Gravity-Specific Adaptation of the Angular Vestibuloocular Reflex: Dependence on Head Orientation With Regard to Gravity

Abstract

The gain of the vertical angular vestibuloocular reflex (aVOR) was adaptively altered by visual-vestibular mismatch during rotation about an interaural axis, using steps of velocity in three head orientations: upright, left-side down, and right-side down. Gains were decreased by rotating the animal and visual surround in the same direction and increased by visual and surround rotation in opposite directions. Gains were adapted in one head position (single-state adaptation) or decreased with one side down and increased with the other side down (dual-state adaptation). Animals were tested in darkness using sinusoidal rotation at 0.5 Hz about an interaural axis that was tilted from horizontal to vertical. They were also sinusoidally oscillated from 0.5 to 4 Hz about a spatial vertical axis in static tilt positions from yaw to pitch. After both single- and dual-state adaptation, gain changes were maximal when the monkeys were in the position in which the gain had been adapted, and the gain changes progressively declined as the head was tilted away from that position. We call this gravity-specific aVOR gain adaptation. The spatial distribution of the specific aVOR gain changes could be represented by a cosine function that was superimposed on a bias level, which we called gravity-independent gain adaptation. Maximal gravity-specific gain changes were produced by 2–4 h of adaptation for both single- and dual-state adaptations, and changes in gain were similar at all test frequencies. When adapted while upright, the magnitude and distribution of the gravity-specific adaptation was comparable to that when animals were adapted in side-down positions. Single-state adaptation also produced gain changes that were independent of head position re gravity particularly in association with gain reduction. There was no bias after dual-state adaptation. With this difference, fits to data obtained by altering the gain in separate sessions predicted the modulations in gain obtained from dual-state adaptations. These data show that the vertical aVOR gain changes dependent on head position with regard to gravity are continuous functions of head tilt, whose spatial phase depends on the position in which the gain was adapted. From their different characteristics, it is likely that gravity-specific and gravity-independent adaptive changes in gain are produced by separate neural processes. These data demonstrate that head orientation to gravity plays an important role in both orienting and tuning the gain of the vertical aVOR.

INTRODUCTION

When the head is rotated, the angular vestibuloocular reflex (aVOR) stabilizes gaze by generating counter-rotation of the eye in the orbit. In response to higher frequency and higher acceleration stimuli, the reflex, characterized by its gain, is close to unity for passive rotations [human, (Aw et al. 1996) monkey, (Minor et al. 1999)]. Generally, however, the aVOR does not precisely compensate for head movement in darkness either in monkeys (Crawford and Vilis 1991; Robinson 1963; Skavenski and Robinson 1973) or in humans (Collewijn and Grootendorst 1978; Gonshor and Melvill Jones 1976b; Melvill Jones and Davies 1976). To overcome this deficiency, the brain has alternate ways to help promote compensation. There is a substantial improvement in aVOR gain during active head movement (Uemura et al. 1980, 1981), during rotation in light (Barnes 1988; Barr et al. 1976), and when subjects imagine an earth-fixed target (Melvill Jones et al. 1984; Schultheis and Robinson 1981). Additionally, the gain of the aVOR can be adaptively increased or decreased so that ocular compensation is more precise (Gonshor and Melvill Jones 1971; Miles and Fuller 1974; Yakushin et al. 2000b). While many previous studies of adaptation have concentrated on the horizontal component of the aVOR, the vertical (Hirata and Highstein 2001; Partsalis et al. 1995a; Snyder and King 1988) and torsional (Berthoz et al. 1981) components of the aVOR are equally modifiable, and the horizontal, vertical, and torsional aVORs can be separately adapted (Bello et al. 1991). Additionally, vestibular and visual stimuli in disparate planes can spatially adapt the aVOR so that the eyes move obliquely in darkness in response to pure horizontal or vertical stimuli (cross-axis adaptation) (Baker et al. 1986, 1987a,b; Harrison et al. 1986a,b; Schultheis and Robinson 1981).

In both monkeys and humans, the first significant gain changes occur as early as 20–40 min after onset of the conditioning procedure (Cohen et al. 1992; Collewijn et al. 1983; Gonshor and Melvill Jones 1976a,b; Partsalis et al. 1995a), and 2 h of adaptation will produce gain changes in the monkey of about 20–25% (Cohen et al. 1992; Partsalis et al. 1995a; Yakushin et al. 2000b). If adaptation is continued for an additional 2 h, there is only a slight additional gain change (~5%). At that point, the gain stabilizes and is unchanged even if stimulation is prolonged for up to 8 h (Bello et al. 1991; Cohen et al. 1992; Godaux et al. 1983; Lisberger et al. 1984; Miles and Eighmy 1980; Nagao 1989; Yakushin et al. 2000b). Beyond that, if adaptation continues, there are further significant gain changes (Berthoz et al. 1981; Gonshor and Melvill Jones 1976b; Lisberger and Pavelko 1986; Lisberger et al. 1983; Melvill Jones and Davies 1976; Miles and Eighmy 1980; Miles and Lisberger 1981a). Despite the possibility that the different periods of adaptation necessary to produce gain adaptation represent separate processes in the nervous system, much has been learned from studying short-term adaptation of the aVOR gain, i.e., the adaptation that takes place within 4 h.

The full range of signals that drive gain adaptation of the aVOR is not known (Highstein et al. 1997; Hirata and Highstein 2001; Ito et al. 1970; Lisberger and Fuchs 1978; Lisberger 1996), but induction of retinal slip over extended periods has been the most effective technique for inducing gain changes (Gonshor and Melvill Jones 1976a; Ito and Miyashita 1975; Miles and Lisberger 1981b; Yakushin et al. 2000b). Such retinal slip can be produced by reversing prisms (Gonshor and Melvill Jones 1973, 1976a,b), by magnifying or reducing lenses (Collewijn et al. 1983; Gauthier and Robinson 1975; Lisberger and Miles 1980; Miles and Fuller 1974), or by passive oscillation of the animal in-phase or out-of-phase with the visual surround (Cohen et al. 1992; Ito et al. 1974; Yakushin et al. 2000a,b). One hypothesis as to why retinal slip induces modification of the aVOR gain is that the visual system does not operate at the same frequencies as the aVOR. The aVOR responds to head movements at frequencies up to 8–10 Hz (Tabak and Collewijn 1994), but the visual system can only directly augment the VOR at frequencies up to ~1–1.5 Hz (Boyle et al. 1985; Fender and Nye 1961; Yakushin et al. 1996). Consequently, the retinal slip associated with such high-frequency head movements cannot be accurately sensed by the visual system. To meet this deficiency, gain values are altered in the compensatory direction even in the absence of precise ocular following (Lisberger et al. 1984). Additionally, retinal error signals, which invoke the saccadic system, could be used to help adapt the aVOR for more accurate compensation.

The otoliths, through activation of compensatory and orienting components of the linear vestibuloocular reflex (lVOR) can also augment the aVOR (Angelaki et al. 2002; Paige and Tomko 1991; Raphan et al. 1996; Wearne et al. 1999; see also Raphan and Cohen 2002, for review). In a frequency range >1.0 Hz, the compensatory lVOR superposes with the aVOR to either enhance or suppress the aVOR (Paige and Tomko 1991; Telford et al. 1997, 1998). In a frequency range of 0.1–1.0 Hz, the gain of the vertical aVOR was larger when cats were oscillated in pitch about a horizontal axis than when the gain was tested with the animals on their sides, rotating around a vertical axis without change in a gravitational component (Tomko et al. 1988). The conclusion of that study was that an orienting component of the lVOR contributes to the aVOR response. Otolith input also contributes to orientation of eye velocity produced by angular rotation or by rotation of the visual surround through velocity storage (Dai et al. 1991; Hess and Angelaki 1997a,b; Raphan and Cohen 1996; Raphan and Sturm 1991; Raphan et al. 1992).

We recently demonstrated that if the gain of the vertical aVOR was adapted in one on-side head position, the gain changes were maximal when the animals were in this position, and there was little or no gain change when animals were tested with their contralateral side down (Yakushin et al. 2000a). This finding shows that static orientation of the head with regard to gravity can influence the high-frequency aVOR response through adaptation of the direct pathway gain, analogous to the way visual following parametrically modifies the high-frequency aVOR pathway. We referred to this phenomenon as gravity-specific aVOR gain adaptation. The purpose of the present study was to investigate the spatial tuning of the gravity-specific adaptation of the vertical aVOR to determine the function that relates such gain changes to head position with regard to gravity and their frequency characteristics. We also wished to determine whether the gravity-dependent gain changes were broadly or narrowly tuned in space to the position in which the gain was adapted and whether the spatial tuning of adapted gain increases and decreases following adaptation to two different gravitational contexts could occur concurrently.

METHODS

Three cynomolgus (Macaca fascicularis, M96012, M98060, and M98078) and two rhesus (M. mulatta, M98063 and M98064) monkeys were utilized in this study. The experiments conformed to the Guide for the Care and Use of Laboratory Animals (National Research Council 1996) and were approved by the Institutional Animal Care and Use Committee. Surgical procedures were performed under anesthesia in sterile conditions. Procedures were performed in two stages. First, a head mount was implanted on the skull to provide painless head fixation in stereotaxic coordinates during testing (Beloozerova and Sirota 1986, 1993; Sirota et al. 1988; Yakushin et al. 2000b). At a second surgery 2 wk later, two three-turn coils were implanted on the left eye. One coil measured the horizontal and vertical components of eye position (Judge et al. 1980; Robinson 1963). Another coil, placed approximately orthogonal to the frontal coil (Cohen et al. 1992), was used to measure the torsional component of eye position. Postoperatively, the animals were treated with analgesics, antibiotics and steroids to relieve pain and inflammation.

Recording of eye movements with search coils

During testing, the monkey’s head was fixed to a plastic frame, which held two sets of field coils that generated orthogonal oscillating magnetic fields at the same frequency. The axes of the field coils were along the interaural (pitch) and dorsoventral (yaw) axes of the head, establishing a head-fixed reference frame for measuring the orientation of the frontal and top search coils. Monkeys were positioned so that the eye with the search coils was at the center of the magnetic fields. To calibrate eye movement, the animals were rotated in light at 30°/s about a spatial vertical axis while upright for yaw, left-side down for pitch and prone for roll. It was assumed that horizontal and vertical gains were close to unity when upright or side-down (Raphan et al. 1979; Robinson 1963), and torsional gains were assumed to be 0.6 when the rotation was performed around a naso-occipital axis aligned with the spatial vertical (Crawford and Vilis 1991; Henn et al. 1992; Yakushin et al. 1995).

Data processing

Eye position voltages and voltages related to the velocity of the chair oscillation as well as to the position of the tilt axes were recorded with amplifiers having a band-pass of DC to 40 Hz. Data were acquired by computer and analyzed off-line. Voltages were digitized at 600 Hz/channel with 12-bit resolution. Voltages related to eye position were digitally differentiated by finding the slope of the least squares linear fit to 11 data points. This corresponds to a filter, which has a 3 dB cutoff >40 Hz, the cutoff frequency of the filters used for data acquisition. Saccades were eliminated using a maximum likelihood ratio criterion (Singh et al. 1981).

Experimental protocol

During testing, the animals sat in a primate chair in a four-axis vestibular stimulator surrounded by an optokinetic drum. Each axis went through the center of rotation of the head. The stimulator used in this study has been described in detail in previous publications (Yakushin et al. 1995, 2000a). In brief, the optokinetic drum had a diameter of 91 cm and contained vertical 10° black-and-white stripes. The axes of rotation of the animals and optokinetic drum were colinear, and when the optokinetic drum rotated around the animal in light, it produced full-field visual stimulation.

Gains were decreased by rotating the animal and visual surround in the same direction and increased by animal and the visual surround rotation in opposite directions. Adaptation was carried out over a 4-h period in each instance. To decrease the aVOR gain, the primate and drum axes were rotated in light with steps of velocity of 60°/s in the same direction for 20 s (Fig. 1A). The animals were then stopped for 5 s and rotated in the opposite direction. This resulted in a reduction in the initial eye velocity and a rapid decline in velocity to zero over 3–8 s. To increase the vertical aVOR gain, the animal and optokinetic drum were first rotated in opposite directions in darkness at 30°/s (Fig. 1B). Two seconds after the onset of rotation, the visual surround was illuminated exposing the animal to a relative visual surround movement of 60°/s (30°/s +30°/s) (Fig. 1B). The light was extinguished 2 s before the end of rotation. Five seconds after the end of rotation, the sequence was repeated in the opposite direction. When the animals were adapted in an upright position with steps of velocity, they were rotated at 30°/s for 5 s around the upright (±75°) to ensure that the fore-aft tilts were maintained within 90°.

FIG. 1.

A: suppression of the vertical eye velocity (top) induced by in-phase rotation of the monkey (middle) and optokinetic surrounding (bottom). B: enhancement of the vertical eye velocity by out-of-phase rotation of the monkey (middle) and optokinetic surrounding (bottom).

In the first set of experiments, the vertical aVOR gain was adapted by oscillating the monkeys (M96012, M98060, and M98078) about an interaural axis in one of three positions: upright, left side down (LSD), or right side down (RSD) in phase or out of phase with the visual surround. Adaptation was done in only one position on any particular day. We refer to this as single-state adaptation.

In a second set of experiments (dual-state adaptation), the vertical aVOR gains were both increased and decreased in the same session. Four animals (M96012, M98060, M98063, and M98064) were first adapted on one side for 15 min to decrease the gain as described in the preceding text. Then their position was shifted so that they were in the opposite side down position, and the gain was increased for 15 min. Adaptation for increases and decreases in gain continued alternately for 4 h. At the end of each hour of adaptation, the aVOR gain was measured in darkness, as described in the following text. When adaptation was completed the aVOR gain was tested in darkness over the next 1–2 h, and the animals were allowed to recover for at least for 48 h before the next experiment.

aVOR gain measurements

Two tests were utilized to measure gravity-specific effects on vertical aVOR gain adaptation [(eye velocity)/(head velocity)]. In the first test, shown by Fig. 2, insets, animals were oscillated sinusoidally at 0.5 Hz 60°/s in darkness about a pitch axis that was either upright or tilted toward side-down positions in roll in 10° increments up to 90°. Because the animals were always rotating in pitch, canal activation was the same in every head orientation in this test, but the direction of the average static pitch (otolith) component varied as a function of the head tilt. Dynamic otolith activation was maximal when the animals were oscillated about an upright position, and gradually decreased to zero as the axis of rotation was tilted toward side-down. In the second test, shown by Fig. 5, insets, the animals were sinusoidally rotated about a spatial vertical axis either while upright or statically tilted in roll with regard to the axis of rotation in 10° increments up to 90°. In this test, canal activation varied as a function of head orientation from yaw to pitch, while the otoliths were activated only statically by the head tilt. Due to the limitations of the equipment, a stimulus frequency of only 0.5 Hz (60°/s) was used for pitch axis rotation in the first paradigm. In the second paradigm, which utilized vertical axis rotation, stimulus frequencies ranged from 0.5 to 4 Hz. The peak stimulus velocity in the second test varied with frequency, being 60°/s at 0.5 Hz, 30°/s at 1.0 Hz, 15°/s at 2 Hz, and ≈7°/s at 4.0 Hz. In all instances, the animals were tested in all head orientations at the lowest frequency first and then at higher frequencies in ascending order.

FIG. 2.

Vertical eye velocities induced by pitch in different head orientations re gravity from left side down (LSD; A and F) to right side down (RSD; E and J) before adaptation (blue traces) and after adaptation (red traces). The stimulus velocity (black) was reversed to facilitate comparison. Insets: the head orientation in tilt during the test. The lines next to the monkey’s head indicate the axis about which head was oscillated. Star, head orientation in which the gain changes were induced. The direction of eye movement is shown by the double-headed arrow on the right of F (upward eye velocity was negative and downward eye velocity positive).

FIG. 5.

A–E: oscillation about a spatial vertical axis with the animal tilted left and right side down before and after the aVOR gain was increased in the RSD position. Stimulus velocities were reversed to facilitate comparison. Eye velocities before adaptation (blue traces) and after adaptation (red traces) were superimposed on the same graph along with reversed stimulus velocity (black traces). Insets: the head orientation in which the gains were tested. F–H: gain of the horizontal (F), vertical (G) and roll (H) aVOR before (blue symbols) and after (red symbols) adaptation. Positions in which the monkey was tested are shown in the inset below H.

Desaccaded eye velocity was fitted with sinusoids to estimate the gain and phase of the response in each head orientation (temporal gain and phase). Changes in gain were expressed as a percentage relative to preadapted level and plotted as a function of head tilt. To analyze the magnitude of the gravity-specific aVOR gain adaptation, we applied a cosine approximation with an unknown bias (C) to this residual function

$$y=A*\text{cos}(x+B)+C$$ (1)

where A is a magnitude of the gravity-specific effect and B is the angle of the head tilt (spatial phase) where the amplitude of the sinusoidal fit through the data were maximal.

Statistical analysis

Depending on the type of experiment, standard t-tests and ANOVA were used to analyze pairs or sets of data, respectively. A generally accepted statistical approach for data-model comparison, the χ² test, provided a robust statistical analysis if there were several hundred data points (Snedecor and Cochran 1967). This assumption failed if the sample size was small. An ANOVA is less sensitive to any nonnormality in the data distribution (Keppel 1991). To avoid possible complications in the statistical analysis of the goodness of the fit of the data with any model-predicted curve and for the data obtained before and after the gain adaptation, we utilized a reduced case of the ANOVA (F statistic) (Yakushin et al. 1995).

RESULTS

Oscillation about an interaural axis (pitch)

When unadapted animals were sinusoidally oscillated about an interaural axis in darkness, the induced vertical eye velocities were independent of the position of the axis of rotation with regard to gravity (Fig. 2, blue traces). Horizontal and roll eye velocities were negligible in all head orientations (not shown). After the vertical VOR gain was decreased with the animal in the LSD position, peak eye velocities were minimal when the animal was in the LSD head orientation (Fig. 2A, star, red trace), and peak eye velocity increased toward its original values as the animal was tilted toward the RSD position (Fig. 2, B–E, red traces). Similarly, when the vertical aVOR gain was increased in the RSD position, the gain changes were maximal when the animal was RSD (Fig. 2J, star, red trace) and there was a gradual decrease in gain toward preadapted values as the animal’s position was shifted from RSD to LSD (Fig. 2, J–F). There was an up-down asymmetry in the gain decreases; the changes were larger for downward (+) than upward (−) slow phases (Fig. 2, A–E). This may be related to differences in suppression and enhancement of upward and downward eye velocity observed among the animals (Fig. 1). This animal also had gain decreases in the downward direction that were present in all head orientations (Fig. 2, A–E). In contrast, gain increases were symmetrical for eye velocity in the upward and downward directions (Fig. 2, F–J).

The gain of the vertical aVOR was calculated for each head orientation in the unadapted state and plotted as a function of head tilt to obtain the spatial responses. Before adaptation, vertical aVOR gains were close to unity regardless of the angle of head tilt (Fig. 3, E). Small variations in gain around the upright position may have reflected a contribution of the lVOR to the response (Tomko et al. 1988), but there was no significant difference between approximations of the data with cosine functions that had a maximum in the upright position and a horizontal line (F statistic at P = 0.05). There was also no significant difference between the average gains over all head orientations for the unadapted state between the trials, which varied from 0.98 to 1.02 for the three animals tested for single-state adaptation, being on average 1.00 ± 0.03. After the vertical aVOR gain had been adaptively increased with the animals upright, the gains were higher (≈1.2) when the animals were tested while rotating upright about a horizontal axis than when the axis of rotation was tilted and they were rotating on their sides (Fig. 3, A–C, ●). Similarly, when the gain was adapted with the animals in LSD or RSD positions (Fig. 3, D–F and G–I, ●), the maximal gain increases were observed in these positions. Thus in all cases, the gain increases were maximal in or around the position in which the gains had been adapted and declined progressively as the head was oscillated in positions that deviated from the adaptation position.

FIG. 3.

Gains of the vertical angular vestibuloocular reflex (aVOR) in 3 animals during pitch, before (E) and after adaptation (●), plotted as a function of head orientation re gravity (abscissa). The aVOR gains were increased in the upright (A–C), LSD (D–F), and RSD (G–I) positions. Insets (on this and other figures): related to the positions on the abscissae and indicate head orientation in tilt during the test.

As shown in Fig. 3, the gain changes were not confined to a specific head position but were distributed over a wide range of head tilts relative to gravity. To characterize the distribution of these gain changes, the percent of the preadapted gain at each head orientation was plotted as a function of head tilt for both increases and decreases for the three animals (Fig. 4), and the data were fit as a summation of a constant value and a cosine function (Eq. 1). Applying this fit, we assumed that there are two types of adaptive changes that occurred simultaneously, a gravity-specific sinusoidal gain change with amplitude (A) and spatial phase (B), as described in the preceding text, and a constant value or bias (C) that was a gain change independent of head orientation. From this analysis, we determined the spatial phase of the observed changes relative to the head orientation in which gain was adapted (Table 1, Test 1). When the gain was increased in an upright position (0°), the average maximal gain change over all monkeys occurred at 6 ± 4° from the upright (Fig. 4A). When the gain was increased with animals on-side (+90° RSD; −90° LSD), maximal gain changes occurred 3 ± 17° from the position of adaptation (Fig. 4, B and C). Similarly, when the gain was decreased with the animals upright, changes were maximal −4 ± 13° from the upright (Fig. 4D), and when the gain was decreased in on-side positions, maximal changes occurred within 24 ± 10° from side-down positions (Fig. 4, E and F). Thus maximal gain changes occurred close to the position in which the gain changes had been induced and were well fit by a cosine function about this peak value (F statistic P < 0.05). Equation 1 could be simplified by assuming that the spatial phase of the sinusoid (B) was equal to the head orientation in which the gain was adapted. The spatial phase was chosen to be zero when the gain was adapted with the animals’ upright, −90° with the animals’ LSD and +90° with the animals’ RSD.

FIG. 4.

Average changes of the vertical aVOR gain (%) in 3 animals in various head orientations re gravity after the aVOR gain was increased (A–C) or decreased (D–F) with the animals upright (A and D), LSD (B and E), and RSD (C and F). The thick lines represent sinusoidal fits through the data. The error bars represent ±SD.

TABLE 1.

Gain changes after adaptation in different head orientations

	Adaptation Upright		Adaptation LSD		Adaptation RSD
	Test 1	Test 2	Test 1	Test 2	Test 1	Test 2
A. Amplitude of gravity-specific gain changes
Gain increase, %
Monkey M96012	13.2	0.4	9.9	15.4	6.3	8.9
Monkey M98060	12.8	2.2	7.0	11.4	15.7	15.4
Monkey M98078	9.6	2.4	10.5	9.5	10.3	11.3
Mean ± SD	11.9 ± 2.0	1.7 ± 1.1	9.1 ± 1.9	12.1 ± 3.0	10.8 ± 4.7	11.9 ± 3.3
Gain decrease, %
Monkey M96012	−11.1	4.7	−11.3	−11.2	−9.1	−8.6
Monkey M98060	−12.3	−1.7	−18.2	−16.2	−6.8	−12.2
Monkey M98078	−8.3	0.3	−7.7	−8.4	−9.2	−8.5
Mean ± SD	−10.6 ± 2.1	1.1 ± 3.3	−12.4 ± 5.3	−11.9 ± 4.0	−8.4 ± 1.4	−9.8 ± 2.1
B. Amplitude of non-specific gain changes
Gain increase, %
Monkey M96012	5.2	11.3	8.7	13.0	9.9	8.3
Monkey M98060	0.3	9.6	7.4	8.1	10.2	14.0
Monkey M98078	5.2	−2.0	7.2	10.1	10.0	7.6
Mean ± SD	3.6 ± 2.8	6.3 ± 7.2	7.8 ± 0.8	10.4 ± 2.5	10.0 ± 0.2	10.0 ± 3.5
Gain decrease, %
Monkey M96012	−9.3	−12.9	−7.4	−13.6	−11.8	−18.8
Monkey M98060	−11.4	−22.3	−29.1	−24.7	−15.0	−23.1
Monkey M98078	−10.9	−19.0	−17.0	−17.2	−22.4	−17.0
Mean ± SD	−10.5 ± 1.1	−18.1 ± 4.8	−17.8 ± 10.9	−18.5 ± 5.7	−16.4 ± 5.4	−19.6 ± 3.1

LSD, left side down; RSD, right side down. Oscillation was about an interaural axis (Test 1) or a spatial vertical axis (Test 2) while animals were upright or tilted at different angle in roll plane toward left or right side down.

Using the simplified Eq. 1 to fit the data, there was little bias when the gain was increased in the upright head position (≈4%, Fig. 4A) but a substantial offset after the gain was decreased (11%, Fig. 4D; Table 1B, Test 1). For side-down adaptation, the gravity-independent gain changes were also smaller when the gain was increased (≈9%, Fig. 4, B and C) than decreased (≈17%; Fig. 4, E and F; Table 2, Test 1). Thus gain changes that were independent of gravity were twice as large for gain decreases as for gain increases in every head orientation.

TABLE 2.

Gravity-specific (A) and non-specific (B) angular vestibuloocular reflex gain changes in three animals tested by rotation about a spatial vertical axis in various head tilts in the roll plane at frequencies that ranged from 0.5 to 4.0 Hz

Stimulus frequency	M96012		M98060		M98078
	LSD	RSD	LSD	RSD	LSD	RSD
A. Gravity-specific aVOR gain changes
Gain increase, %
0.5 Hz	18.5	9.6	12.5	18.9	7.7	12.8
1.0 Hz	18.2	7.6	8.3	17.2	7.6	10.7
2.0 Hz	12.7	11.0	12.0	12.7	12.6	8.3
4.0 Hz	12.3	7.2	12.8	12.8	10.1	13.3
Mean ± SD	15.4 ± 3.4	8.9 ± 1.8	11.4 ± 2.1	15.4 ± 2.1	9.5 ± 2.4	11.3 ± 2.3
Gain decrease, %
0.5 Hz	15.7	7.2	19.9	16.6	9.4	7.9
1.0 Hz	10.1	10.9	17.0	13.0	8.8	7.2
2.0 Hz	8.6	10.8	15.8	9.3	8.5	10.0
4.0 Hz	10.4	5.4	11.9	9.8	6.9	8.8
Mean ± SD	11.2 ± 3.1	8.6 ± 2.7	16.2 ± 3.3	12.2 ± 3.4	8.4 ± 1.1	8.5 ± 1.2
B. Nonspecific aVOR gain changes
Gain increase, %
0.5 Hz	16.3	9.8	9.9	19.8	15.5	15.1
1.0 Hz	12.2	8.9	10.7	16.7	10.7	7.8
2.0 Hz	10.2	12.8	6.0	10.3	9.3	3.5
4.0 Hz	13.2	1.8	5.8	9.0	4.8	3.8
Mean ± SD	13.0 ± 2.5	8.3 ± 4.7	8.1 ± 2.6	14.0 ± 5.2	10.1 ± 4.4	7.6 ± 5.4
Gain decrease, %
0.5 Hz	−14.7	−20.3	−24.4	−26.6	−16.8	−18.7
1.0 Hz	−13.4	−18.6	−27.9	−23.5	−16.0	−18.8
2.0 Hz	−10.4	−18.3	−22.6	−21.8	−17.9	−17.3
4.0 Hz	−15.8	−18.0	−23.9	−20.5	−18.2	−13.0
Mean ± SD	−13.6 ± 2.3	−18.8 ± 1.0	−24.7 ± 2.3	−23.1 ± 2.6	−17.2 ± 1.0	−17.0 ± 2.7

In contrast, the gravity-specific gain changes were comparable for upright and on-side gain increases and decreases. The average gravity-specific gain increase after adaptation in the upright head position was 11.9 ± 2% (Fig. 4A, Table 1A, Test 1). The gravity-specific gain changes were also comparable when the gains were increased in the LSD and RSD positions (9.1 ± 1.9%, Fig. 4B, vs. 10.8 ± 4.7%, Fig. 4C; Table 1A, Test 1; ANOVA, P = 0.591). When the gain was adaptively decreased in the upright head position, the maximal changes were also similar in the LSD and RSD positions (−10.6 ± 2.1%, Fig. 4D, vs. −10.1 ± 3.3%, Fig. 4, E and F, Table 1A, Test 1; ANOVA, P = 0.403), and the gravity-specific, on-side increases and decreases in gain were also comparable (average: 10.5 ± 1.6%). Thus visual suppression of the aVOR, which reduced eye velocity, induced greater gravity-independent adaptation than visual following that increased eye velocity. In contrast, the gravity-specific modulations in gain were independent of suppression or following.

Oscillation about a spatial vertical axis

In the preceding test, in which eye velocities were maintained along the pitch axis, the stimulus to the semicircular canals was constant, but dynamic stimulation of the otoliths varied. Dynamic stimulation was maximal when the animals were upright, but decreased toward zero as animals were tilted toward the on-side positions. We questioned whether the dynamic otolith input had affected the spatial distribution of the vertical aVOR gain adaptation. We also wished to determine whether the frequency of oscillation during testing was important because in the test described in the preceding text, the upper stimulus frequency used for testing was limited to 0.5 Hz due to equipment limitations. These questions were addressed by holding the stimulus to the otoliths constant during oscillation about a spatial vertical axis at frequencies from 0.5 to 4.0 Hz with the animal positioned upright or statically tilted in 10° increments in roll toward LSD and RSD. The paradigm is illustrated in Fig. 5, left, for oscillation at 0.5 Hz. In this test, stimulation of the vertical canals was the same as in the previous test when the animal was positioned LSD and RSD (Fig. 5, A and E). However, vertical canal activation and the amplitude of induced eye velocities were reduced as animal was reoriented toward upright (Fig. 5, B and D). In this experiment, the vertical aVOR gain was adapted in the RSD position (Fig. 5A). Despite this variation of eye velocity, as before, the difference in amplitude of the vertical component of the adapted (red traces) and the unadapted aVOR (blue traces) was maximal in the RSD position (Fig. 5A), and differences between the unadapted and adapted vertical components got smaller as the animals were shifted away from the RSD position (Fig. 5, B–E). As in the previous test, there were larger changes for upward eye velocity after the gain was increased (Fig. 5A). This may also reflect the ability of the animal to pursue OKN stimuli better when they were moving in the upward direction (Fig. 1B). Gains of horizontal (Fig. 5F), vertical (G), and roll (H) components were calculated for each head orientation and plotted as a function of head tilt before (blue traces) and after (red traces) adaptation. Horizontal and vertical gains were approximated with cosine functions, which had maxima for the horizontal aVOR in the upright position (Fig. 5F) and for the vertical aVOR in on-side head orientations (G). Torsional aVOR gains were negligible in all head positions (H).

Differences in vertical aVOR gain in each head orientation were expressed as a percent of the preadapted gain and fit with sinusoids (Eq. 1) to obtain the amplitude (A) and spatial phase (B) of gravity-specific changes and the bias (C) or gravity-independent gain changes (Fig. 6). Vertical responses within ±10° from the upright position were close to zero when the head was oscillated about the spatial vertical in upright. Thus data for this interval were omitted from the analysis. Eliminating this data following adaptation in upright, caused the fits to be erratic because the peak values were missing. Therefore the data obtained using this test after adaptation in upright were not utilized. The average spatial phase of context-specific changes, i.e., the head orientation in which gain changes were maximal, was not significantly different from the orientation in which the gain was adapted (15 ± 25°). This is close to the data obtained in the previous experiment where the changes were within 24 ± 10° from the position in which the gain was modified. When the gain was increased with animals in the on-side position (Fig. 6, A and B), the amount of gravity-specific changes were on average 14 ± 4% among the three animals that were tested. The variation between the animals was not significant (ANOVA, P = 0.297). When the gain was decreased in the on-side head orientation (Fig. 6, C and D), the average gain decrease (12 ± 3%) was similar to the changes observed after the gain increases. Gravity-independent gain changes were larger after gain decreases (19 ± 4%) than after gain increases (10 ± 3%, P = 6.77*10⁻⁴, t-test), similar to findings in Test 1.

FIG. 6.

Average changes in gain after adaptive gain increase (A and B) or decrease (C and D) in LSD (A and C) or RSD (B and D) using the Test 2 paradigm described in Fig. 4. See text for details.

Gravity-specific gain changes were similar at all frequencies that were tested after the vertical aVOR gain was decreased or increased (Fig. 7, A and B). Gain changes varied as a function of frequency in individual monkeys, but on average, there was no significant trend in either gravity-specific gain increases (Fig. 7A, ANOVA, P = 0.773) or decreases (Fig. 7B, P = 0.340). Gravity-independent gain increases (Fig. 7C) were correlated with the frequency of stimulation, being bigger at low frequencies (15% at 0.5 Hz) than at high frequencies (8% at 4 Hz; ANOVA, P = 0.007), but the gravity-independent gain decreases were not correlated with frequency (Fig. 7D, P = 0.790). After on-side adaptation, the gravity-specific changes observed during oscillation about a vertical axis were, on average, similar to the data obtained by oscillation about a horizontal axis (compare Tests 1 and 2 in Table 1, t-test, P > 0.06).

FIG. 7.

Gravity-specific (A and B) and nonspecific (C and D) vertical aVOR gain changes tested at different frequencies after gain increase (A and C) and decrease (B and D).

Thus results were similar regardless of whether or not there was dynamic otolith stimulation during the oscillation about the head orientation in which the adaptation had occurred. Moreover, the frequency at which the animal was tested did not affect the adapted spatial gain distribution. It is important to note that the gain adaptation of the vertical component of eye movement had similar behavior when the animals were tilted in the roll plane and rotated around a spatial vertical axis as when they were tilted in the roll plane and rotated around the pitch axis.

Simultaneous gain increases and decreases (dual-state adaptation)

We next determined how the gravity-specific gain change distributions would be altered by imposing gain changes at two alternate head orientations (LSD and RSD). The aVOR gain was increased with the head in one side-down and decreased with the head in contralateral side down position in single experiments (dual-state adaptation). When gain was concurrently increased in LSD while decreased in RSD, dual-state changes were observed after 1 h of adaptation, although initially decreases had larger percent changes than increases (Fig. 8A). As adaptation proceeded to 4 h, dual-state gravity-specific changes gradually increased and the asymmetry between gain increases and decreases declined or disappeared (Fig. 8, B–D). The data for dual-state adaptation were fit by a sinusoid as in the preceding text. The gravity-independent gain changes were within ±4% of the preadapted gain for all animals at any point in the adaptation process (Fig. 8E), with an average of about −1% (Fig. 8E, thick line). The temporal evolution of the gravity-specific gain changes varied among animals, but all gain changes increased monotonically over 4 h (Fig. 8F). For three of four of the animals tested, the gravity-specific gain changes for the dual-state adaptation were approximately the same as during the single-state adaptation. In one animal (M98060), the gravity specific gain change in the dual-state adaptation paradigm was double that for single-state adaptation (cf. Fig. 8F, squares, and Table 1). However, only two of four animals adapted for dual state were also adapted for single state.

FIG. 8.

A-D: gain changes in response to different durations of dual-state gain adaptation, with increases in LSD and decreases in RSD. ●, the specific values for each head orientation in roll during postadaptive testing and the solid lines represent a sinusoidal fit to the data. E and F: nonspecific (E) and gravity-specific (F) gain changes as a function of time of adaptation obtained in 4 animals (M96012, circles; M98060, squares; M98064, triangles; M98063, inverted triangles). Open symbols: gain increases with the animals RSD and decreases with animals LSD. Closed symbols, gain increases with animals LSD and decreases with animals RSD. The thick line in E is the average value.

To understand the nature of the gravity-specific gain changes as a result of the multiple-state adaptation better, we summated the distributions for RSD and LSD adaptations obtained for gain increases and decreases during single-state adaptation and compared the resultant distributions with those obtained during dual-state adaptation (Fig. 9). For monkey M98060, the summation of increased LSD and decreased RSD gains (Fig. 9A) for single-state adaptation were comparable to the data for dual-state adaptation for gain decreases, but the gain increases were higher for dual-state adaptation (Fig. 9E, open and filled symbols, respectively). There was a residual, which was close to zero for tilts in the right quadrant (Fig. 9I), but the differences became larger (≈10%) for LSD (Fig. 9I). This was due to the shift of the dual-state adaptation, which reduced the bias and made the distribution symmetrical. Similar results were obtained when data for gain increase in RSD was summated with gain decrease in LSD in the same animal (Fig. 9B). For this condition, the differences between summation and the results of dual-state adaptation (Fig. 9F, open and filled symbols, respectively) were substantial for all head orientations, although the maximum residual occurred in LSD (Fig. 9J) as for the previous comparison (Fig. 9I). These results were the same when the averaged data after single-state adaptation (Fig. 9, C and D) were summated (Fig. 9, G and H, open symbols) and compared with the results of dual-state adaptation (Fig. 9, G and H, filled symbols). An observed average residual difference across all animals was about −10% (Fig. 9, K and L). These comparisons indicate that dual-state adaptation imposes a superposition of the gravity-specific effects, excluding the nonspecific changes in gain (bias changes).

FIG. 9.

A–D: single-state adaptation of the vertical aVOR for M98060 (A and B) and for all 3 animals (C and D). The titles above show the positions in which the animals were adapted for gain increases (Up) and decreases (Down). Results of 2 separate experiments were superimposed on each graph. The gain was increased in 1 side-down position in 1 experiment while in another it was decreased with the contralateral side down, and the data were fit with sinusoids. E–H: summation of the gain changes shown in A–D (open symbols) compared with the results of dual-stated adaptation to increases and decreases in the gains in the corresponding directions (filled symbols). I–L: residuals of the gain curves shown in E–H for the corresponding figures.

DISCUSSION

The major finding of this study is that when the gain of the vertical aVOR was increased or decreased in the upright or side-down positions, maximal gain changes occurred when the animals were tested in the head orientation in which the gain was adapted, and the changes decreased continuously as the head was deviated from this orientation. The spatial distributions of the gain changes were broadly tuned and were well approximated by cosine functions, the peak phases of which were close to the angle of head orientation at which the adaptation took place. The sinusoids that described these changes for gain increases and decreases were of similar magnitude, regardless of the position in which the gain had been adapted. Neither dynamic otolith stimulation nor the frequency of oscillation affected the observed gravity-dependent changes. The gravity-dependent gain changes in this study are comparable to those in our previous study (Yakushin et al. 2000a). Amount of gravity-dependent gain increases and decreases obtained after dual-state adaptation in general were comparable to gain changes observed after single-state adaptation. The parameters of the cosine functions were similar, regardless of how the position of adaptation was approached. Thus the gravity specific changes in gain are modifications of aVOR adaptation in response to orientation of the head in three dimensions. We conclude that orientation to gravity is an integral part of the adaptive process of the aVOR and plays a significant role in the expression of the adapted gains. A second, adaptive gain change that was independent of orientation of head position with regard to gravity was also induced by single-state adaptation. Such gain changes were substantial when the gain was reduced, but this bias disappeared if increases and decreases in adaptation were induced simultaneously.

The finding that the adaptive changes were well fitted with a cosine function superimposed on a bias value is of considerable theoretical significance. It shows that the gravity-dependent gain changes are not narrowly tuned to the position of adaptation but are broadly tuned, extending over 180° from side-down and upright positions. This implies that during the process of adaptation, a memory of head orientation is stored in association with the adapted state and is expressed in proportion to the gravitational acceleration, as the head is reoriented relative to gravity. Because the process is continuous, it presumably involves the totality of the network of canal-otolith interaction over the span of polarization vectors. Of note, the gravity-dependent changes were approximately the same in all conditions (mean: 10.5 ± 1.6%), whereas the gravity-independent changes were smaller for gain increase (≈10%) than for gain decrease (≈19%). Additionally, after the gain had been modified with the animals upright, large gravity-dependent gain changes could be associated with minor gravity-independent changes and vice versa. Finally, the bias component in the dual-state adaptation paradigm went to zero, despite the fact that the sum of each adapted state did not summate to zero. These findings strongly suggest that the gravity-specific and nonspecific changes represent two independent processes that are likely to be coded in separate parts of the vestibular system.

How these processes are organized centrally is not known. The flocculus of the vestibulo-cerebellum is known to control aVOR gain adaptation (Lisberger and Fuchs 1978; Partsalis et al. 1995b; Zee et al. 1981) and is probably responsible for the gravity-independent component, i.e., the bias value. The flocculus has a three-dimensional zonal organization (Ito et al. 1977; Van der Steen et al. 1994; Yamamoto and Shimoyama 1977) that is responsible for eye movements in canal coordinates, but whether the flocculus is involved in the gravity specific process is not known. This seems unlikely, because there is no direct otolith input to the flocculus, and as yet, static otolith input has not been demonstrated through secondary vestibulo-floccular connections. Alternatively, the nodulus and rostral ventral uvula could be involved in this process, because both structures have direct and indirect otolith input and these structures are known to control spatial orientation of eye velocity in the aVOR (Wearne et al. 1996, 1998). Proprioceptive pathways could also be involved in the process of induction of the gravity-dependent changes (Yates et al. 2000). Because the eye movements induced by the aVOR are generated in the vestibular nuclei, however, the effects of the gravity-dependent adaptation must be expressed through the neurons that constitute the aVOR. Such cells are located in the superior and medial vestibular nuclei and the Y group (Lisberger and Miles 1980; Partsalis et al. 1995a). The extensive convergence between otolith and canal inputs (Angelaki et al. 1993; Bush et al. 1993; Endo et al. 1995; Kubo et al. 1977; Kushiro et al. 2000; Ono et al. 2000; Sato et al. 2000; Uchino et al. 2000; Wilson et al. 1990; Zakir et al. 2000) could provide the substrate for the gravity-dependent adaptive behavior.

Other studies have also indicated that head position re gravity can influence VOR gain adaptation. The shift in the plane of the eye movements was greater during cross-axis adaptation when it was tested in the head orientation in which the adaptation was induced (Baker et al. 1987a). A similar effect of head tilt on the cross-axis gain adaptation was observed when the animals were oscillated alternatively on one side, then on the other side (Baker et al. 1987b). In that study, the animals were adapted for 2 h, and their head orientation was altered from the left side down to the right side down every 10 min. When animals were on one side, the upward head rotation was in phase with a rightward optokinetic stimulus, and when they were on the opposite side, the upward head movements were in phase with leftward optokinetic stimulation. After adaptation, the direction of the induced oblique eye movements depended on the head orientation during testing. However, cross-axis gain adaptation is a special case of aVOR gain adaptation (Schultheis and Robinson 1981) because disparate planes of the head and surround movement can only occur in artificial conditions. Tan et al. (1992) and Tiliket et al. (1993) demonstrated that when the aVOR gain was modified in humans, with the head tilted 45° forward or 45° left ear down, changes in horizontal aVOR gain were greater when the subject was tested in the head orientation that was used for adaptation, rather than when it was tilted 45° toward the opposite side. Because different canals were stimulated in this experiment with the head forward and backward or tilted to the left and right, it was possible that the observed gain changes could have been either in the context of the head position with regard to gravity or in the context of which set of canals was stimulated. The current experiments using rotation about a horizontal axis rule out this possibility since there was context-specific adaptation with the head in specific head positions re gravity with stimulation of the same set of semicircular canals.

The gain changes induced by forward-backward oscillation about the upright position (±75°) had the same amplitude and distribution of the gravity-dependent gain changes as when animals were adapted on-side. Thus the patterns of gain change were the same from significantly different patterns of dynamic otolith stimulation, realized in the upright and on-side positions. This demonstrates that the gravity-specific changes were independent of dynamic otolith activation. The finding that the gravity dependent gain changes were similar over all test frequencies from 0.5 to 4 Hz (Fig. 7) is likely due to the fact that the steps of velocity used to adapt the aVOR gain were composed of a wide spectrum of frequencies.

The functional significance of this gravity-dependent adaptation may be related to the development of a feedback control strategy to improve the performance of an inherently open-loop aVOR pathway. It has previously been shown that the rapid component of the aVOR is coded in head coordinates and has no spatial orientation components as does velocity storage (Wearne et al. 1997). Thus just as nonspecific gain adaptation changes the gain through visual feedback to adjust the open loop gain properties of the aVOR, gravity-specific adaptation imposes a spatial frame on this otherwise spatially devoid open loop system. Such spatial dependence has been shown in other contexts. Tomko pitched cats in various orientations with regard to gravity and found that the gain of the vertical aVOR was slightly but significantly increased when the cats were upright. He attributed the difference in gain between the side down and upright positions to the combined activation of the otoliths and semicircular canals when the cats were upright (Tomko et al. 1988). Interpreted in another way, his results could also indicate an increase in adapted gain when the cats were in the upright position, the position in which maximal vertical aVOR gain changes were realized. There was no consistent cross-animal change in gain in the upright condition in our monkeys before adaptation (Fig. 3, ○). Because the brain cannot know the direction of gravity in the presence of other linear acceleration, it only senses the summed vector of the linear accelerations, i.e., the gravito-inertial acceleration (GIA) (see Raphan and Cohen 1996, 2002 for review). Thus the observed phenomena could keep the gain of the vertical VOR at a constant level during tilts of the GIA with regard to the head that occur while turning corners (Imai et al. 2001). Similar centrally induced orientation of velocity storage, also causes eye velocity to orient toward the GIA. Insights as to the role of the gravity-specific changes could come from study of adaptive changes in the vertical aVOR in microgravity, where there is orientation to linear acceleration (Moore et al. 1999), but the sensed gravity is negligible (<10⁻⁶ g).

The finding that the non-gravity-specific gain decreases were larger than the gain increases is likely related to control of retinal slip under the different paradigms used to develop the in-phase and out-of-phase changes in gain (Gonshor and Melvill Jones 1976a,b; Ito and Miyashita 1975; Miles and Lisberger 1981b; Yakushin et al. 2000b). When the gain was adaptively decreased, animals were rotated with steps of velocity in a self-stationary surround, and visual suppression began simultaneously with the onset of rotation. When the gain was increased, however, visual enhancement began 2 s after the onset of the step of velocity. Therefore maximal activation of the cupula occurred slightly before the appearance of the enhancing visual stimulus. Moreover, because the aVOR and visual-induced eye velocities were in opposite directions during gain enhancement, the retinal slip was as large as 60°/s, which could exceed the linear range of activation of ocular pursuit (Yakushin et al. 2000b) or of the fast component of OKN (Waespe et al. 1983) (Fig. 1D). Regardless of these differences between enhancement and reduction of gain, enhancement was reliably produced by the adaptation procedure in all instances, and adaptation for both enhancement and reduction of gain were the same in all head positions in which the gains were adapted. Additionally, since sinusoidal rotation was used to measure gain adaptation, the results of both enhancement and reduction of gain could be directly compared. Still unexplained are the finding that the bias component disappeared in each case when the adaptation for both increase and decrease in gain was done on the two sides.

In summary, we have demonstrated a phenomenon that is likely to be a fundamental property of aVOR gain adaptation which is linked to the orientation of the head with regard to gravity. This adaptation was shown to be one of two processes responsible for adaptation of the gain of the aVOR. It is important as a context for aiding aVOR compensation under static head orientation when the otoliths are not dynamically activated.