INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Head movements in a gravitational environment induce linear and angular head accelerations in three dimensions. The vestibular system stabilizes vision by generating a combined linear and angular vestibulo-ocular reflex from the otoliths and semicircular canals. The angular vestibulo-ocular reflex (AVOR) generates compensatory eye rotations to head rotations (Aw et al. 1996a,b, 2001; Cremer et al. 1998; Halmagyi and Curthoys 1988; Tabak et al. 1997), while the linear vestibulo-ocular reflex (LVOR) generates compensatory eye rotations to head translations (Angelaki et al. 2000a; Bronstein and Gresty 1988; Gianna et al. 2000; Lempert et al. 1998; Paige and Tomko 1991; Paige et al. 1998; Telford et al. 1997) and to head orientation with respect to gravity (Paige and Tomko 1991). LVOR is enhanced by near viewing during head translation (Gianna et al. 1997; Paige et al. 1998; Schwarz et al. 1989; Telford et al. 1997).

In normal subjects, head movement stimulates both labyrinths simultaneously. Consequently, the contribution of each labyrinth to the VOR can only be determined by studying the VOR in subjects with just one functioning labyrinth. During high-acceleration head rotations, about 80% of the AVOR arises from the ipsilateral semicircular canal, while about 20% arises from the contralateral semicircular canal (Aw et al. 1996a,b, 2001; Cremer et al. 1998; Halmagyi and Curthoys 1998). Until now, it has not been shown in humans if the same applies to the LVOR arising from the otolith system. The morphological arrangement of the hair cells in the ampullary crista of one semicircular canal are polarized in a single direction that is opposite in direction to the polarization of hair cells in its coplanar contralateral semicircular canal (Lindeman 1969). Hence, when the hair cells in the ipsilateral semicircular canal ampulla are depolarized, those in the contralateral semicircular canal are hyperpolarized. However, in each otolith, the hair cells have opposite polarities across the striola on the macula (Lindeman 1969). As such, the residual responses of the LVOR after the loss of function of one labyrinth may be quite different to the AVOR.

Impulsive stimuli have been used in this study to examine the changes in the LVOR after unilateral vestibular deafferentation (UVD) because the LVOR only shows a robust response to higher frequency stimulus (above approximately 0.5 Hz) (Angelaki et al. 2000a; Telford et al. 1997). There is some evidence that the LVOR evoked by impulsive stimuli is reduced after UVD. Crane and Demer (1998) examined the combined linear and AVOR responses in humans after UVD by using impulsive eccentric yaw whole-body rotations. Although they found a VOR deficit after UVD, the deficit could not be separately attributed to either the linear or the AVOR because the responses to the stimulus were coaxial. Lempert et al. (1998) used a lower linear acceleration stimulus (0.24g) and measured the VOR at 300-500 ms after stimulus onset, which allowed smooth pursuit to augment the LVOR. They showed that the LVOR diminished in humans immediately after UVD, but recovered in the chronic state. Angelaki et al. (2000b) demonstrated in rhesus monkeys that LVOR gain was bilaterally and asymmetrically reduced to impulsive translation stimulus in the acute stage after UVD and that the asymmetry diminished considerably after 3 mo. Using a similar impulsive translation stimulus of about 0.5g, Ramat et al. (2001) tested two subjects after incomplete inactivation of the labyrinthine function with intra-tympanic gentamicin. They demonstrated a reduced symmetrical LVOR in one subject and a bilaterally reduced asymmetrical response in the other subject.

This study aimed to investigate the effects of UVD on the LVOR evoked by an impulsive, whole-body rotation in the roll plane, with the eye positioned 815 mm eccentric to the rotation axis. During this whole-body rotation, the labyrinth was stimulated by an inter-aural linear acceleration of about 0.55g and a roll angular acceleration of about 360°/s². While our linear stimulus was comparable in magnitude to that used by Crane and Demer (1998), we were able to distinguish the LVOR response from the AVOR because our linear and angular stimuli were not coaxial. In this study the response from the LVOR was a horizontal eye rotation, while the response from the AVOR was a torsional eye rotation.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

We studied 11 healthy normal subjects (mean age: 45.0 yr; range: 31-62 yr) with no history of vestibular disorder. These subjects have normal, symmetrical responses to the caloric test and normal individual semicircular canal responses to the high-acceleration head impulse test. We also studied six left and one right UVD subjects (mean age: 50.9 yr; range: 27-68 yr). The seven subjects had undergone UVD either during surgical removal of an acoustic neuroma (5 subjects) or as treatment for Ménière's disease (2 subjects). Six of the UVD subjects were tested more than 2 yr post-UVD and one UVD subject was tested 3 mo after the neurectomy. The UVD subjects do not have any facial palsy or exposure keratopathy. Caloric responses and head impulse tests of individual semicircular canal function were all absent on the deafferented side. The vergence capabilities of the UVD subjects were examined by inspection and were found to be normal. All subjects gave informed consent. The protocol was approved by the Royal Prince Alfred Hospital Human Ethics Committee, in accordance with the Helsinki II Declaration.

Experimental setup

The subjects were stimulated in a manually operated two-axis rotator consisting of an outer ring, which rotated about an earth-horizontal axis, and an inner tube, which rotates about an axis perpendicular to the outer ring. The search coil transmitter, mounted in a cube-like configuration, was attached to one end of the inner tube. Left eye positions were recorded in three dimensions with dual scleral search coils (Skalar, Delft, The Netherlands) using the search coil technique (Robinson 1963) in a two-field magnetic search coil system (660 mm cube, 66 and 100 kHz; CNC Engineering, Seattle, WA). The search coil signals were obtained after preamplification and phase detection with a setting of 3.0 × 0.1 ms, which translated to a 530 Hz3 db position bandwidth with 3 min of arc peak-to-peak noise for a 30° eye movement (CNC Engineering). These signals were then low-pass filtered with anti-alias filters of 0-100 Hz bandwidth and sampled at 1 kHz with 16-bit resolution. The resolution of the recording system was better than 0.1 min of arc. Maximum errors and cross-coupling were <2% (Aw et al. 1996b). The dual scleral search coils were precalibrated in three axes (horizontal, vertical, and torsional) with a Fick gimbal. The gimbal was moved in yaw, pitch, and roll calibration positions ±20° in 5° steps, and the gains and offsets for each coil were determined. We assumed that gains and offsets of the eye coil were the same during in vivo (i.e., with the coil on the eye) and in vitro calibrations. Fixation spots produced by a series of LEDs 2 mm in diameter were located 200, 300, and 600 mm from the center between the subject's eyes. Linear acceleration of the head was recorded in three axes by a tri-axial linear accelerometer (ADXL105, Analog Devices, Norwood, MA) and a capacitative beam device with a frequency response of DC to >2,000 Hz, mounted on the subject's dental impression bite-bar. Signals from the tri-axial linear accelerometer were also low-pass, anti-alias filtered with 0- to 100-Hz bandwidth filters and sampled at 1 kHz with 16-bit resolution using the same system as the search coils. Linear accelerometer signals were calibrated in three axes in a range of ±90° in 10° steps using a gimbal. There was no discernible time delay measured between the dual search coils and the linear accelerometer mounted on a gimbal when the combined linear and angular stimulus was applied. With the subject's head locked to the rotator by the full face mask, the head was only able to rotate in roll during the stimulus. Roll angular head position was sampled synchronously at 12-bit resolution from a shaft encoder (Hengstler, Aldingen, Germany) mounted on the outer ring of the two-axis gimbal. Signals from the dual search coils, tri-axial linear accelerometer, and shaft encoder were all sampled at 1 kHz using in-house data acquisition software written in Labview Version 5.0 (National Instruments, Austin, TX) on a WinNT-based PC.

Experimental protocol

The subject was restrained to a bed using a full-face mask, chest, lap, thigh, and ankle belts and also inflatable air bags along the longitudinal axis of the inner tube such that the subject's eyes were positioned 815 mm from the rotation axis and in the center of the search coil transmitter. To prevent motion-induced artifact during head rotation, the subject's head was secured to the head-holder with an individually molded form-fitting full-face thermoplastic mask (Polyflex II, Roylan Thermoplastic Splinting System, Smith and Nephew, London, UK) bolted to the head holder (Fig. 1A). The dual search coils were placed onto the subject's left eye after application of topical anesthesia (Alcaine 0.5%, Sydney, Australia). Before each experiment, eye movements were recorded for 1 min in darkness to check for spontaneous nystagmus. Throughout the whole-body roll rotation, the subject fixated on an earth fixed green LED in dim illumination. The target was centered between the two eyes at viewing distances of 200, 300, and 600 mm in three separate trials.

View larger version (59K): [in this window] [in a new window]   Fig. 1. A: subject was securely restrained in a manually operated 2-axis rotator during the experiment with the eyes in the center of the transmitter field coils. The subject's head was secured to a head-holder by an individually form-fitted, full-face mask bolted to the head-holder, so that the subject's left eye was 815 mm from the rotation axis. The target was centered between the 2 eyes at viewing distances of 200, 300, or 600 mm. B: typical rightward inter-aural linear head acceleration stimulus (Hacc) during a clockwise roll head rotation (Hacc), from a normal subject. The cartoons on the figure illustrate the roll head position and inter-aural linear head acceleration at stimulus onset and at 2 points after the stimulus onset. At stimulus onset the subject was upright and stationary with 0 roll head position. In this trial, the stimulus reached a peak linear head acceleration of 0.53g at 70 ms after stimulus onset and the roll head position at this point was 0.61°. At 100-ms stimulus onset, the linear head acceleration was 0.46g and the roll angular head position was 1.25° .

Stimulus

The stimulus used was a directionally unpredictable, transient, passive, impulsive eccentric roll rotation. The term roll referred to the direction of rotation defined with reference to the subject. Figure 1B shows an example of a typical trial during a rightward acceleration stimulus in a normal subject. When the head is rotated in roll eccentric to the rotation axis, in addition to the roll angular acceleration stimulus, a tangential and centripetal linear acceleration stimulus is also exerted on the head. This tangential linear acceleration varies inversely with centripetal acceleration as a function of head eccentricity to the rotation axis. In Fig. 1B, the subject's labyrinth was stimulated with a peak tangential linear acceleration along the inter-aural axis of 0.53g, a minute centripetal linear acceleration along the dorsal-ventral axis of 0.02g, a small tilt varying from 0° to 1.25°, and a peak roll angular acceleration of 365°/s² (about 10% of the angular head acceleration used during the "standard" head impulse test). As the centripetal linear acceleration was minute it was not considered in this study. Therefore during the first 100 ms, the stimulus was essentially an inter-aural linear acceleration with a small angular acceleration. Each subject was tested in two directions with reference to the subject: clockwise roll rotation with rightward acceleration and counterclockwise with leftward acceleration. Each subject was tested for a total of 60 trials, i.e., 10 trials per stimulus direction at each viewing distance. All normal and UVD subjects were tested with the same protocol. Peak inter-aural linear acceleration in normal subjects was 0.51 ± 0.02g (SD) at 77 ms after stimulus onset and in UVD subjects was 0.53 ± 0.03g at 78 ms after stimulus onset. The outputs of the linear and AVOR were not coaxial, i.e., the responses of the LVOR were predominantly in yaw, while the responses of the AVOR and tilt were in roll.

Seven of the 11 normal subjects (range: 31-57 yr) and 5 of the UVD subjects (range: 40-68 yr) were also tested in total darkness using the same stimulus as above but with the 600 mm viewing distance target extinguished just before the onset of the eccentric roll rotation. One of the normal subjects and one of the UVD subjects were unable to perform the experiment properly.

Data analysis

Our experimental data were analyzed off-line using custom in-house software written in Labview Version 6.0 (National Instruments) and Splus (MathSoft, Seattle, WA). Three-dimensional (3-D) eye positions in head-fixed coordinates were determined from the search coil voltages and expressed in Euler angles (Haslwanter 1995; Hess et al. 1992). The three components of eye rotation were referred to as torsional, vertical, and horizontal components expressed with respect to the subject and the directions left, down, and clockwise were positive. We determined the horizontal ideal eye position in space-fixed coordinates (i.e., when the eye is looking directly at the target) from geometrical analysis of the setup, head roll position, and known fixation distance. If there was a large blink artifact observed in the response, then that trial was excluded by visual inspection. For each stimulus direction (i.e., rightward or leftward acceleration), a minimum of 6 and a maximum of 10 trials were used for the data analysis. A preliminary stimulus onset for each trial was determined as the time at which the derivative of the inter-aural linear head acceleration signal exceed a threshold value of 0.02g/s. For each trial, the preliminary onset time was used to extract a total of 600 ms of data, commencing 100 ms prior to the preliminary onset.

Spectral analysis of the linear acceleration stimulus signal was performed using a windowed fast Fourier transform (fft). The analysis showed that the majority of the frequency content was in the range of 0-5 Hz, with no signal presence above 20 Hz, supporting the use of the low-pass anti-alias filtering range of 0-100 Hz.

Latency of the LVOR was determined with an automated algorithm as the difference in time between the onset of the inter-aural linear head acceleration stimulus and the onset of the horizontal eye velocity of the LVOR. The onset of each signal was independently determined as the point where the derivative of each signal first exceeded 1 SD of its baseline noise in each trial. The algorithm for computing the latency in each trial was as follows. To detect the onset, a 300-ms slice of linear head acceleration and horizontal eye velocity signal were extracted each head rotation trial (starting 100 ms before the onset of head rotation) and differentiated. The baseline noise was determined as 1 SD of the differentiated signal in a 70-ms region between 30-100 ms prior to the onset of head rotation. The differentiated signal was fitted with a 5th-order polynomial fit. The onset of each signal was then determined as the time when the polynomial fit exceeded 1 SD of the baseline noise. The means ± 1 SD of the normal and UVD groups were computed from the mean latency of each subject in the group.

The acceleration gain of the LVOR was measured for each trial as the ratio of the slope of a line fitted through the horizontal eye velocity data compared with the slope of a line through the calculated ideal horizontal eye velocity data during a 30-ms period starting 70 ms after stimulus onset (Clendanial et al. 2001). The acceleration gain was determined using an automated algorithm in Labview. The period between 70 and 100 ms was chosen because it is in the region of the LVOR response that we are interested in and the response in this portion of the data are fairly linear. The mean of the acceleration gain of the LVOR from 6-10 trials was determined for each stimulus direction at each viewing distance. The group means ± SD of the acceleration gains were then calculated for the normal and UVD subjects.

Statistical analysis

Mean values of the responses from 6-10 trials per stimulus direction at each viewing distance were determined in each subject. The group means ± SE or SD of the normal and UVD subjects were then calculated from the mean of each subject in the group. Student's t-test for differences between two means of independent observations was used to test for differences between normal and UVD subjects using a significance level of P = 0.05 (Winer et al. 1991). Student's t-test for differences between two means of dependent observations was used to test for differences between ipsilesional and contralesional sides within UVD subjects using a Bonferroni adjusted level of significance of (new P)  = 0.025 to minimize the occurrence of type one error due to multiplicity of tests (Sankoh et al. 1997). ANOVA was used to determine if the variations in the latency of occurrence of the initial saccade between the three viewing distances of 200, 300, and 600 mm was significant in normal and UVD subjects.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

3-D responses in normal and UVD subjects

Representative examples of 3-D eye movement responses from a normal and a left UVD subject to leftward and rightward linear head acceleration stimuli are shown in Fig. 2. During the experiment, the subject's left eye was 815 mm from the axis of rotation and 200 mm from the fixation target. The mean peak inter-aural linear head acceleration was about 0.5g. Since this stimulus combined a linear acceleration along the inter-aural axis with an angular acceleration about the naso-occipital axis, the eye movement response observed comprised both linear and AVOR components.

View larger version (33K): [in this window] [in a new window]   Fig. 2. Comparisons of 6 trials each of leftward and rightward inter-aural linear head accelerations (Hacc) and eye rotations at 200-mm viewing distance from a normal and a left unilateral vestibular deafferented (UVD) subject. Counterclockwise (CCW) rotation produces leftward (ipsilesional in UVD subject) head acceleration, while clockwise (CW) rotation produces rightward (contralesional in UVD subject) acceleration. In the normal subject, maximum inter-aural linear leftward and rightward head accelerations were 0.41g and 0.45g. In the UVD subject, the maximum ipsilesional and contralesional accelerations were 0.51g and 0.59g. Horizontal, vertical, and torsional components of eye position (Ehor, Ever, Etor), inverted roll head position (Hroll), and derived ideal horizontal eye position (IEhor[infi],) are shown in the top panels; eye velocities (hor, ver, tor) are shown are shown in the bottom panels.

As the stimulus consisted of predominantly lateral motion, the LVOR observed was comprised of large horizontal eye movement responses and only minute vertical eye movement responses. In the normal subject, apart from the initial period directly after the onset of head motion, the horizontal eye position matches the ideal horizontal eye position reasonably, with little eye position error. The initial saccade to correct any eye position error was observed later than 100 ms after stimulus onset. A fairly robust horizontal eye velocity was observed during rightward and leftward acceleration before the onset of first saccade. At 100 ms after stimulus onset, mean horizontal eye velocity was: 40.1 ± 0.2°/s (SE) for leftward linear acceleration and 40.4 ± 0.1°/s for rightward linear acceleration. In contrast, the horizontal component of the LVOR response was diminished bilaterally in the left UVD subject, especially during the first 100 ms. Consequently, horizontal eye position could not match the ideal horizontal eye position resulting in a large eye position error. In this UVD subject, the initial saccade occurred later, at about 150 ms after stimulus onset. In other UVD subjects, we observed the initial saccade to be between 120 and 150 ms after stimulus onset. Although the UVD subject in Fig. 2 made multiple corrective saccades after 140 ms, she was unable to achieve target fixation even at 400 ms after stimulus onset.

Additionally, there was a compensatory torsional eye movement response from the AVOR and small head roll. Unlike the response to the angular roll head rotation during the "standard" head impulse test, in this study we observed a period of almost zero response between 35 and 70 ms in normal subjects, followed by compensatory mean torsional eye velocities of 5.1 ± 0.2°/s and 9.6 ± 0.1°/s for leftward and rightward acceleration, respectively. In the UVD subject, the mean torsional eye velocity was -14.9 ± 0.2°/s for leftward linear acceleration and 5.3 ± 0.2°/s for rightward linear acceleration at 100 ms from stimulus onset.

LVOR after unilateral vestibular deafferentation

The LVOR responses in the 7 UVD subjects were compared with the 11 normal subjects using the mean horizontal eye velocity response (Fig. 3, Table 1) for all three viewing distances with comparable linear head acceleration. The mean peak inter-aural linear acceleration during the stimulus was 0.51 ± 0.02g, (SE) in normal subjects and 0.53 ± 0.03g in UVD subjects. At 100 ms from stimulus onset, the mean horizontal eye velocity of the LVOR in UVD subjects was significantly lower (P < 0.05) than that of normal subjects at all three viewing distances of 200, 300, and 600 mm, even though the peak linear acceleration was slightly lower in normal than in UVD subjects. This difference in horizontal eye velocity responses between normal and UVD subjects decreased as the viewing distance increased (Table 1). The ipsilesional LVOR in UVD subjects was significantly different from normal subjects at 60 ms from stimulus onset for 200 mm viewing distance, at 76 ms from stimulus onset for 300 mm viewing distance, and at 74 ms from stimulus onset for 600 mm viewing distance. The contralesional LVOR in UVD subjects was significantly different from normal subjects at 46 ms from stimulus onset for 200 mm viewing distance, at 61 ms from stimulus onset for 300 mm viewing distance, and at 76 ms from stimulus onset for 600 mm viewing distance. The difference between ipsilesional and contralesional mean responses was greater at near viewing. At 200 mm viewing distance, mean ipsilesional LVOR response was 80% of contralesional response, at 300 mm viewing distance, mean ipsilesional LVOR response was 88% of contralesional response, and at 600 mm viewing distance, mean ipsilesional LVOR response was 99% or almost equal to contralesional response. However, this difference did not achieve statistical significance at any viewing distance.

View larger version (24K): [in this window] [in a new window]   Fig. 3. Comparisons of the group means ± SE of the horizontal eye velocity of the linear vestibulo-ocular reflex (LVOR; hor) in 11 normal (dashed lines) and 7 UVD subjects (solid lines) at viewing distances of 200, 300, and 600 mm in response to leftward and rightward inter-aural linear head accelerations (Hacc) in normal subjects and ipsilesional and contralesional acceleration in UVD subjects. Horizontal eye velocity in UVD subjects is significantly lower (P < 0.05) than in normal subjects for comparable mean peak linear head acceleration stimulus of 0.51 ± 0.02g (mean ± SD) in normal subjects and 0.53 ± 0.03g in UVD subjects at each of the 3 viewing distances.

The latency of occurrence of the initial saccade from stimulus onset to correct any eye position error was extremely variable (range: 101-418 ms) in both normal and UVD subjects. ANOVA performed on the latency of the occurrence of the initial saccade showed that the variations between the three viewing distances of 200, 300, and 600 mm was not significant either in normal or in UVD subjects. The latency of the occurrence of the initial saccade in normal subjects was 166 ± 41 (SD) ms and in UVD subjects was 177 ± 15 ms.

Modulation of the LVOR by viewing distance

In response to a peak inter-aural linear acceleration of 0.51 ± 0.02g, the horizontal eye velocity of the LVOR, at 100 ms from onset of head movement, in normal subjects increased by 45% when the viewing distance decreased from 600 to 300 mm and almost doubled (95% increase) when the viewing distance decreased from 600 to 200 mm (Figs. 3 and 4; Table 1). For leftward acceleration, modulation of the LVOR became significantly different (P < 0.05) at 50 ms after onset of head acceleration when viewing distance changed from 200 to 300 mm, at 48 ms when viewing distance changed from 200 to 600 mm, and at 57 ms when viewing distance changed from 300 to 600 mm. For rightward acceleration, modulation of the LVOR became significantly different (P < 0.05) at 42 ms after stimulus onset when viewing distance changed from 200 to 300 mm, at 36 ms when viewing distance changed from 200 to 600 mm, and at 41 ms when viewing distance changed from 300 to 600 mm.

View larger version (15K): [in this window] [in a new window]   Fig. 4. Group means ± SE of the horizontal eye velocity of the LVOR at 100 ms after the onset of leftward and rightward linear head acceleration in normal subjects, and ipsilesional and contralesional acceleration in UVD subjects at viewing distances of 200, 300, and 600 mm in 11 normal and 7 UVD subjects. The horizontal eye velocity was significantly enhanced by near viewing in normal subjects. In UVD subjects, horizontal eye velocity was not significantly enhanced between 300 and 200 mm viewing distance, but was significantly enhanced between 600 and 200 mm viewing distance.

In UVD subjects, the mean peak linear acceleration stimulus was 0.53 ± 0.03g, which was comparable to that in normal subjects. The horizontal eye velocity of the LVOR was bilaterally reduced in UVD subjects, but was only significantly modulated by viewing distance (Fig. 4) changes between 600 and 200 mm. In response to ipsilesional linear acceleration, horizontal eye velocity at a viewing distance of 200 mm was significantly greater (P < 0.05) than the response at a viewing distance of 600 mm at 71 ms from stimulus onset, but not significantly different (P > 0.05) from the response at a viewing distance of 300 mm. Both the 200 and 300 mm viewing distance responses were 54% greater than the 600 mm velocity. In response to contralesional linear acceleration, horizontal eye velocity at a viewing distance of 200 mm was not significantly different (P > 0.05) from the response at 300 mm but was significantly greater (P < 0.05) than the response at 600 mm viewing distance at 62 ms from stimulus onset. At 100 ms after onset, the velocity at the 300 mm viewing distance was 70% greater than the 600 mm response, while the 200 mm velocity was 93% greater than that at 600 mm.

Seven of 11 normal subjects and 5 of 7 UVD subjects were also tested in total darkness with the 600 mm viewing distance target extinguished just before the onset of eccentric head rotation, using the same stimulus as those in the presence of a fixation target. One normal subject and one UVD subject demonstrated an inability to perform the task adequately and there was no consistent pattern to their 3-D eye movement responses. At 100 ms from stimulus onset, the mean ± SE of horizontal eye velocity in response to leftward and rightward acceleration in the six remaining normal subjects was 11.9 ± 2.3°/s and 13.0 ± 2.8°/s. These horizontal eye velocities were significantly lower (P < 0.05) from the horizontal eye velocity in the presence of a target at all viewing distances. For the remaining four UVD subjects, the mean ± SE of horizontal eye velocity for ipsilesional and contralesional acceleration in total darkness was 5.8 ± 1.8°/s and 8.8 ± 2.4°/s, respectively. There was no statistically significant difference between the ipsilesional and contralesional responses. These velocity responses at 100 ms from stimulus onset were significantly different from the responses at 200 and 300 mm viewing distances but were not significantly different from the responses at 600 mm viewing distance.

Latency of the LVOR

The group mean and SD of the latencies of the LVOR were computed in the 11 normal subjects and 7 UVD subjects. In normal subjects, the latencies for leftward and rightward acceleration were 34.2 ± 6.4 ms for 200 mm viewing distance, 36.2 ± 5.7 ms for 300 mm viewing distance, and 38.7 ± 6.5 ms for 600 mm viewing distance. In normal subjects, there was no significant difference (P > 0.05) between the latencies from each of the three viewing distances. In UVD subjects, the latencies for ipsilesional and contralesional acceleration were 36.7 ± 7.8 and 41.2 ± 7.6 ms, respectively, for 200 mm viewing distance, 39.4 ± 6.6 and 36.2 ± 9.0 ms, respectively, for 300 mm viewing distance, and 36.6 ± 9.9 and 40.8 ± 8.3 ms, respectively, for 600 mm viewing distance. There was no significant difference (P > 0.05) between the latency for ipsilesional and contralesional linear acceleration in UVD subjects at each of the viewing distances and also no significant difference (P > 0.05) between the latencies of normal and UVD subjects at each of the three viewing distances.

Acceleration gain of the LVOR

The acceleration gain of the LVOR was determined as the ratio of the slope of the horizontal eye velocity compared with the slope of the calculated ideal horizontal eye velocity, for a 30-ms period starting 70 ms from stimulus onset. The mean acceleration gains for 11 normal subjects in response to leftward and rightward accelerations were 0.98 ± 0.26 and 1.01 ± 0.25, respectively, for 200 mm viewing distance, 1.03 ± 0.29 and 1.09 ± 0.32, respectively, for 300 mm viewing distance, and 1.06 ± 0.27 and 1.07 ± 0.32, respectively, for 600 mm viewing distance. There was no significant difference (P > 0.05) between the acceleration gain of the LVOR at each of the three viewing distances for either leftward or rightward linear accelerations in normal subjects.

The mean acceleration gains for seven UVD subjects in response to ipsilesional and contralesional accelerations were 0.42 ± 0.14 and 0.60 ± 0.28, respectively, for 200 mm viewing distance, 0.50 ± 0.23 and 0.63 ± 0.27, respectively, for 300 mm viewing distance, and 0.56 ± 0.29 and 0.66 ± 0.30, respectively, for 600 mm viewing distance. For each viewing distance, the acceleration gain of the LVOR in UVD subjects was significantly lower (P < 0.05) than in normal subjects. Using the Bonferroni adjustment of the significance level of  = 0.025, the ipsilesional acceleration gain was not significantly different (P > 0.025) at 200, 300, or 600 mm viewing distances in UVD subjects.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We have shown here that UVD causes a marked bilateral and almost symmetrical reduction in the LVOR to 40-60% of normal values in response to impulsive inter-aural linear head acceleration. What is unexpected and surprising is not that the LVOR to the deafferented side is diminished, but that the response to the intact side is also reduced significantly. This is in contrast to the AVOR where the response to intact side is well preserved, with only a slight reduction of about 15%, while the response to the deafferented side is severely defective, with a deficit of 60% (Aw et al. 1996a, 2001; Cremer et al. 1998; Halmagyi and Curthoys 1988).

In the otolithic system, a potential for redundancy exists with the morphological arrangement of the hair cells and the central pathways. Every otolith organ is a bi-directionally sensitive structure with opposing polarization of hair cells (Lindeman 1969) and sensitivity vectors (Fernandez and Goldberg 1976) across the striola in the utricle and saccule. Therefore signals for opposing directions of stimulation can be provided from either side of the striola in each otolith organ and from either labyrinth. Cross-striolar inhibition across the utricle and saccule has been shown to enhance the sensitivity of the otolith system, but this effect has been shown to be a less dominant circuit for increasing sensitivity in the utricle than in the saccule (Ogawa et al. 2000). Commissural inhibition, however, may play a more important role in augmenting the sensitivity of the vestibular neurons in the utricular system. A recent study by Uchino et al. (2001) showed that more than one-half of the utricular-activated second-order vestibular neurons and about one-half of the third-order neurons received commissural inhibition from stimulation of the contralateral utricular nerve, while the remaining neurons showed no response to contralateral utricular nerve stimulation. Only a few of these second- and third-order vestibular neurons received facilitation from stimulation of contralateral utricular nerve. As our data show that loss of function from one utricle decreased the output of the LVOR by about 40-60% depending on the viewing distance, we hypothesize that it could be due to the loss of increased sensitivity provided by similarly polarized hair cells on the medial side of the contralateral utricle. This sensitivity increase is probably largely mediated by commissural inhibition of utricular-activated vestibular second- and third-order neurons in vestibular nucleus as commissural facilitation was found to play only a very small role (Uchino et al. 2001).

Previous studies using high-frequency and high-acceleration angular stimuli, such as the head "impulse" or head "thrust" tests, have demonstrated that both labyrinths are required to produce a normal AVOR. After UVD, the response is profoundly reduced to the lesioned side but remains close to normal to the intact side (Aw et al. 1996a, 2001; Cremer et al. 1998; Halmagyi and Curthoys 1988; Tabak et al. 1997). Our present findings suggest that for the LVOR, the utricles in the two labyrinths both contribute to the response during inter-aural linear head acceleration. Unilateral loss of utricular function effectively halves the input, and hence the output from the system is also nearly halved (approximately 40-60%), with the deficit being more pronounced at closer viewing distances. Our present findings also show that compensation of the LVOR in humans is incomplete and there is a permanent bilateral deficit after UVD in response to a high acceleration of approximately 0.5g linear stimulus.

Although the mean response in this study was slightly lower during stimulation toward the ipsilesional side than toward the contralesional side, we were not able to show any significant difference between the two responses in our UVD subjects tested more than 3 mo after UVD (range: 3 mo to 14 yr). In the AVOR, it is well established that compensation is incomplete after UVD, with both an inadequate eye velocity response as well as an incorrect axis of eye rotation (Aw et al. 1996a, 2001; Cremer et al. 1998; Halmagyi and Curthoys 1988; Tabak et al. 1997), even though there is recovery from the acute vestibular shutdown and restoration of resting discharge after vestibular injury (Curthoys and Halmagyi 1995). Similarly, Angelaki et al. (2000b) reported that in primates, the LVOR responses in the acute stage, i.e., 1 wk after unilateral labyrinthectomy, were bilaterally and asymmetrically reduced, with a larger deficit to the ipsilesional than to the contralesional side in response to transient lateral translational motion. This response asymmetry may be attributed to the absence of a resting discharge during the acute stage (Curthoys and Halmagyi 1995). After 3 mo, although the LVOR was still reduced bilaterally, the response asymmetry was greatly diminished (Angelaki et al. 2000b), which could be attributed to the restoration of the spontaneous central neuronal discharge (Ris et al. 1995).

Others have also shown deficits in the human LVOR after UVD. Ramat et al. (2001) found a reduced LVOR in response to a transient lateral translational stimulus in two human subjects after intra-tympanic gentamicin. The responses were symmetrically reduced in one, but asymmetrically reduced in the other, immediately after intra-tympanic gentamicin. The lesions in these two patients were probably incomplete, because the extent of the ablation of labyrinthine function by intra-tympanic gentamicin is not independently quantifiable. Therefore it is difficult to surmise on the difference in their responses. Using an eccentric whole-body yaw rotation stimulus, Crane and Demer (1998) reported a reduction in combined ipsilesional linear and AVOR after UVD. As the linear and AVOR responses were coaxial, the study was unable to determine if the LVOR was deficient after UVD. However, by varying viewing distances and head eccentricities from the rotation axis, Crane and Demer (1998) identified a component attributable to the LVOR that was reduced after UVD. Lempert (1998) showed a diminished ipsilesional response to lateral whole-body translation 1 wk after UVD. These responses regained symmetry and completely recovered 6-10 wk later. A factor that may account for this "recovery" is that the subjects' ocular responses in the study were analyzed between 300 and 500 ms after stimulus onset. At 300 ms after onset of head motion, subjects could employ other strategies such as smooth pursuit to minimize retinal slip (Carl and Gellman 1987). Smooth pursuit strategies could observed in our study at times later than 100 ms after the onset of head motion in both normal and UVD subject (Fig. 2). For example, in Fig. 2, the normal subject was able to generate the initial saccade just after 100 ms after the onset of head acceleration. We observed that both normal and UVD subjects showed both patterns of smooth pursuit strategy, by either producing an early saccade or by increasing slow phase velocity.

We showed that the mean latency of the LVOR is about three times longer than the AVOR. It was found to be about 36 ms in normal subjects and about 39 ms in UVD subjects, but the difference in the latency of the LVOR between normal and UVD subjects was not statistically significant. Aging may be one of the factors attributable to the prolongation of latency (Tian et al. 2002), because our UVD subjects are slightly older than our normal subjects. The latency that we measured is comparable to the range of >30 ms previously reported in humans (Bronstein and Gresty 1988; Crane and Demer 1998; Wiest et al. 2001). Wiest et al. (2001) reported the latency is also unchanged in subjects with cerebellar ataxia even though their LVOR responses were reduced. In primates, Synder and King (1992) showed a comparable latency of about 30 ms, which was longer than the 12 ms reported by Angelaki and McHenry (1999).

Measurement of VOR latency in humans is generally associated with problems such as slippage of measuring devices, e.g., eye coil and linear head accelerometer, the sampling frequency, the resolution of the data acquisition system, and the criteria employed in the data analysis. All of these factors can vastly influence the results. We tried to minimize any artifacts during our latency measurement by minimizing slippage of the linear head accelerometer by coupling it to the subject's head with a dental impression bite-bar; restraining the subjects' heads with individualized thermoplastic masks, using a high sampling rate of 1 kHz; using high resolution 16-bit sampling of head and eye signals; and using an automated algorithm for detecting the onsets of head and eye movements with criteria that minimize artifacts to calculate latency. Our automated algorithm is less susceptible to artificial prolongation of the LVOR latency because we determined the onset as the point at which the signal first exceeded 1 instead of 3 SD (Angelaki and McHenry 1999; Crane and Demer 1998; Wiest et al. 2001). To increase the sensitivity of detection, we also used the derivatives of the inter-aural linear head acceleration signal and the horizontal eye velocity signal for detection of onset, instead of using eye position (Crane and Demer 1998; Wiest et al. 2001) or eye velocity directly(Angelaki and McHenry 1999).

Our data also showed that the normal LVOR response is enhanced by near viewing, as has been shown previously in humans (Gianna et al. 2000; Paige et al. 1998) and in primates (Angelaki et al. 2000a; Schwarz et al. 1989; Telford et al. 1997). Modulation by viewing distance results in significantly different horizontal eye velocities between all three viewing distances in normal subjects. Although the response sensitivity was bilaterally reduced in UVD subjects, the modulation of the LVOR by near viewing was still significant between the responses at 200 and 600 mm viewing distances. This was consistent for linear acceleration toward both the intact and lesioned side and was consistent with primate data (Angelaki et al. 2000b). We did not find a significant difference between the responses for targets at 200 and 300 mm viewing distance and attributed this absence to the lower response sensitivity after UVD. A limitation of this study was that we did not have the facility for binocular search coil recordings so we were unable to determine if the subjects, especially the older subjects, were verging on the target. Since the modulation of the LVOR response remains even after UVD, it is suggestive that the central inputs from a single utricle are responsible for this modulation, and commissural effects probably do not play a role in it.

In conclusion, we have shown that UVD causes a bilateral and largely symmetrical reduction in the LVOR to 40-60% of the normal response; however, its properties, such as latency and dependence on viewing distance, remain intact. In contrast, a complete loss of unilateral labyrinthine function results in a strongly asymmetrical AVOR output with a profound loss of output from the lesioned side and near normal output from the intact side.