the optokinetic system in teleosts has been described as similar to other species in that it appears to function independently and support the vestibuloocular reflex to hold the environment still during head movement (Dieringer et al. 1992; Easter 1972; Pastor et al. 1992; Schairer and Bennett 1986a). However, a major difference with mammals exists because a pursuit-like following system is absent in the majority of teleosts, as reflected by the absence of a fovea, pontine nuclei, and associated ponto-cerebello-vestibular loops (Finger 1978, 1983). This difference allows for direct examination of the pure optokinetic response not affected by the pursuit system. Recent behavioral work in the goldfish has ascribed vestibulo- and visuooculomotor function to two separate sites in the hindbrain (Pastor et al. 1994b). These observations suggest that a detailed, perhaps complete, picture of the neuronal circuitry underlying the VOR and OKR might be obtained by single cell structure/function techniques in the brain stem and cerebellum. This supposition provided one compelling rationale for providing a thorough analysis of the OKR in the velocity and frequency domain including during visual adaptation to complement that existing for the VOR (Pastor et al. 1992; Schairer and Bennett 1986a,b). Only when the analysis of the two oculomotor behaviors, as well as the structure/function studies, are complete can a total understanding be achieved of the neuronal organization within the brain stem along with the cellular and molecular basis of plasticity.

All vertebrates, fish in particular, exhibit continual growth and development throughout life. Thus the oculomotor system constantly must be recalibrating VOR and OKR sensitivity to maintain proper visual function. Prior work has demonstrated a robust plasticity in VOR pathways in which visual stimuli interact with head rotations of various magnitudes (Davies and Melville Jones 1976; Ito 1993; Lisberger 1988; Miles and Eighmy 1980). The majority of these experiments have suggested that adaptations largely are acquired and stored in brain stem neurons/circuits as opposed to in the cerebellum; however, hypotheses remain that place learning either solely in the cerebellum or in both the cerebellum and brain stem (Du Lac et al. 1995; Llinas and Welsh 1993; Partsalis et al. 1995). In general, any type of motor plasticity, visuo-vestibuloocular in particular, is of interest in respect to the neuronal mechanisms responsible. Hence, teleosts also offer a unique opportunity to assess the cellular basis for plasticity in a preparation that itself is attractive for rigorously addressing the pharmacological analysis at well-identified synaptic sites (Faber et al. 1991). Therefore the current experiments were designed as the initial stage of a qualitative and quantitative measurement of the visuomotor responses in the goldfish, a representative cypriniform teleost without any retinal specializations (i.e., lacking foveation). Both monocular and binocular eye movements were studied in response to quite different basic visual stimuli to assess the phase and gain relationships while quantifying the latency and early/delayed components of the optokinetic reflex behavior. After establishing the eye velocity responses to each set of stimuli, specific features such as habituation, sensitization, and adaptation were determined independently. In addition, eye velocity in response to head movement (VOR) was compared with eye velocity generated by visual mechanisms.

The results show that the goldfish OKR responds quite comparably in both magnitude and frequency as that described in mammals (Collewijn 1981; Fuchs 1967; Lisberger et al. 1981). A major difference observed was that a rather substantive adaptation of eye velocity occurred in response to continuous presentation of visual stimuli. Interactions of visual and vestibular motor training suggested that some, but not all, gain elements might be shared in the VOR and OKR. Finally, an entrainment of eye movements specific to the training period of individual stimuli was observed, and a diverse set of experiments demonstrated this plasticity to be period, as opposed to frequency, tuning. Part of this work has appeared elsewhere in abstract form (Marsh et al. 1994; Pastor et al. 1994c).

	METHODS

Abstract Introduction Methods Results Discussion References

General procedures

Goldfish (Carassius auratus) were obtained from an authorized supplier (Hunting Creek Fisheries, Thurmont, MD) and maintained at 20°C on a 12 h light/dark cycle with the water quality monitored biweekly. The findings reported herein represents analysis of data from 35 experiments. Goldfish were restrained by fitting the mouth onto a tapered tube and holding the body, wrapped in gauze for protection, in a Plexiglas rigging lined by moistened sponges. The circular tank, 25 cm in diameter, had untextured white walls that acted as a background for the visual stimulus. A 2 cm attachment for the respiration tube, directly ahead of the fish, was the only interruption in the otherwise plain background. Aerated water, at 20°C, was passed over the gills for ventilation. After preparation for recording, fish were acclimated to the apparatus for 15-30 min before experimental intervention.

Eye movement recording

Eye movements were recorded using the scleral search coil technique with a bandwidth of 0-200 Hz and a sensitivity of 0.2°. An 80-turn, 1.6-mm diameter insulated coil (Sokymat S. A.) was sutured at two points onto the upper scleral margin, with 6.0 ophthalmic silk, under 4% lidocaine anesthesia. Obstruction of the visual field was avoided by careful suturing of the eye coils and testing for the normal range of vestibular and visual eye movements (Pastor et al. 1992). The tank was placed into the center of the magnetic field coils, and three complementary procedures were used for calibrations. First, the field coils were rotated about stationary eye coils at known angles. Second, the eye coils were rotated at known angles. Third, the long-term accuracy and reproducibility of the calibrations were verified by using the same set of eye coils on different fish in sequential experiments with the detection system settings held constant.

Visual and vestibular stimulation

Visual stimulation was provided by a planetarium that could be rotated 360° around the vertical axis at speeds ranging from 1 to 60°/s constant velocity. The planetarium projected a random light spot pattern on to the walls of the water tank. Fiber optic illumination was selected because it provided a light of uniform intensity and color that generated little heat and could be turned on/off (<5 ms) by a shutter at the source. A fiber optic cable connected the planetarium with the 20-W halogen bulb light source. A solid 2-mm fiber optic machined into a cone was fit tightly to an emitter fiber optic within the planetarium. This arrangement resulted in a uniform projection of the light spots onto the walls. The light spots (dots) subtended 1 cm on the wall of the tank with an intensity of 1.9 × 10⁹ quanta of green light per square centimeter (measured using a calibrated thermopile). The intensity of the light was 400% over the determined threshold for eliciting eye movements. The planetarium projected 360° around the tank, above and below the water level. Because the goldfish visual field was found to be 320° with 30° binocular overlap (Easter 1972), ~29° was obstructed by the tapered tube holder. As a result, ~90% of the visual field was stimulated by the planetarium. Two function generators (Exact-337 and Wavetek) were used to drive the planetarium for the visual stimuli. Sinusoids and either uni- or bidirectional velocity steps (i.e., triangles of position) were the primary waveforms (see Fig. 6). Other stimuli employed were velocity steps with either different half cycle durations (e.g., 10 vs. 4 s) or with random delays between cycles (see Fig. 9). Visual stimulation occurred at frequency ranging from 0.004 to 4 Hz spaced at octave intervals and at amplitudes of 1 to 64°/s.

View larger version (38K): [in this window] [in a new window]   FIG. 6. Adaptive modification of the OKR. A: eye velocity to left recorded at 15, 30, 50 and 60 min after training with unidirectional velocity steps at 0.14 Hz. Solid arrows point to change in early rise (OKRe). Changes in later part of delayed component are illustrated by length and direction of dashed arrows. Eye velocity to right (Con, 60 min) is shown after leftward visual training. B: control and adapted (Con, 60 min) eye velocity are shown after visual step training at 0.125 Hz. C: eye velocity response increased to sinusoidal stimulus after visual step training. D: plot of frequency vs. gain before and after training (n = 8). E: a chart showing amplitude of early response expressed as a percent of total gain before and after training (n = 10 ± SE); typical time to build-up of late component; maximum acceleration of OKRe; and average acceleration of OKRe. *Significant changes.

View larger version (37K): [in this window] [in a new window]   FIG. 9. Distinguishing between period and frequency tuning. A-C: 3 different training paradigms are illustrated, each set containing 1 control record and 2 after training. A: control eye velocity (top) in response to a bidirectional step visual stimulus with unequal half cycles of 4 and 12 s durations. After training for 3 h with 4 s/12 s visual stimulus, eye velocity was increased (middle). Period tuning was tested by extending visual stimulus symmetrically to 16 s (0.062 Hz, bottom). Dashed line, training period for all traces in this figure. B: control eye velocity response to bidirectional step stimuli with delay training paradigm (arrows show delay periods). After training, OKR gain was increased in both directions (left and right), but asymmetrically (middle). Period tuning was tested by removing random delay and doubling stimulus cycle to 16 s (0.062 Hz, bottom trace). C: control eye velocity response to a 0.062-Hz visual stimulus (top). OKR gain increased after 8 h of training to a 0.062-Hz bidirectional step visual stimulus in unrestrained condition (middle). Extending training period to 0.031 Hz resulted in an OKR response that exhibited tuning (bottom).

Vestibular stimulation was provided by a servo-controlled (14 ft-lb) motor attached directly to a rate table that could rotate 360° around the vertical axis at a rate of 1 to 48°/s constant velocity. The tank placement resulted in the vertical axis of the table passing through the head at the level of the semicircular canals. Sinusoids and velocity steps were employed at a frequency range of 0.031 to 8 Hz at octave intervals and velocities ranging from 4 to 48°/s. The planetarium and vestibular rate table were interfaced by means of a dual channel digital wave form generator (Exact-337) that could independently control the phase and amplitude relationship of the two stimuli.

Visuomotor training consisted of presenting one of the visual stimulus waveforms for 1-3 h and testing at 15- to 30-min intervals. Unrestrained goldfish (i.e., freely swimming) were used in one series of experiments to test for visual plasticity under "natural" visual and vestibular input conditions. In the free swimming experiments, the tank was similar to the normal experimental setup, but without any obstructions of the visual world. To avoid visual interactions with other light sources in the essentially dark laboratory, a nonreflectant black box covered the entire preparation.

Data analysis

Eye and head position and planetarium velocity signals were recorded on a digital audio tape (Cygnus Technology) at 0-2.5 kHz band-pass for off-line analysis. For on-line observation of eye and head velocity, the eye and head position signals were differentiated using analog electronics (9 ms time constant) and displayed on either a storage oscilloscope, a digital oscilloscope (Data 6000), or on a Macintosh 840AV computer using GW Instruments software. Gains of the eye velocity response to either visual or vestibular stimuli were obtained by two methods, and their comparison ensured accuracy of measurement. Gain was calculated as peak-to-peak eye velocity/stimulus velocity. One method (Fig. 1A) was to alter manually the amplitude of the stimulus velocity traces (either head or planetarium) such that the stimulus waveform superimposed on the outline of the eye velocity waveform (see dotted line in 1A, ). The original and altered amplitudes were divided resulting in the gain measurement. The second method (Fig. 1B) was computer analysis using a program (GW Instruments software) that allowed accurate unbiased measurement of the amplitude of the eye movements. As shown in Fig. 1, the slow phase eye velocity was not necessarily smooth, therefore to calculate slow phase eye velocity, the computer selected the portion between the fast phases. This was accomplished by comparing the slope of the waveform of the slow and fast phases. The intervening slow phase eye velocity (see Fig. 1B, insets) then was averaged. Because the overall gain was of importance, the computer averaged the slow phase eye velocity during the sustained portion of each half cycle and then the result was divided by the stimulus velocity. The two methods of measuring gain were essentially identical, with <5% difference between the procedures (e.g., in Fig. 1B, the gain measured by the computer was 0.29 and by manually scaling, 0.3). Phase measurements of eye velocity during sinusoidal visual stimuli was determined by comparing the 0 crossing of the eye and stimulus velocity for a minimum of four cycles, with the average time difference used for the calculation of the phase lag (see Fig. 2A, arrows illustrate the phase lag by pointing to the peaks instead of the 0 crossing).

View larger version (35K): [in this window] [in a new window]   FIG. 1. Vestibuloocular reflex (VOR) and optokinetic response (OKR) performance of a naive goldfish. A: eye velocity () in response to sinusoidal (31.4°/s peak velocity at 0.125 Hz) vestibular () and visual (planetarium, L) stimuli. Dotted line superimposed on lower () depicts 1 way in which gain was measured (see RESULTS). B: left eye velocity (E) during 1 cycle of ±16°/s triangle of position visual stimulus (L). Early and delayed components of eye velocity step are indicated by arrows. Insets: enlargements of slow phase eye velocity regions between fast phases. Dashed line, average slow phase eye velocity computed for each section.

View larger version (19K): [in this window] [in a new window]   FIG. 2. Bode plot analysis of OKR. A-D: eye velocity response to 1-, 0.5-, 0.125-, and 0.062-Hz sinusoidal visual stimuli, respectively. Arrows show phase shift at 1 and 0.5 Hz. E and F: Bode plots for OKR. Gain (E) and phase (F) vs. frequency plot for 10 different experiments. Average of all experiments is shown by diamonds and connecting line.

The latency of the optokinetic eye velocity response was measured using unidirectional velocity steps as the visual stimulus. Latency of individual responses was measured and averaged (8 steps per animal were averaged, n = 11 goldfish). Through out the paper, n is for the number of animals used to determine the averaged value, unless otherwise noted. Onset of eye velocity after the visual stimulus was determined by differentiation of eye velocity with a five-point differentiation algorithm followed by a smoothing algorithm on the computer (GWI software). The first 0 crossing of the acceleration profile after the stimulus onset was considered as the initiation of eye movement. Early and dynamic components of the OKR and VOR were measured by capturing individual cycles and either overlying the waveforms or averaging those cycles without fast phases in the first 400 ms. Early component amplitude, acceleration, and rise time were measured with the same procedures. The time of the delayed component was measured for each cycle in an experiment and means and standard errors were calculated. Statistical measures used to check for significance were calculations of means, standard error/deviation, and Student's t-test. In all figures, 0°/s of eye velocity is in the center of each trace. Care was taken to select traces in which no velocity bias was present and the eye velocity responses were symmetrical.

	RESULTS

Abstract Introduction Methods Results Discussion References

OKR response to sinusoidal stimuli

The optokinetic response was characterized by projecting a random pattern of light spots uniformly on to the wall of the tank containing the restrained goldfish. Representative eye velocity responses to sinusoidal stimuli are illustrated in Fig. 1A for 0.125 Hz, and Fig. 2, A-D, for 1.0, 0.5, 0.125, and 0.062 Hz, respectively. Sinusoidal and velocity step wave forms were used to compute the velocity, frequency, latency, and storage components of the OKR. Measurements were obtained over many cycles of raw data, as illustrated for sinusoidal stimuli in Fig. 1A. Phase and gain (eye velocity/stimulus velocity) were described using a Bode plot of the OKR in the frequency range from 0.031 to 4 Hz at 31.4°/s peak-to-peak planetarium velocity (Fig. 2, n = 10). Phase and gain were calculated from the average of the total raw data and not the single cycles presented in Fig. 2. The eye velocity traces (Fig. 2, A-D) show that OKR gain decreased and phase lagged with increasing frequency. Bode plots showed the average gain to be 0.65 at 0.031 Hz and to decrease at 0.07 per octave reaching 0.15 at 4 Hz (Fig. 2E). Phase was unaltered until 0.5 Hz and then changed rapidly at 83° per octave, reaching 300° at 4 Hz (Fig. 2F). The quantitative results are consistent with previous observations in goldfish (Pastor et al. 1992; Schairer and Bennett 1986b).

OKR response to visual step stimuli

Visual velocity step stimuli of varying durations were used to gather information about the dynamics of the visuomotor responses. Typical eye velocity responses to either, short, 4 s, or long, 200 s, velocity step (triangle of position) visual stimuli are shown in Figs. 1B and 3. In response to a prolonged visual step stimulus that jumped from 0 to 32°/s, eye velocity exhibited a small early rise (arrows in Fig. 3A, eye velocity traces) followed by a steady increase in eye velocity for 80 s. The early rise (OKRe) was a consistent feature of the optokinetic response to step stimuli (Lisberger et al. 1981; Miles and Busettini 1992). OKRe could be defined as that part of the step response initiated after onset of the stimulus and continuing for ~400 ms at which time eye acceleration returned to 0 deg·s¹·s¹ (seeFig. 1B).

View larger version (45K): [in this window] [in a new window]   FIG. 3. Eye velocity responses to prolonged visual step stimuli. A: left and right eye position (E, E), eye velocity (), and difference in eye positions (EDiff; arrows show 0°/s eye velocity) are shown in response to ±32°/s visual velocity step (L). Solid lines (in RE) demarcate underlying spontaneous saccadic rhythm. Arrows indicate early component. B: early component of eye and planetarium velocity are superimposed for 8 consecutive velocity steps. Arrows indicate a latency of 76 ms.

From the end of the early component at 500 ms until the completion of the stimulus cycle, a constant velocity visual stimulus produced a compensatory slow phase eye velocity defined as the delayed component (OKRd; Fig. 1B) (Cohen et al. 1977; Lisberger et al. 1981; Miles and Busettini 1992). Together the early and delayed eye velocity responses generated an eye velocity that was less than, but approached, the 32°/s planetarium velocity stimulus (Fig. 3A). In naive animals, eye velocity built-up to, and saturated between, 24-32°/s even after presentation of 40-64°/s stimulus velocity for 5 min. The time required to build-up to maximum eye velocity ranged between 40 and 70 s with a mean of45.9 ± 6.2 s (mean ± SE, n = 6).

Eye movements in response to planetarium movement were conjugate (Fig. 3A, right). Eye velocity was typically symmetrical as shown by the 0 velocity difference (E_Diff) between the LE and RE position traces. The slow phase portion of eye movements (see arrow in Fig. 3A) was also of similar amplitude as indicated by the essentially flat E_Diff trace, but fast phases consistently exhibited a temporal-nasal overshoot (interruptions in E_Diff). Temporal-nasal overshoot of fast phases also was described previously in the VOR (Pastor et al. 1992). Because differences in slow phase amplitude in the two eyes were either absent or minimal with the presentation of binocular stimuli, only representative records from one eye are shown in Figs. 4-10. Eye position traces in Fig. 3A show that a rhythm identical to that seen during spontaneous saccade generation underlies the OKN during the prolonged visual stimulus presentation (solid lines). Because the frequency of OKN fast phases and scanning saccades appear independent and always nonoverlapping, a downstream convergence of these two signals is suggested to occur on the premotor circuitry controlling oculomotor nuclei.

View larger version (32K): [in this window] [in a new window]   FIG. 4. Frequency and amplitude characteristics of OKR in response to velocity steps and trapezoids. A: superimposition of eye velocity responses to 4 visual velocity steps in which stimulus velocity ranged from 8 to 32°/s. Inset: illustration that early components of 3 responses were of similar amplitude. Magnitude of delayed component eye velocity saturated at 12°/s. B: eye velocity responses to a velocity trapezoid to left are shown with constant velocity to left. Inset: 4 superimposed steps shown. C: plot of gain vs. frequency for triangle of position stimuli (n = 9). D: graph of gain vs. amplitude at 3 representative frequencies (0.125, 0.062, and 0.031 Hz: n = 9). E: plot of amplitude of early component against stimulus amplitude (n = 9). Fast phases were truncated to make viewing figure easier.

The latency, rise time, and acceleration of OKRe was determined by employing a repetitive, short-duration (4 s) constant visual velocity step stimulus (0.125 Hz 0-16°/s). Eye velocity exhibited an early rise to 60 ± 10% (mean ± SE, n = 9) of the maximum velocity (Fig. 4, A and B). The latency for the response to the visual step stimulus was76 ± 3.45 ms (n = 11) (arrows in Fig. 3B). Both the variation in latency and rise time in consecutive trials are shown in the superimposed eye velocity records (n = 8 cycles). The time to peak velocity of the early component (the rise time) was 319.1 ± 14.5 ms (n = 9). The acceleration profile of the early component had a mean max acceleration of 59.1 ± 3.4 deg·s¹·s¹ (n = 9) and a mean average acceleration of 31.4 ± 2.0 deg·s¹·s¹ (n = 9) as summarized in the table (Fig. 6).

The amplitude of the early (OKRe) and delayed (OKRd) components were measured in response to visual step stimuli ranging from 2 to 48°/s (Fig. 4, A and B). Using either bidirectional (Fig. 4A) or unidirectional visual step stimuli (Fig. 4B) the amplitude of OKRe appeared to be independent of stimulus amplitude (Fig. 4, A and B, inset). Plotting the mean amplitude of the early component against stimulus velocity resulted in a straight line (Fig. 4E) for stimulus velocity >4°/s (n = 9). Less than 4°/s, the amplitude of the early component equaled that of the visual step (not illustrated).

During OKRd, 0.125-Hz velocity step stimuli of <6.4°/s always resulted in an eye velocity response profile that matched stimulus velocity (a linear response), and often, with some animals, the response was linear at even higher velocities. In these cases, eye velocity was saturated and did not increase to stimulus velocity >12°/s (n = 9) (Fig. 4, A and B). The maximum eye velocity generated in naive animals ranged from 5 to 12°/s at 0.125 Hz (Fig. 4, A-C), whereas 24-32°/s was eventually reached at longer intervals, e.g., 1/200th Hz in Fig. 3A. Eye velocity achieved during velocity step stimuli with different durations is plotted in Fig. 4D for stimulus amplitudes ranging from 4 to 48°/s in eight experiments. The increase in eye velocity with longer periods of stimulus presentation demonstrates a role of velocity storage in generating the delayed OKR component.

Optokinetic afternystagmus

The time course of optokinetic afternystagmus (OKAN) decay was measured and used to characterize the neural processing responsible for the build-up of eye velocity during OKRd. OKAN has been postulated to represent the discharge of a central velocity storage integrator (Cohen et al. 1977; Collewijn et al. 1980). After build-up of OKRd, the light was extinguished when eye velocity had approximated the stimulus velocity of 32°/s for ~30 s. Eye velocity decayed in the direction of the stimulus to 0°/s (Fig. 5, A and inset in B). The "time constant" of decay was measured as the entire duration of the decay because the fall-off, as determined by best fit algorithms was essentially linear (Collewijn et al. 1980) (GW Instruments curve fit algorithm with smallest error coefficient chosen). Mean decay time was 77 ± 6.2 s (n = 8) for the initial OKAN measurement (Fig. 5, A and B). When OKAN testing was repeated, each subsequent trial resulted in a reduction of the decay time. After eight trials, the decay plateaued at a mean of 20 ± 6 s (n = 6; Fig. 5, A and B). The decay time for each animal remained unaltered throughout any subsequent experimental intervention. This type of OKN behavior called, habituation, also has been observed in mammals (Cohen et al. 1992; Collewijn et al. 1980; Jager and Henn 1981).

View larger version (28K): [in this window] [in a new window]   FIG. 5. Optokinetic afternystagmus (OKAN). A: build-up and decay of eye velocity to a continuous 32°/s visual stimulus to the right. Decay in trials 1-4 and 8 of raw data are shown in A, and fitted lines shown in B were computed with second order polynomials. Trial number is indicated on left. B: decay time (n = 8) is plotted vs. trial number illustrating OKAN habituation.

Visuomotor adaptation

Prolonged presentation of visual stimuli as either unidirectional (Fig. 6A) or bidirectional step waveforms (Fig. 6B) were used to induce modification of the OKR response (Marsh et al. 1994; Pastor et al. 1994c). Visuomotor training was found to produce three major changes in the previously described eye velocity profiles: an increase in amplitude, a patterning of eye velocity tuned to stimulus duration, and an increase in vestibuloocular amplitude.

The amplitude of eye velocity was increased during visual stimulus training paradigms in all experiments. The increase in gain consisted of three distinct but related changes. As early as 15 min and continuing for 60 min, the amplitude of the delayed optokinetic response (OKRd) increased until eye velocity equaled planetarium velocity (i.e., gain = 1.0; Fig. 6A). For example, an OKRd gain before training at 0.125 Hz of 0.41 ± 0.03 (n = 8) was increased significantly to 0.77 ± 0.06 (n = 8; P < 0.001, Student's t-test) after 60 min. The increase in gain produced at the training frequency always appeared to be the most significant. For instance, visuomotor plasticity extrapolated closely to frequency one octave away from the training frequency (e.g., 0.063 Hz, control vs. adapted gain; con = 0.51 vs. adp = 0.83, P < 0.001); but changes were smaller but still significant at other frequencies tested (Fig. 6D).

A second modification in eye velocity that contributed to the overall change in OKR gain (as measured above) was an increase in amplitude of the early component (OKRe) (Fig. 6, A, B, and E). The early component increased significantly from 60.9% of the control response at 0.125 Hz to 77.5% after training (P < 0.005). A third change contributing to the overall OKRd amplitude was a decrease in time required for OKRd build-up of stored eye velocity. The time course shortened from 3.2 ± 0.16 s to 1.3 ± 0.07 s, and this significant change is illustrated during the presentation of a 0.125 Hz constant velocity step stimulus in Fig. 6, A, B, and E (P < 0.005). All changes in OKRe and OKRd components of eye velocity were seen after either unidirectional (Fig. 6A) or bidirectional step (Fig. 6B) presentation, and this training always translated to an increase in eye velocity to sinusoidal waveforms of visual stimuli (Fig. 6C). Significant changes were not observed in either the maximum or average acceleration profile of OKRe before or after training (P < 0.25) (Fig. 6E).

Period tuning

As early as 15 min after the onset of velocity step visual training, eye velocity began to decrease before the change in visual stimulus direction as indicated by the slope of the arrows in Fig. 6, A and B. The decrease in eye velocity was first noted after 15 min and, by extrapolation, began ~1 s before the change in stimulus direction at a rate of 1.1 deg·s¹·s¹ (Fig. 6A). With continued training, the rate increased to 3.2 deg·s¹·s¹, and the onset appeared earlier (e.g., 2.5 s before the turn-around at 60 min in Fig. 6A). The decrease was independent of training duration but not training period. The relationship between training period and the onset of the predictive eye velocity was linear (points could be fit to a line; y = 17.1 + 203.4x, R² = 0.998). As stimulus duration increased, the decrease in eye velocity occurred later, but the percentage of the cycle decreased. For a 0.125 Hz cycle, eye velocity began to decrease 1.7 ± 0.1 s before the change in stimulus direction. This interval represents 42.7% of the stimulus cycle. For 0.062 and 0.031 Hz, the times were 2.35 ± 0.1 s and 3.8 ± 0.2 s, respectively, but the percentage of the cycle decreased from 29.3% at 1/16 Hz to 23.8% at 1/32 Hz. When the apparent stimulus frequency was lowered by lengthening the period after 1 h of training (Fig. 8A), eye velocity continued to decrease toward, and often reached, 0°/s eye velocity. Hence, this phenomenon consisted of both a "predictive" decrease in eye velocity during training and a "continued" decrease in eye velocity toward 0°/s when the training stimulus was extended. The predictive-like decrease in eye velocity specific to, and occurring before, the end of the training period was determined to be period as opposed to frequency tuning (see later section).

View larger version (44K): [in this window] [in a new window]   FIG. 8. Eye velocity after visual stimulus training. A: eye velocity records after 1 h training at 0.125 Hz with a triangle of position visual stimulus. Gains increased and "predictive" eye velocity responses are illustrated by inclined arrows. B: control eye velocity records at 0.062 Hz. The 0.125 Hz cycle is superimposed (dashed line). C: eye velocity at 0.062 Hz after 1 h of training at 0.125 Hz. Dashed line, training period. Eye velocity decrease toward 0 after period extension is evident. D: periodic eye velocity (arrows) recorded in dark after training to 0.125 Hz. E: eye velocity in response to a 7-s velocity trapezoid visual stimulus to left after training for 1 h. F: extension of period to 12 s to left. Dashed line, training period. G: after training, eye velocity oscillates in dark only to left and at frequency of training stimulus.

Period tuning was detected by training at periods that ranged from 2 s (0.25 Hz) to 128 s (0.0039 Hz). Training periods <2 s never resulted in detectable period tuning. Tuning also was seen after training to a velocity, 4°/s, that did not induce a change in the eye velocity performance (data not shown). These two points, as well as the fact that eye velocity can be maintained at high velocity for an indefinite period of time (see Figs. 3A and 5A) argue that the tuning is not due to the inability of the animal to maintain the increased eye velocity. In the presence of scanning saccadic eye movements, period tuning was maintained for 40-50 min after training in stationary light conditions and for 60-70 min in the dark. Much shorter extinction times of <20 min were found when the initial training was followed immediately by conditioning to another period. For example, three consecutive periods werelearned in one experiment at 1 h intervals (e.g., 0.125-, 0.063-,and 0.031-Hz velocity steps).

OKR and VOR plasticity

In naive goldfish, normal vestibuloocular reflex (VOR) gain (eye velocity/head velocity) averaged 0.85 ± 0.04 (Pastor et al. 1992, 1994c) (Fig. 7, A and E). In response to a single velocity step vestibular stimulus, eye velocity exhibited an early dynamic component (VORd) that began at a latency of 20 ms and attained stimulus velocity within 150 ms (small arrow in Fig. 7A). The abrupt rise to compensatory eye velocity was followed by a sustained velocity component (VORs) (Fig. 7A, connected arrows) that was shown to last for 4-6 s before decaying to baseline (Pastor et al. 1992). After interactive visual-vestibular training paradigms, VOR plasticity was observed in both the dynamic (including latency) and sustained eye velocity components.

View larger version (35K): [in this window] [in a new window]   FIG. 7. Effect of OKR modification on VOR gain. A: control eye velocity response to a step in head velocity in dark (: VOR gain = 1.0). Arrows indicate dynamic (Dyn) and sustained (Sus) components, respectively. B: eye velocity in response to steps of head velocity in dark after 1 h of bidirectional velocity step visual training at 0.125 Hz. C: eye velocity response to a bidirectional step in head velocity in dark 1 h after unidirectional velocity step visual training to left at 0.125 Hz. D: dynamic component of VOR eye velocity was averaged after both bidirectional step (Adp) and unidirectional step (uniD) visual training. Large arrows indicate latency for adapted eye velocity to diverge from control eye velocity. Small arrows show latency of VOR. E: graph of VOR sustained gain versus frequency before and after bidirectional velocity step visual training.

Training with only a visual stimulus was observed to produce changes in the untrained VOR (Fig. 7) (Pastor et al. 1994c; Schairer and Bennett 1986b). After 1 h of training with a velocity step visual stimulus at 0.125 Hz and 24°/s, the VOR was tested at 0.125 Hz ± 16°/s (Fig. 7B). Control and adapted VORs averaged 0.85 ± 0.04 and 1.3 ± 0.05, respectively (P < 0.001, n = 9; Fig. 7E). The increase in VORs gain was significant over a range of frequency between 0.031 and 1.0 Hz (at 0.063 Hz, P < 0.001, n = 6; 0.25 Hz, P < 0.001, n = 9; and 0.5 Hz, P < 0.001, n = 9). The increase in VORd gain occurred at a latency ranging from 40 to 80 ms (mean was 55.5 ± 4.0 ms, n = 8) after the onset of the vestibular stimulus as opposed to the earliest possible latency as observed during VOR training (Fig. 7D, solid arrow, for traces to left and right labeled "Adp") (Pastor et al. 1992, 1994c). However, eye velocity increased during both the dynamic and sustained components of the VOR response profile (Fig. 7B). Period tuning was detected rarely in the VOR even after lengthening the step vestibular stimulus and several hours of either visual or visual-vestibular training (not illustrated). When the tuning occurred in the VOR, it was often difficult to distinguish between a shortened vestibular storage time constant and a true predictive decrease in eye velocity.

Unidirectional visual step training also resulted in a VOR gain increase, however, eye velocity only increased in the direction of the training. In Fig. 7C, two continuous vestibular cycles to the right and left are compared showing that, in the direction opposite to visuomotor training, VORs gain remained unchanged. Gain changes with unidirectional velocity step training, like the use of bidirectional steps, also occurred 55.5 ms after the onset of the stimulus (Fig. 7D, traces labeled UniD). The increase in VOR gain after visuomotor training suggests overlap in modifiable gain elements responsible for the horizontal VOR and OKR.

Tuning: period or frequency?

Period tuning was distinguished from frequency tuning by employing various training paradigms that showed eye velocity to be best associated with the period of the stimulus, not the frequency. In Fig. 8A, OKRd gain was increased from 0.4 to 0.9 after 1 h of training at 0.125 Hz with velocity step stimuli. After adaptation, predictive eye velocity responses (as in Fig. 6B) varying about a peak-to-peak amplitude of ±15°/s were apparent at the training period (Fig. 8A, arrows). Before the training at 0.125 Hz, the eye velocity response at 0.063 Hz was normal (Fig. 8B); however, after training with a visual stimulus, the cycle period was extended from 4 to 8 s (0.063 Hz) and eye velocity continued to decrease toward 0°/s (Fig. 8C). Eye velocity was calculated to decrease at a mean rate of 1.8 deg·s¹·s¹ (range of 1.1-4 deg·s¹·s¹), which varied both within individual trials and between animals. After training for 1 h, eye velocity oscillated in the dark at a frequency approximating the visual training period (0.13 Hz in Fig. 8D). This oscillation was observable in the dark in over 50% of the experiments. Eye velocity changes were symmetrical ~0°/s as indicated by the dotted line.

When an unidirectional velocity step visual stimulus was used in the training paradigm either to the left or right, eye velocity exhibited predictive changes only in the trained direction irrespective of whether uni- or bidirectional steps were presented (Fig. 8, E-G). For instance, after 1 h of training with a 16°/s unidirectional velocity step to the left, eye velocity increased from 6°/s (i.e., as in Fig. 6A) to 14°/s peak eye velocity (Fig. 8E). This unidirectional training induced a "predictive" fall-off in eye velocity before the planetarium step to 0°/s (Fig. 8E). When the visual stimulus period was extended from 7 to 12 s, eye velocity continued to decline toward 0°/s (Fig. 8F). Eye velocity oscillated in the dark at the training period employed, but unlike visual step conditioning, only in the direction of the stimulus(Fig. 8G).

Another stimulus waveform that was used to distinguish this type of visual plasticity as either frequency or period tuning was a repetitive velocity step set to different half cycle widths (4 and 12 s in Fig. 9A). When these "asymmetric" stimuli were first presented, OKRe and OKRd eye velocity were appropriately normal for the length of time of each half cycle (Fig. 9A, top). After 3 h of training, eye velocity increased in amplitude to 16°/s in both the 4 s and 12 s directions (Fig. 9A, middle). When the 4/12 s cycle was extended to a temporally symmetrical 16-s cycle, 0.031 Hz (Fig. 9A, bottom), eye velocity was 16°/s for the 4-s training direction then decreased to 10°/s for the remainder of the cycle. After 12 s, in the other direction, eye velocity decreased toward 0. Six different asymmetric training paradigms were measured, ranging from 2/12 to 6/8 s, and in every case, the results suggested that periods of visual training and not an average frequency were acquired.

The most defining stimulus waveform used to distinguish period from frequency tuning was a velocity step with a random delay inserted between each cycle (Fig. 9B, top and middle). The stimulus was presented at either 0.125 Hz or 0.063 Hz with the planetarium stationary at 0°/s for 2-8 s between each cycle (see arrows, Fig. 9B, top and middle). Control eye velocity, during both half cycles was asymmetric (4 and 8°/s; see Fig. 9B, top) and increased after 3 h of training with the random delay paradigm to 12 and 16°/s, respectively (Fig. 9B, middle). This random delay stimulus demonstrated that a repetitive stimulus cycle was not required for period tuning. The asymmetry in eye velocity was due to eye velocity starting from rest at stimulus onset but from a higher velocity during the directional change. The predictive decline at the end of the stimulus cycle was present in both directions but more obvious to the right (Fig. 9B, middle). A 32-s cycle (0.031 Hz) stimulus showed predictive eye velocity changes (Fig. 9B, bottom) similar to those seen with the symmetrical velocity step visual stimulus; however, the asymmetry in amplitude persisted (arrow in Fig. 9B, bottom).

Because predictive eye velocity changes occurred with a randomly displayed stimulus and a stimulus of different durations, the important parameter for tuning appears to be the length of time (i.e., the period) of stimulus presentation and not the frequency. Hence, these data suggest that the central pattern generator was not simply acquiring a repetitive frequency, but a completely patterned stimulus, which, in all cases, exhibited a definite length of time (i.e., a temporal encoder).

Adaptation in freely swimming goldfish

To establish that the conditioning visuomotor paradigm was independent of head restraint and inducible in the presence of angular VOR, animals were trained while freely swimming in the observation tank (Fig. 9C). Eye velocity measurements were obtained before and after visual training with the head restrained. The training stimulus utilized was 0.062-Hz, 16-s period, visual step, ±24°/s. Presentation of velocity step visual stimuli (0.031-0.125 Hz) produced a distinct whole body optomotor response for 1 h (eye movements were observed but not measured). Within 3 h, body motion had ceased and the animals gently rocked back and forth at the training frequency (not illustrated). In all experiments, OKRe and OKRd eye velocity (Fig. 9C, top and middle) after training was similar to the changes observed in those animals trained with head restrained. OKR gain increased from 0.6 to 1.0 and eye velocity decreased at the end of the stimulus cycle. Period tuning appeared little different from that observed in restrained animals (Fig. 9C, middle and bottom). VOR gain also was increased. These data suggest that visuomotor plasticity is a natural adaptive response to environmental manipulation.

Oscillations in eye velocity

After completion of visuomotor training using any combination of experimental paradigms, a low-amplitude eye velocity was generated at the training period in the dark (Figs. 10D and 8, D and G). Another type of oscillation became prominent in all experiments after visuomotor training. This oscillation was fixed at a frequency of 2.4 Hz (0.114SE, n = 15) and was extinguished entirely when the lights were turned off (Fig. 10D). In naive healthy fish, oscillations in eye movements in the light or dark during the normal rhythm of the spontaneous saccades were not found before training (Fig. 10, A and B). The 2.4-Hz oscillation was measured by recording the time between peaks of multiple cycles and calculating individual cycle frequency (, Fig. 10C). Averaging the eye velocity traces in the light by triggering off the fast phases showed that the 2.4-Hz oscillation clearly was constant during saccade free intervals (_Ave, Fig. 10, C and D). This suggests that the eye velocity oscillation and saccadic generator are linked together; however, whether saccades act to reset, or are just in phase with, the oscillations was not established. To determine if the 2.4-Hz oscillation was conjugate and symmetrical, right and left eye positions were subtracted in the light and dark after training (E_diff, Fig. 10, C and D). The resultant eye position traces canceled out suggesting that the oscillation might be generated by a single central source.

View larger version (23K): [in this window] [in a new window]   FIG. 10. Oscillations in eye velocity. A and B: control eye position and velocity records are shown in light and in dark, respectively. Oscillations were not found in either condition. C and D: eye position, difference in eye position, and velocity records after training at 0.125 Hz with a bidirectional velocity step. Solid arrows show oscillations. Average eye velocity traces (n = 12), obtained by triggering of fast phases, illustrate frequency of oscillation. D: after adaptation, 2.4 Hz oscillation is not present in dark (large arrow), but eye velocity oscillated at approximately training frequency (small arrows, see Fig. 8).

	DISCUSSION

Abstract Introduction Methods Results Discussion References

A broad quantitative analysis of the normal and adapted optokinetic response in a representative teleost possessing robust visual and vestibular reflexes was performed with consideration of three main objectives. One goal was to establish the relationships between the properties of the optokinetic system in different species. A second purpose was to explore the unique range of plasticity present in the goldfish visuomotor system (Pastor et al. 1992; Schairer 1980). The third objective was to record the interactions between the visuo- and vestibulomotor system over a wide range of oculomotor dynamics. An additional advantage in determining the dynamic profile of visuooculomotor performance in goldfish was to expedite the structure/function analysis of brain stem and cerebellar neurons.

Characteristics of the goldfish optokinetic system

Binocular presentation of either sinusoidal or velocity step visual stimuli elicited symmetrical, conjugate eye movements (Fig. 3). Quantification of the OKR was performed initially by generating Bode plots with frequencies ranging from 0.031 to 4 Hz. The phase and gain profiles generated in this study were similar to prior measurements obtained in goldfish (Marsh et al. 1994; Schairer and Bennett 1986b). Gain and phase responses also were matched to those in other species, with the major differences largely seen in the maximum eye velocity attained in each particular animal; rabbit (Collewijn 1969, 1971), cat (Evinger and Fuchs 1978; Godaux et al. 1983a; Maioli and Precht 1984), rat (Hess et al. 1985; Tempia et al. 1991), and primate (Cohen et al. 1977; Lisberger et al. 1981; Robinson 1968). Bode plot similarity corroborates the viewpoint that the major function of the optokinetic system has been conserved phylogenetically. The primary evolutionary focus appears to have been intent on complementing VOR dynamics (Baarsma and Collewijn 1974; Collewijn 1989; Robinson 1975; Walls 1962), with the OKR working in a range in which the VOR is inadequate. The differences in the eye velocity that the various species can achieve could be due to a number of variables including differences in either the visual system, the eye plant, or central circuitry.

Using visual step stimuli, dynamics, other than phase and gain, were quantified. Normal OKR performance consisted of a 4-12°/s early step in eye velocity and a maximum eye velocity of 24-32°/s, which could be reached after 80 s with prolonged constant velocity visual stimulation. The initial jump in eye velocity was termed the OKR early component and has been observed in all animals exhibiting an OKR (Cohen et al. 1977; Collewijn 1969; Fuchs 1967; Godaux et al. 1983a; Hess et al. 1985). As with the sinusoidal OKR response, the amplitude of the step response in goldfish was different from other vertebrates with the early jump being larger than for rabbits (Collewijn 1969) but smaller than for cats, rats, or monkeys (Cohen et al. 1977; Fuchs 1967; Godaux et al. 1983a; Hess et al. 1985). After the early component, eye velocity continued to increase with a time course of 40-100 s, reaching a maximum velocity, which could be maintained indefinitely. This gradual rise in eye velocity was termed the delayed component (Miles and Busettini 1992), and in goldfish, the OKRd exhibited a linear time course. Whenever the visual stimulus was extinguished, after build-up of eye velocity, OKAN decay also followed a linear time course of 20-100 s (Fig. 4). In goldfish, both rise of the delayed component and OKAN appeared essentially linear and is similar to that described in rabbit (Collewijn et al. 1980). Curve fitting algorithms confirmed that all the time courses appeared linear independent of either the animal or state of habituation. Other authors have described this build-up/decay as being exponential (Cohen et al. 1977; Raphan et al. 1979). The linearity seen in goldfish and rabbit could be due to the lower maximum eye velocity in these animals compared with that of monkey, cat, and humans. In fact, in the lower eye velocity trials in monkeys, the decay was also essentially linear (Cohen et al. 1977). Therefore, the oculomotor system may not simply follow either linear or exponential dynamic characteristics, but rather employ a more complex algorithm for visuomotor signaling.

Conjugate eye movements

The response to both step and sinusoidal visual stimuli was conjugate and symmetrical. Goldfish eye movements were reported not to be symmetrical in response to lower velocity visual stimuli (<6°/s) (Dieringer et al. 1992). This finding also was noted at even lower eye velocity (<4°/s) (Easter 1972). However, the current measurements and other work supports the finding that conjugate eye movements seem likely to be the normal behavior (McElligott et al. 1995; Michnovicz and Bennett 1983; Pastor et al. 1994c; Schairer and Bennett 1986b). In a few cases, there was a difference in the slow phase velocity of the two eyes at <6°/s, but this was the exception not the rule and did not influence the analysis because quantification was essentially at higher eye velocities that were always symmetrical. There was, consistently, a slight nasal-temporal asymmetry in the amplitude of the early component (OKRe) but not in the sustained (OKRd). The asymmetry found in the OKR was far less significant then the asymmetry in nasal-temporal dynamics found after vestibular stimulation (VORd) (Pastor et al. 1990, 1992). We propose that visual signals may use different pathways that, for example, largely are effected through Area II [a presumed structural and physiological homolog to the mammalian prepositus (Delgado-Garcia et al. 1989; McCrea 1988)]. These neural signals could project directly to the abducens motoneurons and internuclear neurons thereby bypassing the vestibular nuclei. OKR conjugacy also implies that the visuomotor signal should be distributed equally to both abducens motoneurons and internuclear neurons, irrespective of the visual premotor source. In the VOR, second order vestibular signals project directly onto both the medial and lateral rectus motoneurons as well as projecting to Area II (Pastor et al. 1994b). This difference in synaptic arrangement, including efficacy, possibly leads to the nasal overshoot in VORd eye velocity.

Latency of OKR

The latency of the early OKR component was found to be 76 ± 30 ms (Fig. 3, C and D) and as such is similar to that reported in other species (Collewijn 1972b; Evinger and Fuchs 1978; Fuchs 1967). As in mammals variability in latency was large and even occurred when the eye velocity records chosen all began from 0°/s of eye velocity (data analysis not illustrated). In addition, there was no correlation between eye position and latency excluding eye position in the orbit as a contributing factor. Therefore, the variation must be due to neural processing delays [e.g., retinal processing by the direction selective ganglion cells (Oyster 1968)].

Velocity storage and habituation in OKR

OKRd is believed to represent the central storage of eye velocity in circuits exhibiting long time constants. Shorter time constant direct pathways produce the VORd and OKRe and simultaneously provide the input to this hypothesized storage integrator (Cohen et al. 1977; Collewijn 1972a; Lisberger et al. 1981; Raphan et al. 1979). The neuronal site of storage has been suggested to reside in either the prepositius nucleus or the vestibular nucleus, but, in either case, the integrity of storage may depend on neural loops between the two areas (Galiana and Outerbridge 1984; Godaux et al. 1993; Mettens et al. 1994). In the goldfish, two discrete sets of neurons were found in the caudal ventral hindbrain that had properties similar to the prepositus nucleus of mammals (Pastor et al. 1994b). Reversible lidocaine inactivation of the most rostral subgroup in the goldfish (Area II) supports the view that eye velocity storage is generated and stored in this presumed homologue to the prepositus nucleus (Pastor et al. 1994b).

Not only has velocity storage been thought to be represented by OKRd, but also by OKAN. Using these two behaviors as markers, velocity storage in goldfish exhibited significant habituation on repetitive presentation of the visual stimulus. The habituation seen in goldfish was similar in time course and in amplitude to that found in previous studies of either post rotary nystagmus or OKAN in both rabbits and primates (Jager and Henn 1981; Kleinschmidt and Collewijn 1975). Other studies in primates showed less decline in amplitude with a much longer time course in post rotary nystagmus (Cohen et al. 1977; Cohen et al. 1992). The function of habituation as well as whether it largely reflects central or peripheral phenomena has been often debated. If habituation is viewed as an essential adaptive property of the velocity storage system (Cohen et al. 1992; Kleinschmidt and Collewijn 1975), then it must be equally vital to both the vestibular and visual motor system as suggested by its presence in both aquatic and terrestrial vertebrates. Habituation, however, is not correlated with the mechanisms mediating oculomotor plasticity because both gain changes and period tuning occurs independent of the habituated state. We agree with the suggestion that the habituated oculomotor reflexes may be primarily an artifact of the restrained condition and may not reflect the more normal state of the oculomotor (gaze also) system found in freely moving animals (Cohen et al. 1992; Kleinschmidt and Collewijn 1975). Mechanistically, we note that unique intrinsic properties of the brain stem neurons may be required to produce velocity storage, and these properties are not normally faced with either prolonged, repetitive visual or vestibular stimuli. As a result of constant negative feedback, there is a reduction in the gain of the premotor neurons. Hence, unnatural stimuli rapidly and with long-lasting effect can modify the electrophysiological and molecular properties that control this neuronal circuitry.

Separate evolution for OKRe and OKRd

Early and delayed components of eye velocity are thought to represent two separate steps of neural processing underlying the visuomotor response to visual velocity steps (Cohen et al. 1977; Lisberger et al. 1981). The early component in goldfish is an abrupt, large change in eye velocity elicited by sudden changes in the whole visual world. OKRe appears to have a maximum velocity and short-duration (milliseconds) largely independent of stimulus amplitude. The low level for saturation supports the view of an anatomically separate, dynamic visuomotor synaptic pathway (Lisberger et al. 1981). By contrast, the delayed component develops slowly (seconds), and this build-up likely is the result of a brain stem velocity storage system. The delayed component is thought to be largely responsible for maintaining compensatory eye movements in the absence of retinal slip, especially in response to head rotation. However, this view may be too simplistic because the OKR and VOR are rarely independent under natural situations (Collewijn 1989).

The evolutionary basis of these two components of the OKR has been under theoretical consideration for years. In the oculomotor system, vestibular and visual signals may have evolved separately, sharing only motoneurons as their common central targets. Co-adaptation could have led to the acquisition of different (not shared) central neuronal pathways each allowing for either short and quick, or long and slow, compensatory eye motion during either world or body motion. Thus the early and late components of the OKR might be viewed as discrete parts of the oculomotor system, not necessarily different neurons/circuits, that are primarily designed to work conjointly with their VOR counterparts. In primates, smooth pursuit dominates visuomotor function, and the early component of the OKR is thought to be related to smooth pursuit. Miles and Busentii (Miles 1994; Miles and Busettini 1992) proposed that the early and late components of ocular following were evolutionary related to the acquisition of translational VOR and rotational VOR, respectively. This hypothesis, however, appears unlikely because both the early and delayed components of the OKR are found in the goldfish, which has neither smooth pursuit nor the need for translational VOR. A translational VOR in the goldfish is unlikely to exist because goldfish lack a fovea and therefore would not experience visual flow that would stimulate a translational VOR. The observation of early/delayed components in goldfish supports the suggestion that OKRe/d may have evolved for another purpose and/or were more simply designed to match up with VOR dynamics. Nevertheless, there appears to be some significant overlap in central OKR and VOR pathways in the goldfish because gain increases are shared between the OKR and VOR, notably after training with only a visual stimulus (Fig. 7).

OKR adaptation

Training with a variety of visual stimuli produced an increase in OKR gain. Eye velocity could be increased with unidirectional, symmetric bidirectional, and asymmetric bidirectional visual velocity steps even with a random delay inserted between each cycle (Figs. 6-9). The increase in eye velocity was manifest by changes in the amplitude and time course of OKRe and OKRd. An increase in the amplitude of the early component produced an increase in the velocity of eye movements. Decreasing the time to reach maximum eye velocity (i.e., shortening of the time course of the delayed component) resulted in an increased gain in response to a short visual velocity step. Changes in the amplitude and time course of OKR eye velocity always occurred in parallel. The magnitude of gain changes noted previously in goldfish (Schairer and Bennett 1986b) were similar to those reported here and have not been observed in other animals. However, changes in VOR and OKR gain after prolonged visual stimulation has been seen in rabbits (Collewijn and Grootendorst 1979).

Visual training modifies the VOR

After visuomotor training the VOR also showed an increase in gain (Fig. 7). The changes in VOR gain were similar to a previous goldfish study (Schairer and Bennett 1986b), but in the rabbit, VOR gain increases were much less (Collewijn and Grootendorst 1979). A second important finding was that the change in VOR gain after visual stimulus training did not occur at the earliest possible latency of 17 ms as seen with visuo-vestibuloocular adaptations (Pastor et al. 1992) but at ~56 ms. This difference between visuo-vestibular conflict and visual adaptation paradigms is significant. In VOR plasticity, the site of adaptation was hypothesized to be in the vestibular nucleus based on the eye velocity changes occurring at the earliest latency and the effects of cerebellectomy (Pastor et al. 1994a). Changes in VOR gain after OKR plasticity occurred at a later time, suggesting either a location other than the vestibular nucleus in the brain stem or the involvement of the cerebellum as the site of the adaptation. The site of adaptation is not likely in the cerebellum because after cerebellectomy VOR gain changes are retained (Marsh et al. 1994). Therefore, the changes induced by visual stimulus training likely occur in either Area II or within a presumed loop between Area II and the vestibular nucleus (Cheron 1991). A related point concerning the shared gain changes is that the increases found in the OKR and VOR after visual stimulus training are larger for the OKR but smaller for the VOR than after visual-vestibular training. The gain increase in the two oculomotor subsystems occur in a 1:0.6 ratio (visual:vestibular). Nearly, the same difference was seen in rabbit (Collewijn and Grootendorst 1979), cat (Demer 1981), and monkey (Lisberger et al. 1981). These studies support the hypothesis for a shared site of adaptation between the VOR and OKR, however, our data also suggests an additional separate gain site in the goldfish visuomotor system.

Frequency selectivity

In previous studies on VOR and OKR plasticity, either large increases or decreases in gain were observed at the training frequency and significantly smaller changes at any other frequency tested (Collewijn and Grootendorst 1979; Godaux et al. 1983b; Lisberger et al. 1983; Schairer and Bennett 1986a; Wallman et al. 1982). This effect was hypothesized to be due to either frequency selective channels (Lisberger et al. 1983) or to storage of the stimulus pattern (Collewijn and Grootendorst 1979). The results presented here show a broader range of frequencies to be affected by the visual stimulus training. Training with velocity steps, which include a broadband of frequencies, may in part be responsible, however, the hypothesized shared and separate gain elements also may contribute. If the vestibular nucleus were to be the site of frequency selective gain changes and Area II the site for broadband gain changes, then an adaptation paradigm that increased gain at both sites (i.e., OKR adaptation) would result in a broadband gain increases. Adaptation only within the vestibular nucleus, i.e., occurring during VOR adaptation, would result in frequency selective effects.

Entrainment of eye velocity

After prolonged visual motor training, eye velocity increased to match stimulus velocity (i.e., compensatory). Eye velocity then appeared to decrease before the end of the visual stimulus period (i.e., a predictive phenomenon). In most cases, eye velocity oscillated in the dark at the training frequency. This peculiar type of tuning was observed as early as 30 min after the start of training and became increasingly more pronounced over time. Eye velocity clearly decreased >1 s preceding the change in direction of the visual stimulus (Figs. 6 and 8). After 1-2 h of training, eye velocity decreased toward 0 eye velocity coincident with the turn around of the training period and remained at 0°/s for any extended duration of the visual stimulus (e.g., from 4 to 8 s; Figs. 6 and 8). These data alone suggest that the period of the training stimulus had been entrained in the OKR circuit to produce a decrease in eye velocity even though retinal slip velocity was continuously becoming larger. Because retinal slip induces the OKR and the main purpose of the OKR is to eliminate retinal slip, then this type of plasticity might be viewed as paradoxical. Certainly the entrainment of a fixed frequency of repetitive visual motion may not be beneficial to either the optomotor or vestibular reflex system. However, the observation that the amplitude and time course of the OKR can be modified along with the rhythmic characteristics is significant because the cellular mechanisms responsible for such oscillation clearly represent another oculomotor learning paradigm. This type of plasticity also could be relevant to understanding visuomotor signal processing, and thus it was important to distinguish whether this behavior was correlated to the period or frequency of visual stimulation.

The decrease in eye velocity observed after visual stimulus training had not been reported previously in other species. However, the oscillation at the training frequency in the dark has been observed in rabbits (Collewijn and Grootendorst 1979; Kleinschmidt and Collewijn 1975) and goldfish (Schairer and Bennett 1986b). An oscillation in complex spike modulation of Purkinje cells at the VOR training frequency was found in rabbits for only low frequency (Barmack and Shojaku 1992). Although these oscillations were offered as explanations for the frequency selectivity of the gain changes, in no case, was the training period extended and a change in eye velocity reported. The occurrence of an oscillation in eye velocity after training does suggest that brain stem-cerebellar circuitry can "memorize" the training period, but how this is accomplished and at what level this interaction influences the gain elements is not clear. The battery of experimental procedures used herein suggest that the period or duration of the eye velocity response is learned and not the frequency of the training stimulus.

Period versus frequency tuning

Period and frequency were distinguished in these experiments by using two different stimulus waveforms. One was an asymmetrical repetitive velocity step, where the planetarium moved for different periods of times in each direction (e.g., 4 s to the left and 10 s to the right). The second training protocol was a symmetrical velocity step with a random pause of 1-8 s between each cycle. Training with both of these stimuli resulted in an eye velocity decrease as observed with the repetitive step stimuli (Fig. 9), with the only difference being variable durations of tuning to the left and to the right. In both cases, there was no frequency of stimulus presentation, only a length of time for light motion, strongly arguing for learning of the period and not the frequency of training. These observations, along with the fact that step stimuli contain all frequency components, eliminating a visual frequency selectivity, strongly argued for this behavior to be termed "period tuning." We further would suggest that the important factor in generating period tuning may not be the visual stimulus, but rather the internal representation of eye velocity (i.e., the so called efference copy) (McCrea 1988).

Period tuning could be elicited at periods as short as 2 s and as long as we tested: 128 s. At periods <2 s, period tuning could not be observed but there were small increases in gain. Stimulus durations <2 s did not allow for build-up of much velocity storage, suggesting that velocity storage is involved in the behavior. Hence, the learning or imprinting of the eye velocity may be occurring within the storage circuitry, which appears to be an easily modifiable synaptic pathway.

Period tuning is a very plastic phenomena, because one animal can learn sequentially (and forget) many periods within 45 min after starting each training session. Period tuning was extremely labile because tuning never lasted >50 min after individual training sessions. In contrast, the gain increases out lasted the training sessions and required additional training to reverse (data not shown). These observations suggest that gain changes and period tuning are independent of each other implying either two separate sites of modification and/or separate cellular mechanisms generating each behavior.

Period tuning was not an artificial response of the restrained goldfish because three training sessions were performed with the fish freely swimming around the circular tank. In these experiments, period tuning was normal (Fig. 9) and the gain of the response increased as well.

After training, a 2.4-Hz oscillation was found in the eye movements when the light was on and stationary (Fig. 10). Turning off the light caused the oscillation to stop. The oscillation appears to be correlated directly with an up-regulation of visuomotor gain because it does not appear with vestibular training (unpublished observations). This oscillation parallels the increase in visuooculomotor gain, which fits well with the idea that the oculomotor system functions like a velocity servo (Collewijn 1972a; Robinson 1975), in that if the closed loop gain of the system becomes too high, eye velocity begins to oscillate.

In conclusion, the goldfish optokinetic response is qualitatively similar to that of other species, but it has two distinct robust adaptive processes associated with the OKR. The adaptive gain changes are shared with the vestibulomotor system, probably indicating a joint gain modification site that can be accessed with either visual or vestibular training. Period tuning is unique to the OKR possibly reflecting a second plastic mechanism in the oculomotor system.

	ACKNOWLEDGEMENTS

  The authors would like to thank Dr. J. I. Simpson for his comments and suggestions on the manuscript.

  This work was supported by National Institute of Neurological Disorders and Stroke Grant NS-13742.

	FOOTNOTES

  Address for reprint requests: R. Baker, Dept. of Physiology and Neuroscience, New York University Medical Center, 550 First Ave., New York 10016.

  Received 31 July 1996; accepted in final form 29 October 1996.

	REFERENCES

Abstract Introduction Methods Results Discussion References

• BAARSMA, E. A., COLLEWIJN, H. Vestibulo-ocular and optokinetic reactions to rotation and their interaction in the rabbit. J. Physiol. Lond. 238: 603-625, 1974.[Abstract/Free Full Text]
• BARMACK, N. H., SHOJAKU, H. Vestibularly induced slow oscillations in climbing fiber responses of purkinje cells in the cerebellar nodulus of the rabbit. Neuroscience 50: 1-5, 1992.[Medline]
• CARPENTER, R.H.S. In: Movements of the Eyes., . London: Pion LTD, 1977
• CHERON, G. Effects of incision in the brainstem commissural network on the short-term vestibulo-ocular adaptation of the cat. J. Ves. Res. 1: 223-239, 1991.
• COHEN, B., MATUSO, V., RAPHAN, T. Quantitative analysis of the velocity characteristics of the optokinetic nystagmus and optokinetic after-nystagmus. J. Physiol. Lond. 270: 321-344, 1977.[Abstract/Free Full Text]
• COHEN, H., COHEN, B., RAPHAN, T., WAESPE, W. Habituation and adaptation of the vestibuloocular reflex: a model of differential control by the vestibulocerebellum. Exp. Brain Res. 90: 526-538, 1992.[Medline]
• COLLEWIJN, H. Optokinetic eye movements in the rabbit: input-output relations. Vision Res. 9: 117-132, 1969.[Medline]
• COLLEWIJN, H. The optokinetic system of the rabbit. Doc. Ophthalmol. 30: 206-236, 1971.
• COLLEWIJN, H. An analog model of the rabbit's optokinetic system. Brain Res. 36: 71-88, 1972a.[Medline]
• COLLEWIJN, H. Latency and gain of the rabbit's optokinetic reactions to small movements. Brain Res. 36: 47-57, 1972b.[Medline]
• COLLEWIJN, H. In: The Oculomotor System of the Rabbit and its Plasticity., . Berlin: Springer-Verlag, 1981
• COLLEWIJN, H. The vestibulo-ocular reflex: an outdated concept? Prog. Brain Res. 80: 197-209, 1989.[Medline]
• COLLEWIJN, H., GROOTENDORST, A. F. Adaptation of optokinetic and vestibulo-ocular reflexes to modified visual input in the rabbit. In: Progress in Brain Research. Reflex Control of Posture and Movement, edited by R. Granit, and O. Pompeiano . Amsterdam: Elsevier, 1979, p. 772-781
• COLLEWIJN, H., WINTERSON, B. J., VAN DER STEEN, J. Post-rotatory nystagmus and optokinetic after-nystagmus in the rabbit linear rather that exponential decay. Exp. Brain Res. 40: 330-338, 1980.[Medline]
• DAVIES, P., MELVILLE JONES, G. Adaptive neural model compatible with plastic changes induced in the human vestibulo-ocular reflex by prolonged optical reversal of vision. Brain Res. 103: 546-550, 1976.[Medline]
• DELGADO-GARCIA, J. M., VIDAL, P. P., GOMEZ, C., BERTHOZ, A. A neurophysiological study of the prepositus hypoglossi neurons projecting to oculomotor and preoculomotor nuclei in the alert cat. Neuroscience 29: 291-307, 1989.[Medline]
• DEMER, J. L. The variable gain elements of the vestibulo-ocular reflex is common to the optokinetic system of the cat. Brain Res. 229: 1-13, 1981.[Medline]
• DIERINGER, N., REICHENBERGER, I., GRAF, W. Differences in optokinetic and vestibular ocular reflex performance in teleosts and their relationship to different life styles. Brain Behav. Evol. 39: 289-304, 1992.[Medline]
• DU LAC, S., RAYMOND, J. L., SEJNOWSKI, T. J., LISBERGER, S. G. Learning and memory in the vestibulo-oculo reflex. Annu. Rev. Neurosci. 18: 409-441, 1995.[Medline]
• EASTER, S. S. Pursuit eye movements in goldfish (Carassius Auratus). Vision Res. 12: 673-688, 1972.[Medline]
• EVINGER, C., FUCHS, A. F. Saccadic smooth pursuit, and optokinetic eye movements of the trained cat. J. Physiol. Lond. 285: 209-229, 1978.[Abstract/Free Full Text]
• FABER, D. S., KORN, H., LIN, J. W. Role of medullary networks and postsynaptic membrane properties in regulating Mauthner cell responsiveness to sensory excitation. Brain Behav. Evol. 37: 286-297, 1991.[Medline]
• FINGER, T. E. Efferent neurons of the teleost cerebellum. Brain Res. 153: 608-614, 1978.[Medline]
• FINGER, T. E. Organization of the teleost cerebellum. In: Fish Neurobiology, edited by R. G. Northcutt, and R. E. Davis . Ann Arbor: The University of Michigan Press, 1983, p. 261-284
• FUCHS, F. A. Saccadic and smooth pursuit eye. Movements in the monkey. J. Physiol. Lond. 191: 609-631, 1967.[Abstract/Free Full Text]
• GALIANA, H. L., OUTERBRIDGE, J. S. A bilateral model for central neural pathways in vestibuloocular reflex. J. Neurophysiol. 51: 210-241, 1984.[Abstract/Free Full Text]
• GODAUX, E., GOBERT, C., HALLEUX, J. Vestibuloocular reflex, optokinetic response, and their interactions in the alert cat. Exp. Neurol. 80: 42-54, 1983a.[Medline]
• GODAUX, E., HALLEUX, J., GOBERT, C. Adaptive change of the vestibulo-ocular reflex in the cat: the effects of a long-term frequency selective procedure. Exp. Brain Res. 49: 28-34, 1983b.[Medline]
• GODAUX, E., METTENS, P., CHERON, G. Differential effect of injections of kainic acid into the prepositus and the vestibular nuclei of the cat. J. Physiol. Lond. 472: 459-482, 1993.[Abstract/Free Full Text]
• HESS, B.J.M., PRECHT, W., REBER, A., CAZIN, L. Horizontal optokinetic ocular nystagmus in the pigmented rat. Neuroscience 15: 97-107, 1985.[Medline]
• ITO, M. Cerebellar flocculus hypothesis. Nature Lond. 363: 24-25, 1993.[Medline]
• JAGER, J., HENN, V. Habituation of the vestibulo-ocular reflex (VOR) in the monkey during sinusoidal rotation in the dark. Exp. Brain Res. 41: 108-114, 1981.[Medline]
• KELLER, E. L., PRECHT, W. Visual-vestibular responses in vestibular nuclear neurons in the intact and cerebellectomized, alert cat. Neuroscience 4: 1599-1613, 1979a.[Medline]
• KELLER, E. L., PRECHT, W. Adaptive modification of central vestibular neurons in response to visual stimulation through reversing prisms. J. Neurophysiol. 42: 896-911, 1979b.[Free Full Text]
• KLEINSCHMIDT, H. J., COLLEWIJN, H. A search for habituation of vestibulo-ocular reactions to rotatory and lenear sinusoidal accerlerations in the rabbit. Exp. Neurol. 47: 257-267, 1975.[Medline]
• LISBERGER, S. G. The neural basis for learning of simple motor skills. Science Wash. DC 242: 728-734, 1988.[Abstract/Free Full Text]
• LISBERGER, S. G., MILES, F. A., OPTICAN, L. M. Frequency-selective adaptation: Evidence for channels in the vestibulo-ocular reflex? J. Neurosci. 3: 1234-1244, 1983.[Abstract]
• LISBERGER, S. G., MILES, F. A., OPTICAN, L. M., EIGHMY, B. B. Optokinetic response in monkey: underlying mechanisms and their sensitivity to long-term adaptive changes in vestibuloocular reflex. J. Neurophysiol. 45: 869-890, 1981.[Free Full Text]
• LLINAS, R., WELSH, J. P. On the cerebellum and motor learning. Curr. Opin. Neurobiol. 3: 958-965, 1993.[Medline]
• MAIOLI, C., PRECHT, W. The horizontal optokinetic nystagmus in the cat. Exp. Brain Res. 55: 494-506, 1984.[Medline]
• MARSH, E., PASTOR, A., BAKER, R. Cerebellar contribution to goldfish visuomotor adaptation. Soc. Neurosci. Abstr. 20: 858, 1994.
• MCCREA, R. The nucleus prepositus. In: Neuroanatomy of the Oculomotor System, edited by and J. A. Buttner-Ennever . Amsterdam: Elsevier/North Holland Biomedical, 1988, p. 203-223
• MCELLIGOTT, J. G., WEISER, M., BAKER, R. The effect of temperature on adaptation of the vestibulo-ocular reflex in the goldfish. J. Neurophysiol. 74: 1463-1472, 1995.[Abstract/Free Full Text]
• METTENS, P., GODAUX, E., CHERON, G., GALIANA, H. L. Effect of muscimol microinjections into the prepositus hypoglossi and the medial vestibular nuclei on cat eye movements. J. Neurophysiol. 72: 785-802, 1994.[Abstract/Free Full Text]
• MICHNOVICZ, J. J. A Cerebellar Role in the Control of the Goldfish Vestibulo-Ocular Reflex (PhD thesis). Bronx, NY: Albert Einstein College of Medicine, 1983.
• MILES, F. A. Stimulus specificity in the primate optokinetic system. In: Information Processing Underlying Gaze Contol, edited by J. M. Delgado-Garcia, E. Godaux, and P.-P. Vidal . Tarrytown: Pergamon, 1994
• MILES, F. A., BUSETTINI, C. Ocular compensation for self-motion. In: Sensing and Controlling Motion: Vestibular and Sensorimotor Function, edited by B. Cohen, D. L. Tomko, and F. Guedry . New York: N. Y. Acad. Sci., 1992, p. 220-232
• MILES, F. A., EIGHMY, B. B. Long-term adaptive changes in primate vestibuloocular reflex. I. Behavioral observations. J. Neurophysiol. 43: 1406-1425, 1980.[Free Full Text]
• OYSTER, C. W. The analysis of image motion by the rabbit retina. J. Physiol. Lond. 199: 613-635, 1968.[Abstract/Free Full Text]
• PARTSALIS, A. M., ZHANG, Y., HIGHSTEIN, S. M. Dorsal Y group in the squirrel monkey. I. Neuronal responses during rapic and long-term modifications of the vertical VOR. J. Neurophysiol. 73: 615-631, 1995.[Abstract/Free Full Text]
• PASTOR, A. M., DE LA CRUZ, R. R., BAKER, R. Characterization and adaptive modifications of the goldfish vestibulo-ocular reflex by sinusoidal and velocity step vestibular stimulation. J. Neurophysiol. 68: 2003-2015, 1992.[Abstract/Free Full Text]
• PASTOR, A. M., DE LA CRUZ, R. R., BAKER, R. Cerebellar role in adaptation of the goldfish vestibulo-ocular reflex. J. Neurophysiol. 72: 1383-1394, 1994a.[Abstract/Free Full Text]
• PASTOR, A. M., DE LA CRUZ, R. R., BAKER, R. Eye position and velocity integrators reside in separate brainstem nuclei in the goldfish hindbrain. Proc. Natl. Acad. Sci. USA 91: 807-811, 1994b.[Abstract/Free Full Text]
• PASTOR, A. M., MARSH, E., BAKER, R. Role of the cerebellum in visuo-motor and vestibulo-ocular adaptation is independent of oculomotor performance in three orders of teleosts. Soc. Neurosci. Abstr. 20: 857, 1994c.
• PASTOR, A. M., WEISER, M., MCELLIGOTT, J. G., BAKER, R. Velocity step training of the goldfish vestibulo-ocular reflex suggests multiple sites are involved on gain modification. Soc. Neurosci. Abstr. 16: 1083, 1990.
• PRECHT, W. Vestibular mechanisms. Annu. Rev. Neurosci. 2: 265-289, 1979.[Medline]
• RAPHAN, T., MATSUO, V., COHEN, B. Velocity storage in the vestibulo-ocular reflex arc (VOR). Exp. Brain Res. 35: 229-248, 1979.[Medline]
• ROBINSON, D. A. Eye movement control in primates. Science Wash. DC 161: 1219-1224, 1968.[Free Full Text]
• ROBINSON, D. A. Oculomotor signal controls. In: Basic Mechanisms of Ocular Motility and Their Clinical Implications, edited by G. Lennerstrand, and P. Bach-y-Rita . New York: Pergamon, 1975, p. 337-374
• SCHAIRER, J. O. Adaptive Gain Control of the Vestibulo-Ocular Reflex in Goldfish (PhD thesis). Bronx, NY: Albert Einstein College of Medicine, 1980.
• SCHAIRER, J. O., BENNETT, M.V.L. Changes in gain of the vestibulo-ocular reflex induced by combined visual and vestibular stimulation in goldfish. Brain Res. 373: 164-176, 1986a.[Medline]
• SCHAIRER, J. O., BENNETT, M.V.L. Changes in gain of the vestibulo-ocular reflex induced by sinudoidal visual stimulation in goldfish. Brain Res. 373: 177-181, 1986b.[Medline]
• SCUDDER, C. A., FUCHS, A. F. Physiological and behavioral identification of vestibular nucleus neurons mediating horizontal vestibulo-ocular reflex in trained Rhesus monkey. J. Neurophysiol. 68: 244-264, 1992.[Abstract/Free Full Text]
• TEMPIA, F., DIERINGER, N., STRATA, P. Adaptation and habituation of the vestibulo-ocular reflex in intact and inferior olive-lesioned rats. Exp. Brain Res. 86: 568-578, 1991.[Medline]
• WALLMAN, J., VELEZ, J., WEINSTEIN, B., GREEN, A. E. Avian vestibulo-ocular reflex: adaptive plasticity and developmental changes. J. Neurophysiol. 48: 952-967, 1982.[Abstract/Free Full Text]
• WALLS, G. L. The evolutionary history of eye movements. Vision Res. 2: 69-80, 1962.

0022-3077/97 $5.00 Copyright ©1997 The American Physiological Society