Spatial and temporal properties of eye movements produced by electrical stimulation of semicircular canal afferents

Richard F. Lewis, Csilla Haburcakova, Wangsong Gong, Faisal Karmali, Daniel M. Merfeld


To investigate the characteristics of eye movements produced by electrical stimulation of semicircular canal afferents, we studied the spatial and temporal features of eye movements elicited by short-term lateral canal stimulation in two squirrel monkeys with plugged lateral canals, with the head upright or statically tilted in the roll plane. The electrically induced vestibuloocular reflex (eVOR) evoked with the head upright decayed more quickly than the stimulation signal provided by the electrode, demonstrating an absence of the classic velocity storage effect that improves the dynamics of the low-frequency VOR. When stimulation was provided with the head tilted in roll, however, the eVOR decayed more rapidly than when the head was upright, and a cross-coupled vertical response developed that shifted the eye's rotational axis toward alignment with gravity. These results demonstrate that rotational information provided by electrical stimulation of canal afferents interacts with otolith inputs (or other graviceptive cues) in a qualitatively normal manner, a process that is thought to be mediated by the velocity storage network. The observed interaction between the eVOR and graviceptive cues is of critical importance for the development of a functionally useful vestibular prosthesis. Furthermore, the presence of gravity-dependent effects (dumping, spatial orientation) despite an absence of low-frequency augmentation of the eVOR has not been previously described in any experimental preparation.

  • vestibular nerve
  • velocity storage
  • spatial orientation

velocity storage was first described three decades ago (Cohen et al. 1977; Robinson 1977; Waespe and Henn 1977) as a neural circuit in the brain that prolongs rotational information provided by vestibular and visual cues, thereby improving the performance of the vestibuloocular reflex (VOR) during low-frequency head rotations. When the upright head is rotated in yaw at a constant velocity about an Earth-vertical axis, for example, the VOR decays more gradually than the signal encoding head velocity in the vestibular nerve. These temporal characteristics are strongly modulated by the orientation of the rotational axis relative to gravity. If the head is tilted in the roll plane after rapidly decelerating from a constant-velocity yaw rotation, for example, the duration of the horizontal postrotatory nystagmus is attenuated (“dumped”) compared with the condition in which the head remains upright (Angelaki and Hess 1994; Benson and Bodin 1966; Merfeld et al. 1993b). Eye movements produced by vestibular and visual rotational cues also show spatial orientation properties whereby the eye's rotational axis shifts toward alignment with gravity (Cohen et al. 1999). Postrotational tilt in the roll plane, for example, results in the nystagmus shifting direction from the initial horizontal, head-centered axis toward the Earth-vertical axis defined by gravity (Angelaki and Hess 1994; Merfeld et al. 1993b).

It is widely accepted that the same velocity storage mechanism that improves the low-frequency VOR also underlies the interaction between canal and otolith cues that is responsible for tilt suppression and spatial orientation of vestibular and optokinetic eye movements (reviewed in Cohen et al. 1999). These temporal and spatial properties have never been dissociated experimentally. In particular, lesions in the brain that eliminate (Katz et al. 1991; Wearne et al. 1997) or disinhibit (Angelaki and Hess 1995; Wearne et al. 1998) the effects of velocity storage on the low-frequency VOR also block tilt suppression and the development of the cross-coupled eye movements that serve to shift the rotational axis toward alignment with gravity. As part of our effort to develop a vestibular prosthesis, we studied the temporal characteristics of eye movements produced by electrical stimulation of lateral canal afferents in squirrel monkeys with plugged lateral canals and found that the resultant VOR time constants were shorter, rather than longer, than the input provided by the periphery (Lewis et al. 2010; Merfeld et al. 2007). These results suggested that velocity storage may not be activated by electrical stimulation of canal afferents. This is a critical issue for the potential utility of a vestibular prosthesis, since an interaction between the artificial rotational cue provided by the prosthesis and otolith inputs is necessary if a canal prosthesis is to improve the brain's estimate of head orientation relative to gravity, a process that underlies the brain's ability to correctly interpret otolith signals as translational or gravitational in origin (Angelaki et al. 1999; Merfeld et al. 1999).

To investigate this issue further, in the present study we measured the spatial and temporal characteristics of eye movements in squirrel monkeys that were produced by electrical stimulation of lateral canal ampullary nerves with the head upright or tilted. The goal was to determine whether tilt suppression and spatial orientation of the electrically induced VOR (eVOR) occurred despite the brain's inability to augment the low-frequency eVOR response. We found that both tilt suppression and spatial orientation of the eVOR were present, although they differed in subtle ways from normal responses. These results demonstrate that, despite the short VOR time constant elicited by electrical stimulation, the rotational cue produced by this stimulation does activate the velocity storage mechanisms responsible for both the temporal and spatial interactions between rotational and gravitational cues. The observation that prosthetic rotational cues combine with graviceptive cues in a qualitatively normal manner, alongside findings that the gain and axis of the eVOR adapt during chronic stimulation (Dai et al. 2011a; Lewis et al. 2010), provide strong evidence that rotational information provided by electrical stimulation of canal afferents is used by the brain in a manner that mimics normal sensory integration.


Many of the methods used in this study have been described in other publications from our laboratory (e.g., Lewis et al. 2010; Merfeld et al. 2007), so they are summarized briefly here. The novel elements of this study are described in detail. These experiments were performed in two adult squirrel monkeys, and all aspects of the study were approved by the institutional animal care and use committee and were in accordance with US Department of Agriculture guidelines.


Surgeries were performed under general anesthesia provided by either intravenous pentobarbital or inhaled isoflurane and included implanting a head bolt into the skull, insertion of a frontal eye coil in one eye, plugging of both lateral canals (Lewis et al. 2010; Rabbitt et al. 1999), and insertion of a stimulating electrode into each lateral canal through a second opening more proximal to the ampulla than the canal plug. As previously described (Lewis et al. 2010; Merfeld et al. 2007), the electrodes were made from platinum wire with ∼150 μm of Teflon insulation stripped from the tip. The electrodes were slowly advanced toward the ampulla while electrical pulse trains were applied and eye movements were observed. Each electrode was fixed in the position that elicited maximum horizontal eye movements with minimal vertical eye movement or facial nerve activation.

Electrical Stimulation

Charge-balanced biphasic current pulses (200 μs cathodic, 200 μs rest, 200 μs anodic) were used to stimulate the lateral canal ampullary nerve, and the amplitude of the pulses was chosen by previously described criteria (Lewis et al. 2010). In each monkey, the electrode that elicited larger horizontal eye movements was used. In monkey R, the right electrode was used with pulse amplitudes of 130 μA; for monkey S the left electrode was used with pulse amplitudes of 140 μA. Both monkeys were studied during chronic prosthetic stimulation, as previously described (Lewis et al. 2010; Merfeld et al. 2007). Briefly, head velocity in the plane of the lateral canals was measured by a piezoelectric angular rate sensor in the prosthesis and high-pass filtered with a time constant of 5 s (to simulate the dynamics of the normal canals), and this filtered angular head velocity was used to modulate the rate of current pulses applied unilaterally by the implanted electrode. The motion-modulated component (based on a hyperbolic tangent function with a slope of 1.8 pulses/s/°/s in monkey R and 1.2 pulses/s/°/s in monkey S over the linear range) was superimposed on a tonic baseline stimulation rate (e.g., the rate of stimulation with the head stationary) of 200 pulses per second (pps) in monkey R and 250 pps in monkey S. A nonzero tonic baseline rate was chosen since it allowed us to encode bidirectional head rotations with unilateral stimulation (e.g., by increasing the stimulation rate for head turns ipsilateral to the ear with prosthetic input and decreasing the stimulation rate for contralateral head turns).

Experimental Protocol and Data Analysis

For all experiments, the head was pitched forward by 17°, thereby placing the lateral canals nearly parallel to the Earth-horizontal and the vertical canals close to the Earth-vertical (Blanks et al. 1985). Horizontal and vertical eye movements were recorded from one eye with standard search coil methods. Eye movements were filtered at 50 Hz, sampled at 200 Hz, and analyzed off-line with an interactive program that allowed accurate removal of quick phases. Eye movement direction was defined with the right-hand rule sign convention, with positive values being leftward and downward. Eye coils were linearly calibrated based on a fixed relationship between angular deviation from zero and the recorded voltage (20° defined to equal 5 V).

The prosthesis was initially activated with the monkey upright, stationary, and in the dark. After waiting 30 min for the spontaneous nystagmus to attenuate (Lewis et al. 2010; Merfeld et al. 2006), each monkey was tested with angular velocity steps to the right and left (accelerated from 0 to 80°/s over 1 s, then rotated at a constant velocity for 90 s), with the time constant of the prosthesis angular velocity high-pass filter set to 5 s. The monkeys then entered the period of chronic prosthetic stimulation, during which the prosthesis continuously provided pulsatile electrical stimulation to the lateral canal ampullary nerve at a rate that was modulated by the animal's head velocity while it moved freely in its normally lit cage (see Lewis et al. 2010 for further details). Monkey R was first tested on the task that is the focus of the present study on day 16 of chronic stimulation and was tested six times over a period of 66 days. Monkey S was first tested on day 377 and completed six tests over the ensuing 13 days. Despite the different durations of chronic stimulation, the results in the two monkeys were comparable. Monkey S did have weaker eye movement responses to electrical stimulation, but this was not due to the longer duration of chronic stimulation in this animal, as this difference was observed during the initial test after stimulation was activated.

The monkeys were tested with the head stationary and in the dark. The ear receiving chronic prosthetic stimulation was given an additional step of stimulation superimposed on the tonic, baseline rate. For monkey R the step was +167 pps, raising the stimulation rate in the right ear from 200 to 367 pps. For monkey S, the step was −224 pps, lowering the stimulation provided to the left ear from 250 to 26 pps. These steps of stimulation decayed gradually with a time constant of 80 s and had durations of 60 s.

A time constant of 80 s was chosen because we previously observed that the horizontal eye movements produced with a step of electrical stimulation that decayed with the normal cupular time constant of 5 s attenuated very rapidly and essentially resolved within 10 s (Haburcakova et al. 2005; Lewis et al. 2010). Postrotational tilt experiments in squirrel monkeys, however, have shown that the cross-coupled eye movement component does not begin to develop for ∼10 s and peaks at ∼20 s (Merfeld et al. 1993b). We therefore chose to use an 80-s stimulus time constant because it would allow us to extend the duration of the eye movement response produced by the step of electrical stimulation so that the vertical cross-coupled component could be observed if it were to develop. Furthermore, previous experiments using the vestibular prosthesis with an 80-s stimulus time constant produced VOR responses with time constants in the normal range (Haburcakova et al. 2005), so the 80-s time constant allowed us to produce VOR responses with normal low-frequency dynamics. As discussed in detail below, this approach offers a compromise solution to the question we wished to study—by extending the afferent cue provided by electrical stimulation (with an 80-s time constant), we should be able to observe gravity-dependent effects on the eVOR if they occur, but in tandem we are providing an afferent signal that differs from the normal postrotational canal cue (which decays much faster with a time constant of 5 s).

The electrical stimuli were provided with the head in one of two orientations, either upright or tilted in roll toward the right-ear-down orientation (30° tilt for monkey R, 45° tilt for monkey S). Eye movement responses were analyzed for their temporal characteristics by calculating the rate at which the slow-phase eye velocity decayed and for their spatial characteristics by plotting horizontal versus vertical eye velocity and calculating the peak shift in the eye's rotational axis. Since the eye velocity traces were poorly fit by single exponentials or combinations of exponentials, the “time constant” of the decay was defined empirically as the time required for the eye velocity to decay to a value of 1/e times the maximal value. The total eye velocity in the frontal plane was calculated as [(horizontal velocity)2 + (vertical velocity)2]1/2.



Prior to ear surgery, the VOR produced by a step of angular head velocity decayed gradually in both monkeys (Fig. 1, top) and therefore exhibited the classic feature of velocity storage whereby the eye movement response is prolonged beyond that of the peripheral input (Fig. 1, bottom). In contrast, after canal plugging and electrode implantation, when the prosthesis was first activated with a time constant of 5 s and the monkeys were subjected to the same velocity steps (Fig. 1, bottom), the eVOR decayed more rapidly than the input stimulus in both animals (Fig. 1, middle), with a time constant of 3.5 s in monkey R and 3.2 s in monkey S. Similar to our previous reports (Lewis et al. 2010; Merfeld et al. 2007), the eVOR time constant remained relatively stable over time during long-term prosthetic stimulation.

Fig. 1.

Normal and prosthesis-mediated vestibuloocular reflexes (VORs) measured in the dark and with the head upright. Head motion stimuli (bottom, solid lines) were velocity steps from 0 to 80°/s. Top: the VOR response in the normal monkeys. Middle: the prosthesis-driven VOR response, which was measured after 30 min of tonic electrical stimulation of 1 lateral canal ampullary nerve with the animal stationary. Although the motion-modulated prosthesis stimulation was filtered with a time constant of 5 s to approximate normal cupular dynamics (bottom, dashed lines), the prosthesis-mediated VOR decayed more rapidly than the input signal in both monkeys while the normal VOR decayed more slowly than the input signal. All eye movement traces in this and subsequent figures follow the right-hand rule, with positive values representing leftward or downward eye movements.

Figure 2 shows sample eVOR responses in both monkeys when the lateral canal was activated with a step of electrical stimulation that decayed with a time constant of 80 s. The monkeys were stationary during these tests and were positioned either upright or tilted 30° (monkey R) or 45° (monkey S) in roll toward the right-ear-down orientation. These raw data traces demonstrate that when the monkeys were upright (Fig. 2, top), a large horizontal (leftward) eye movement response was elicited by the step of stimulation but little vertical response occurred. Monkey R did have a very small upward eye response at the onset of stimulation, while monkey S had a weak downbeat nystagmus with upward slow phases of ∼3°/s. Since the spontaneous downbeat nystagmus in monkey S was constant in the dark and was independent of the electrical stimulation, it was subtracted from the vertical velocity traces in all subsequent analysis. When the monkeys were stimulated while tilted in roll (Fig. 2, bottom), a substantial vertical eye movement response with positive (downward) slow phases occurred, which shifted the eye's rotational axis toward alignment with gravity. The horizontal responses were not obviously affected by the head tilt, although close inspection of the raw data suggests that an early plateau in horizontal eye velocity present in the upright state during the first 15–30 s of stimulation was attenuated in the tilted position.

Fig. 2.

Electrically induced VOR (eVOR) responses produced by electrical stimulation of 1 lateral canal ampullary nerve with the monkeys stationary and in the dark. The animals were either upright or tilted in the roll plane toward the right-ear-down position by 30° (monkey R) or 45° (monkey S). The electrical stimulation consisted of a step increase (200 → 367 Hz, monkey R) or decrease (250 → 26 Hz, monkey S) from the tonic level of stimulation provided by the prosthesis, which decayed with a time constant of 80 s. H, horizontal; V, vertical.

Below we consider in detail the spatial and temporal characteristics of the eVOR elicited by steps of electrical stimulation with the animals upright or tilted in roll. The eVOR responses in these two monkeys are compared with normative squirrel monkey data obtained from a study of postrotational tilt previously performed by one of the authors (Merfeld et al. 1993b).

Spatial Characteristics of the eVOR

Representative eVOR responses are plotted in polar coordinates for monkeys R and S in Fig. 3. In these examples, it is evident that the eye's rotational axis closely aligned with gravity when the monkeys were upright (peak axis shift of −4.6° in monkey R and 3.8° in monkey S, where positive values represent a clockwise shift in the rotational axis). Similarly, in a normal monkey the rotational axis of the nystagmus produced by rapid deceleration from a constant velocity was closely aligned with gravity when the head remained upright (Fig. 3; peak axis shift of 4.4° in this example). When the head was tilted in the roll plane prior to electrical stimulation, the eVOR demonstrated a large axis shift toward alignment with gravity, with the axis shift generally exceeding the amplitude of the head tilt. For example, the responses illustrated in Fig. 3 demonstrate a peak axis shift of 35° in monkey R for a 30° head tilt and a 53.7° axis shift in monkey S for a 45° head tilt. The mean eVOR axis shifts in the upright and tilted orientations for all trials are summarized in Table 1. On average, the peak axis shift for monkey R exceeded the 30° amplitude of the head tilt by 13%, and for monkey S the peak axis shift was 34% larger than the 45° head tilt. In contrast, postrotational tilt in normal squirrel monkeys elicits an axis shift that closely approximates the amplitude of the head tilt (e.g., 30.5° for 30° head tilt, 44.9° for 45° head tilt; Merfeld et al. 1993b), although for the example in Fig. 3 the axis shift was atypically large and exceeded the head tilt angle.

Fig. 3.

Spatial characteristics of representative eVOR responses (monkeys R and S) and the postrotational tilt responses in a normal monkey. Quick phases have been removed from all data, and eye velocity is plotted in head-centered coordinates, with horizontal velocity on the y-axis and vertical velocity on the x-axis. The orientation of the Earth-vertical relative to the head is indicated by the arrows, and the peak axis shift is indicated by the lines. Head tilt angles were 30° (monkey R), 45° (monkey S), and 32° (normal monkey).

View this table:
Table 1.

Rotational axis of the eVOR

As detailed below, the dynamics of the horizontal and vertical eVOR components during head tilts differed from normal postrotational tilt responses. The vertical component of the eVOR response decayed more rapidly than the horizontal component in both monkeys (more dramatically in monkey R), such that the vertical response was extinguished before the horizontal component. The eye velocity vector therefore shifted away from gravity and back toward the head-vertical orientation during the latter portions of the eVOR response (Fig. 3, monkeys R and S). In contrast, in both normal squirrel monkeys (Fig. 3) and rhesus monkeys (Angelaki and Hess 1994), the eye velocity vector decays toward the origin along the angle of the peak axis shift, which approximates the orientation of gravity.

Temporal Characteristics of the eVOR

Figure 4 illustrates the slow-phase eye movement responses for representative electrical stimulation trials (monkeys R and S) and for a normal monkey. The top and second rows in Fig. 4 show the horizontal and vertical eye movement responses, respectively, when upright or tilted and the [upright − tilted] difference. Like the eVOR responses produced by rotation at the onset of stimulation (Fig. 1), the time constant of the horizontal eVOR with the head upright was always substantially shorter than the time constant of the peripheral input. In particular, the mean horizontal eVOR time constant during upright stimulation was about 24% (monkey R) and 33% (monkey S) of the 80-s input time constant provided by the electrical stimulation (Table 2).

Fig. 4.

Slow-phase eye movement responses produced by electrical stimulation (left and center) and a standard postrotational tilt paradigm in a normal monkey (right). Each plot includes responses with the head upright (solid lines) and tilted (dashed lines) and the [upright − tilted] difference (dotted lines). Total eye velocity in the frontal plane is calculated with the approach described in methods. Bottom: the electrical stimulation provided by the prosthesis in monkeys R and S (step change followed by a decay to baseline with an 80-s time constant) and the approximate change in lateral canal afferent firing in a normal squirrel monkey subjected to a 160°/s velocity step (estimated from Goldberg and Fernandez 1971).

View this table:
Table 2.

Time constants for the eVOR

For both monkeys R and S, it is evident that the horizontal component (Fig. 4, top row) decayed initially at similar rates in the upright and tilted orientation, but between ∼5 and 30 s the traces diverged, with the horizontal response in the tilted orientation decaying more rapidly. This is most clearly observed in the [upright − tilted] traces, which show a positive deflection peaking at ∼5°/s and returning to zero after ∼30–35 s. After ∼30 s the horizontal responses were nearly the same in the upright and tilted orientations and persisted throughout the 60-s period of stimulation (reflecting the 80-s time constant of the input stimulus). The average change in the horizontal time constant produced by tilting the head was a reduction of 19.5% in monkey R and 57.9% in monkey S (Table 2). The vertical eye movements (Fig. 4, second row) displayed minimal responses in the upright orientation but a substantial positive response in the tilted position, peaking at about 12°/s in both monkeys (Fig. 4) and resolving after about 20 s in monkey R and 30 s in monkey S (Table 2).

Unlike the eVOR, the horizontal response in the normal monkey after postrotational tilting (Fig. 4, top row) immediately deviated from the upright response, attenuated more rapidly, and resolved after ∼30 s (reflecting both the shortened velocity storage time constant and the high-pass characteristics of the normal canal input). The reduction in the horizontal time constant produced by postrotational tilts in the normal monkey was 38% for 30° head tilt (e.g., greater than the eVOR reduction) but was 54% for 45° head tilt (similar to the eVOR reduction). The vertical response after postrotational tilt in the normal monkey also differed from eVOR responses (Fig. 4, second row), as it peaked and resolved later than the vertical eVOR response.

To determine whether the eye velocity was attenuated by head tilts and not simply reoriented in the frontal plane, we calculated the total eye velocity in the upright and tilted orientations (Fig. 4, third row; Table 2). The total velocity in the upright position was nearly identical to the horizontal velocity, since minimal vertical responses occurred. In the tilted position, the time constant of the total response was shortened (Table 2) and the total eye velocity trace displayed an early, transient peak that resulted from the vertical eye movement evoked in this orientation. The [upright − tilted] eVOR trace for total eye velocity was biphasic, with an early negative deflection followed by a more prolonged positive deflection.

The afferent stimulation that elicited these eye movement responses is illustrated in the bottom row of Fig. 4 for monkeys R and S, and the lateral canal afferent firing rate is estimated for the normal squirrel monkey (Goldberg and Fernandez 1971) based on a 160°/s velocity step (Merfeld et al. 1993b). It is evident from these plots that the afferent cues differed substantially between the two monkeys that received electrical stimulation of the lateral canal afferents and the normal monkey that was tilted after being rapidly decelerated. In particular, given its 80-s decay time constant, the electrical stimulation rate changed little in monkeys R and S during the time frame of the horizontal VOR tilt suppression and the development of the cross-coupled vertical VOR. In contrast, the postdeceleration modulation in afferent firing in the normal monkey decayed much more quickly (time constant of 5 s) and therefore had largely resolved before the primary tilt suppression and cross-coupled effects developed.

The mean and SE for the [upright − tilted] differences for the horizontal, vertical, and total eye velocity traces for all eVOR trials are illustrated in Fig. 5 and summarized in Table 3. These responses were quite consistent over multiple stimulation trials. The horizontal [upright − tilted] difference (Fig. 5, left), which can be considered the component of the response that was “dumped” when the head was tilted away from the Earth-vertical, developed relatively rapidly but decayed more slowly, particularly in monkey S. In both animals the vertical components (Fig. 5, center) developed and decayed more rapidly than the “dumped” horizontal components (Table 3), resulting in biphasic total velocity traces with a small negative deflection followed by a more prolonged positive deflection (Fig. 5, right).

Fig. 5.

Slow-phase eVOR responses, [upright − tilted] difference, for both stimulated monkeys. Traces represent the mean (dark lines) ± SE and are calculated from 6 upright and tilted trials for each monkey.

View this table:
Table 3.

Time constants for the “velocity storage” component of the eVOR


The primary findings of this study are that 1) rotational cues produced by electrical stimulation of semicircular canal ampullary nerves interacted with graviceptor cues, evidenced by tilt suppression and spatial orientation of the eVOR responses that were qualitatively normal, and 2) the horizontal eVOR evoked with the head upright decayed more rapidly than the input signal from the periphery and hence did not show evidence of the classical velocity storage effect. Below we discuss the implications of our findings regarding the potential utility of a vestibular prosthesis, consider possible explanations for the dissociation we observed between the gravity-dependent and low-frequency temporal features of the eVOR, and discuss potential future directions for this work.

Implications for the Potential Utility of a Vestibular Prosthesis

Earlier studies have shown that chronic electrical stimulation of semicircular canal afferents provided by a canal prosthesis can generate compensatory angular eVOR responses that reduce motion of images on the retina during head rotation (Dai et al. 2011b; Lewis et al. 2010; Merfeld et al. 2007). Furthermore, visual error signals can clearly drive adaptation of the oculomotor responses produced by the prosthetic rotational cue, since both the gain (Lewis et al. 2010; Merfeld et al. 2007) and axis (Dai et al. 2011a; Lewis et al. 2010) of the eVOR change over time in a manner that serves to reduce retinal image motion. No prior studies, however, have shown any evidence that electrical stimulation of canal afferents engages central velocity storage mechanisms. This is a significant issue since velocity storage is believed to underlie the influence of otolith cues on rotational responses (see, e.g., Merfeld et al. 1993a; Wearne et al. 1997) and may also mediate the role of rotational cues in interpreting otolith signals (Green and Angelaki 2003; Merfeld et al. 1993a). This latter possibility is of great functional significance since the brain's ability to distinguish head tilts relative to gravity from linear translation appears to rely on this interaction (Angelaki et al. 1999; Merfeld et al. 1999). More specifically, the otoliths respond identically to shifts in gravitational force (caused by head tilt) and to inertial force (caused by linear acceleration), and there is substantial evidence that rotational information derived from the canals is used by the brain to differentiate between tilt and translation (Angelaki et al. 1999; Merfeld et al. 1999).

Since a canal prosthesis can in principle supply the brain with an accurate representation of head angular velocity in three dimensions, canal inputs should be adequate to reduce oscillopsia associated with head rotation in vestibulopathic subjects by generating a high-frequency compensatory angular VOR that operates in head-centered coordinates. Conversely, more complex behaviors that are mediated in part by vestibular cues such as postural control and perception of head orientation require information derived from interactions between canal-mediated rotational cues and graviceptive signals (Merfeld et al. 1999, 2005). The sensory integration between the prosthesis-generated rotational cues and otolith inputs we documented in this study is therefore essential for a canal prosthesis to improve central estimates of head orientation relative to gravity in vestibulopathic subjects. This process contributes directly to the perception of head orientation and indirectly to the perception of head translation (most likely via an internal model; Angelaki et al. 1999; Merfeld et al. 1999). Furthermore, postural control also depends on an accurate estimate of head orientation in space, since recent work has demonstrated that body orientation relative to gravity, which is derived by combining vestibular (head orientation relative to gravity) and neck proprioceptive (head orientation relative to the body) signals, is an important parameter regulated by the postural control system (Peterka 2002; Stapley et al. 2006). In summary, in prior eye movement studies using the canal prosthesis, the apparent absence of velocity storage raised the possibility that the prosthetic rotational cue may not be synthesized appropriately with graviceptive inputs. The present study clearly shows that rotational cues generated by electrical stimulation of canal afferents do interact with otolith inputs in a manner that is qualitatively normal. This is an important finding because it suggests that a canal prosthesis could potentially improve perception of head motion and orientation as well as postural control in vestibulopathic subjects.

Potential Mechanisms Underlying Differences between the eVOR and Normal VOR

The principal difference between the normal VOR and the eVOR was the duration of the horizontal response when the head was upright. The former showed normal velocity storage characteristics (e.g., the time constant of the VOR was greater than the peripheral input), while the latter showed no evidence of low-frequency augmentation (e.g., the eVOR time constant was shorter than the peripheral input). Despite this difference, qualitatively normal tilt suppression and spatial orientation of the eVOR were observed. This dissociation between low-frequency augmentation and gravity dependence of the VOR has not been previously reported either in normal monkeys or after lesions in the medulla that inactivate velocity storage (Katz et al. 1991; Wearne et al. 1997), since both the storage and orientation-dependent properties of velocity storage were lost after these lesions. Furthermore, lesions in the cerebellar nodulus and uvula (Angelaki and Hess 1995; Wearne et al. 1998) increase the VOR time constant with the head upright and concurrently block tilt suppression and spatial orientation of eye movements. The closest another study has come to demonstrating a dissociation between low-frequency dynamics and gravity-dependent VOR properties occurred after medial lesions of the nodulus and uvula (Wearne et al. 1998), which blocked the cross-coupled responses that shift the eye's rotational axis but did not affect the dynamics or tilt suppression of the eye movement response in head-centered coordinates.

Since electrical stimulation of canal afferents produced qualitatively normal tilt suppression and spatial orientation of the eVOR without augmenting the low-frequency eVOR dynamics, it is clear that the elements of the velocity storage network that are responsible for these behaviors are not identical, as they can be separated in certain conditions. Three elements of our experiment design could potentially be responsible for these findings.

Plugging of the lateral canals.

This procedure was used to render the canals unresponsive to angular head velocity over the frequency range we studied during the chronic prosthesis experiment (Sadeghi et al. 2009). Prior work has shown that canal plugging alters the dynamics of the low-frequency VOR in a manner suggesting that canal rotational cues no longer engage the velocity storage integrator or, alternatively, that velocity storage is not capable of compensating adequately for the dynamic changes in the VOR produced by the plugging procedure (Angelaki et al. 1996; Paige 1983). It remains plausible, however, that despite its effects on VOR dynamics, canal plugging could leave the interaction between canal and otolith signals intact, and this could explain the persistence of the gravity-dependent eVOR features in our experiment. It is important to note that this hypothesis could not be tested in otherwise normal canal-plugged animals, since the gravity-dependent features of the VOR are low-frequency phenomena but the VOR response is so markedly attenuated in the low-frequency range after canal plugging (Angelaki et al. 1996) that no meaningful observations about tilt suppression or spatial orientation are possible.

Dynamics of the canal cue produced by electrical stimulation.

As noted above, for these experiments we chose as our stimuli a step change in the rate of electrical stimulation that decayed to baseline slowly with a time constant of 80 s. While this allowed us to observe the gravity-dependent features that might have not been evident if the stimulation decayed faster, it is important to emphasize that our stimuli differed substantially from the canal afferent cue in the normal monkey, which decayed much more rapidly (this difference is illustrated in the bottom row of Fig. 4). We cannot be certain that similar gravity-dependent features of the eVOR would be observable if the electrical stimulus decayed with a physiological time constant of 5 s, since we did not perform this experiment. It seems more probable, however, that the 80-s time constant simply allowed us to observe these effects by extending the duration of the eye movement response and was not itself responsible for the interaction between the electrically stimulated canal cue and graviceptive signals. In contrast, many of the quantitative differences observed between the dumping and spatial orientation responses of the eVOR and the VOR in the normal monkey are likely due to the substantial difference in the dynamics of the canal afferent cue.

Other nonphysiological characteristics of vestibular nerve activation.

The afferent fibers that were stimulated by the prosthesis most likely discharged simultaneously and in synch with the applied current pulses (Litvak et al. 2003), which could have the effect of introducing a nonphysiological correlation in the noise carried by the group of modulated fibers. It has been suggested that the velocity storage integrator is leaky to discharge the bias generated by integrating noisy input signals (Laurens and Angelaki 2011). If the input noise were correlated rather than random, its integral might accumulate more rapidly and therefore might need to be discharged more rapidly to minimize bias in the system. This line of reasoning suggests that the eVOR time constant should be shorter than that of the normal VOR, and indeed a recently implemented particle filter model of the VOR predicted that correlating the noise in afferent vestibular fibers should reduce the time constant of the VOR (Karmali and Merfeld 2012). Similarly, correlated noise could be considered equivalent to increasing the amount of noise, and simulations have suggested that the VOR time constant could become shorter if the amount of noise carried by canal afferents increased (MacNeilage et al. 2008).

A second potential explanation is that high-frequency electrical stimulation could reduce the sensitivity of peripheral or central vestibular neurons (Tykocinski et al. 1995) to changes in the stimulation rate, which could result in eye movement responses with a shortened duration. This could presumably reflect a habituation effect (Jeannerod et al. 1976), degeneration of vestibular neurons (Shepard and Clark 1987), or synaptic changes such as depletion of vesicles as suggested by Davidovics and colleagues (Davidovics et al. 2011). To investigate this possibility, we tested the eVOR with velocity steps after only 30 min of tonic stimulation and found that even with acute, short-term stimulation the eVOR time constant was shorter than the input signal. While reduced sensitivity due to chronic stimulation almost certainly affects the dynamics of the VOR during long-term stimulation (Lewis et al. 2010), these acute experiments demonstrate an absence of the velocity storage effect on the low-frequency eVOR near the onset of stimulation. Habituation or cell degeneration seems unlikely over this time frame, although depletion of vesicles remains a possible explanation.

Future Directions

We have shown that an artificial rotational signal provided to semicircular canal afferent neurons via pulsatile electrical stimulation interacts with graviceptive cues in ways that mimic normal canal-graviceptive sensory integration, as evidenced by qualitatively normal tilt suppression and spatial orientation of the eVOR. It would be fruitful in future experiments to modify the characteristics of the electrical stimulation in an attempt to better simulate normal peripheral or central canal-mediated rotational cues. To improve the physiological characteristics of the afferent signal, for example, noise could be superimposed on the biphasic pulses or high-frequency stimulation could be used to desynchronize firing in the stimulated afferents. Similar approaches have been used with electrical stimulation of the cochlea (e.g., Rubinstein et al. 1999) and have modestly improved the quality of sound encoding in the brain (Morse and Evans 1996; Paglialonga et al. 2010). The dynamics of the central canal-mediated rotational cue could be adjusted by altering the time constant of the high-pass filter applied by the prosthesis to the measured head velocity. In our prior chronic experiments (e.g., Lewis et al. 2010), we chose a time constant of 5 s to simulate the dynamics of normal canal afferents. As noted in that paper and the present study, however, the eye movement responses (which reflect the central rotational cue mediated by the canal input) displayed poor low-frequency dynamics with time constants in the 3–4 s range. One option would be to filter the angular head velocity in a manner that results in eye movements with low-frequency dynamics similar to the normal VOR, rather than attempting to simulate the dynamics of the normal afferent input. Our prior work has shown that filtering head velocity with a time constant of 80 s resulted in eye movements with time constants in the normal range (Haburcakova et al. 2005), indicating that the dynamics of the central rotational cue were normalized when the peripheral input decayed much more gradually than normal. An alternate approach would be simply eliminate all high-pass filtering and allow the measured angular head velocity to directly determine the rate of electrical stimulation. Clearly, more experiments are necessary to determine which, if any, of these approaches would result in more physiological behavioral responses.

While oscillopsia is symptomatic to patients with bilateral vestibular hypofunction, they also experience markedly impaired postural control and percepts of head motion and orientation. To develop a clinically useful vestibular prosthesis, we must expand our testing beyond the VOR to include more complex and integrative behaviors such as balance and perception that depend in part on vestibular information. As emphasized above, the demonstration that prosthesis-mediated canal rotational cues interact at the oculomotor level with graviceptive cues is an important indicator that these more complex behaviors may also improve when a canal prosthesis is implemented in vestibulopathic subjects, a possibility that is supported by our preliminary postural and perceptual studies in rhesus monkeys (Lewis et al. 2011).


This work was supported by National Institute on Deafness and Other Communication Disorders Grants DC-6909 and DC-8362 to R. F. Lewis and DC-8167 to D. M. Merfeld.


No conflicts of interest, financial or otherwise, are declared by the author(s).


Author contributions: R.F.L., W.G., and D.M.M. conception and design of research; R.F.L., C.H., W.G., and D.M.M. performed experiments; R.F.L., C.H., and F.K. analyzed data; R.F.L., F.K., and D.M.M. interpreted results of experiments; R.F.L. and C.H. prepared figures; R.F.L. drafted manuscript; R.F.L., C.H., W.G., F.K., and D.M.M. edited and revised manuscript; R.F.L., C.H., W.G., F.K., and D.M.M. approved final version of manuscript.


We thank L. Zupan, D. Channer, and R. Terry.


View Abstract