INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION APPENDIX REFERENCES

Transmastoidal galvanic vestibular stimulation (GVS) induces eye movements with horizontal (Buys 1909) and torsional (Hitzig 1898) components by activating the vestibuloocular reflex (VOR). On the side of the cathode, the applied current is believed to depolarize the vestibular afferent's spike trigger site, which extends 10-50 µm from the synapse to the first level of myelinization (Goldberg et al. 1984). The thus depolarized afferent membrane leads to an increase in the firing rate. On the anodal side, this process is reversed. The galvanic activations are similar for the afferents of every vestibular end organ (Goldberg et al. 1984; Kleine and Grüsser 1996). Unilateral excitation of either utricular or semicircular canal (SCC) nerve branches induces compensatory eye movements with a predominantly torsional component, with the upper side of the bulbus rotating away from the stimulated side (Suzuki et al. 1969). Transmastoidal galvanic stimulation thus mimics a head movement toward the side of the cathode, with utricular fibers signaling a tilt and SCC fibers signaling a rotation of the head. Since irregular afferents have a smaller postspike recovery time constant, their galvanic sensitivity is sixfold greater than that of regular afferents (Goldberg et al. 1984). The contribution of irregular fibers to the horizontal VOR is significant if the head is rotated with velocity steps (Angelaki and Perachio 1993) but diminishes if the stimulus is sinusoidally modulated (Angelaki and Perachio 1993; Minor and Goldberg 1991).

Sustained current steps induce two distinct types of torsional eye movements: first, a tonic ocular torsion (OT) of both eyes, and second, superimposed torsional nystagmus. The tonic OT has been attributed to the activation of otolith afferents (Watson et al. 1998; Zink et al. 1997, 1998). In some individuals, GVS can evoke more tonic OT than nystagmic eye movements, whereas in others a considerable nystagmus response can be observed. Two explanations are discussed in the literature for this interindividual variability of OT patterns. One proposed explanation is that the threshold for the activation of SCC afferents varies (Zink et al. 1997, 1998). Another proposed explanation is that the contribution of the otolith afferents varies (Kleine et al. 1999). We recently proposed a third possibility, namely that interindividually variable nystagmus frequencies and amplitudes could lead to the observed phenomenon. On the basis of theoretical considerations, we concluded that the effect of SCC afferents on OT prevailed over the effect of utricular pathways, although all vestibular afferents are activated equally by GVS (Schneider et al. 2000a).

The purpose of the present study was to test this hypothesis by a direct comparison of torsional eye movements elicited by GVS and by natural SCC stimulations. If all the observed OT patterns could be explained by a dominant SCC contribution during GVS, similar results should be obtained with a pure rotational stimulation of the head that mimics the afferent firing rate induced by GVS. Since blinks during GVS induce transient torsional eye movements (Schneider et al. 1999), we additionally hypothesized that blinks should have the same effect during pure SCC stimulations. The preliminary findings of this study were presented at the 26th meeting of the Society for Neuroscience (Schneider et al. 2000b).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION APPENDIX REFERENCES

Six healthy subjects (1 female, 5 males; 32 ± 2.5 yr of age, mean ± SD) participated in the experiments. The subjects gave their informed consent after being briefed about the examination. The experiments were approved by the local ethics committee (approval numbers 87/96 and 212/96).

The subjects kneeled on a rotating chair (Toennies turntable) with their heads restrained in a pitched nose-down position by a forehead and a chin rest (schematic drawings of heads at top of Figs. 1 and 4). The distance between the chin rest and the center of rotation was 0.5 m. We determined the orientation of the Reid plane (Blanks et al. 1975) by applying one black skin marker to the inferior margin of the left orbit and another one to the left tragus at the level of the auditory canal center. After each experiment we recorded the subject's head with a video camera. From these images we calculated the angle between a vertical line in the background and a line connecting the two markers. The thus calculated mean deviation of the Reid plane from the earth horizontal was 58 ± 6°. With the angle of 25 ± 6° between the lateral SCC planes and the Reid stereotaxic planes obtained from 10 human skulls (Blanks et al. 1975), the deviation for the lateral SCC planes of the participants in our experiment from the earth horizontal was roughly estimated to be 33 ± 8°. In RESULTS we simplified the estimation of afferent GVS sensitivities by approximating these anatomical data with the mathematically more convenient alignment of SCC planes outlined in Fig. 1.

View larger version (39K): [in this window] [in a new window]   Fig. 1. Schematic orientation of all semicircular canal (SCC) planes relative to gravity (g) during a pitched forward position of the head. The head (top) is inclined forward in the sagittal plane. Each SCC (middle) is plotted with an optimal activation vector. These vectors are perpendicular to their corresponding SCC plane and point in a direction determined from the right-hand rule (circles with arrows). The activation vectors of the right labyrinth are plotted both starting from the center of their corresponding canal (middle) and along the upper edges of a cube (bottom). Only the right activation vectors are plotted along this cube. The cube diagonal is the vectorial sum of the 3 activation vectors. It has a length of 3 · l if the length of the activation vectors is l. To simplify calculations, 5 approximations were used: 1) the 3 SCC planes of the vestibular end organ of one side are perpendicular to each other, 2) all activation vectors have the same length, 3) the projections of the axis of rotation onto the individual SCC activation vectors are proportional to the activity changes of the individual SCC afferents, 4) the 2 activation vectors of an SCC pair are parallel and point in opposite directions, and 5) the vectorial sum of the three SCC activation vectors  and the earth-vertical axis of rotation, i.e., the gravity vector g, are parallel. If  is aligned with g, a rotation around g will cause equal activations of the canals of one side. R, right; L, left; A, anterior; P, posterior; H, horizontal; e.g., RPC, right posterior canal.

Two stimulation paradigms were used to induce torsional eye movements. In the first paradigm current steps of 2 mA were applied by means of two electrodes placed binaurally over both mastoids. The electrodes had a surface of 15-20 mm² and were made of rubber foam permeated with a NaCl-based electrode gel.

The second paradigm consisted of angular head rotations (aVOR) around a combined roll-yaw axis parallel to the gravity vector with the subject's head in the same position as during GVS. This head position was chosen to allow a roughly unimodal stimulation of the canals on one side, i.e., an activation (deactivation) of the right canals and a deactivation (activation) of the left canals during a (counter-) clockwise rotation around the earth-vertical axis. The stimulation profile consisted of a sustained velocity step on which a velocity ramp was superimposed. The slope of the ramp was calculated from the inverse torsion pendulum model of the cupular dynamics with a dominant time constant of  = 6 s (Dai et al. 1999; Fernández and Goldberg 1971). This model-based approach was assumed to keep the afferent activation constant at the level induced by the initial velocity step (Fig. 4, bottom row). Initially, velocity steps of  = 5°/s were used. The corresponding ramp acceleration was calculated with a = / = 5/6°/s². For the torsional eye movements obtained from aVOR and GVS, we determined means across time for both quick phase amplitudes e and slow phase velocities p. For each subject a scaling factor k = (e_GVS/e_aVOR + p_GVS/p_aVOR)/2 was calculated, then multiplied by  to obtain the step amplitude of a second aVOR stimulation. With the thus adjusted second aVOR step velocities of 4-12°/s and ramp accelerations of 0.66-2°/s², we hypothesized that torsional eye movements would be obtained with similar properties to those observed during GVS. The maximal utricular shear forces expected from the off-axis head rotation were on the order of 0.003 g and were thus negligible. Only the effects of the second aVOR stimulation are reported in RESULTS.

GVS and aVOR stimulations lasted 20 s and were repeated four times with alternating polarity or direction. In addition, the subjects were asked to intentionally blink as soon as they heard an acoustic signal, which was triggered 10 s after each stimulation onset under both conditions.

Experimental equipment

An off-the-shelf digital video camcorder (Sony DCR-7000E PAL) was used to monocularly measure movements of the left eye. The camera was attached to a rotating chair and was adjusted so that the subjects were able to look directly into the camera optics. The so-called "NightShot" functionality of the camera provided the means for recording the eyes in total darkness with an infrared illumination. A fluorescent fixation point in a straight ahead position was used to ensure pure cyclo-rotations of the eye while horizontal and vertical eye movements were suppressed. The OT was calculated from a pair of markers applied to the sclera (Fig. 2) (Clarke et al. 1999). The markers consisted of an infrared absorbing cosmetic pigment (Chronos Vision, Berlin, Germany). They were applied to the sclera by means of a sterile surgical pen. Prior to the application the pen was permeated with a solution prepared from a small amount of pigment and a drop of water. The images were recorded on tape with a sampling rate of 50 Hz (25 Hz interlaced) and were later analyzed by a custom-made image processing software package (APPENDIX and Fig. 2).

View larger version (39K): [in this window] [in a new window]   Fig. 2. Original video images with scleral markers (top) and the corresponding ocular torsion (OT) trajectories (bottom). Dark markers were artificially applied to the sclera. The marker coordinates were obtained from a center-of-intensity calculation in the 2 black colored regions of interest, which were positioned outside the left and the right edges of the iris (left image). The image processing algorithm draws crosses or lines at the detected pupil and marker coordinates. Compared with the OT trajectories, which were obtained from a cross-correlation of 16 iral segments (noisy trajectory in the left OT plot), the scleral marker method delivered OT data with a significantly lower amount of noise. When the eye is open, torsional nystagmus can be seen during galvanic vestibular stimulation (GVS; left). To obtain the OT gain and the time constant of the slow phase trajectories, a corrected OT (dashed line) is calculated by adding inverse nystagmus to the original OT (black line). Note that with an optimally chosen (fitted) time constant the corrected OT lines are continuous, even in the presence of nystagmus. A blink with a total occlusion of the pupil induced a quick phase with a large amplitude (middle). A small blink that did not occlude the pupil also induced a quick phase (right).

Both the current source for the GVS and the servo-driver for the rotating chair were controlled by a computer with an accuracy of 12 bits and an output rate of 200 Hz. With its microphone input, the video camera recorded the acoustic trigger for a blink on the tape. This audio signal was later used to synchronize the video images with the stimulation data.

Data analysis

To analyze the OT gain and dynamics, we detected torsional quick phases and added inverse nystagmus to the original OT recordings to obtain a corrected OT (APPENDIX and dashed lines in Figs. 2 and 4). From the detected quick phases, we calculated the mean nystagmus frequencies and amplitudes as well as the mean nystagmus intensity (defined as mean frequency times mean quick phase amplitude) for each subject and each stimulation. Similarly, the blink-induced quick phase amplitudes were extracted. After removal of all quick phases from the original OT, we determined the mean slow phase velocities (SPV) by digital differentiation. The slope of the SPV was deduced from a straight line that was fitted to the SPV trajectory. The nystagmus beating field was obtained by computing the mean of the nonprocessed OT from the last 7 s of the stimulus. We used a repeated measures ANOVA to determine the statistical significance of differences between the parameters obtained from the two types of stimulation. The stimulation type was used as the first repeated measures ANOVA factor, and the four repetitions were used as the second factor. Differences between parameters were considered significant at a confidence level of 5%.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION APPENDIX REFERENCES

Figures 3 and 4 clearly show that appropriately adjusted rotational stimuli were able to induce OT patterns that resembled those induced by GVS steps. During both stimulations, an interindividually variable amount of phasic OT, i.e., nystagmus, was observed, which was superimposed on a tonic OT. In both experiments blinks triggered quick phases with amplitudes larger than the quick phase amplitudes of normal nystagmus. In the subsequent paragraphs we quantitatively compare the OT parameters from GVS and aVOR, which are summarized in Table 1.

View larger version (20K): [in this window] [in a new window]   Fig. 3. Original stimulus and videooculography recordings of ocular torsion (OT) from subject TS. We were able to adjust the properties of the head rotation stimulus so that the OT patterns from GVS and natural head rotations cannot be distinguished. GVS, galvanic vestibular stimulation; aVOR, angular head rotation.

View larger version (25K): [in this window] [in a new window]   Fig. 4. Original videooculography recordings of ocular torsion (OT) from 2 subjects during GVS (left) and aVOR (right). The traces for the afferent firing rate (AFR) were estimated from the stimuli (bottom; see METHODS). A spontaneous firing rate of 90 Hz was assumed (Fernández and Goldberg 1971). While the 2 subjects show different patterns of elicited OT, the intrasubjective similarities between the aVOR and the GVS stimulations can be recognized. Subject SG has a high and almost constant nystagmus frequency during both GVS and aVOR, while subject AH shows a more infrequent and random pattern of nystagmus. In both subjects, blinks have a stronger effect on OT than normal nystagmus does. By adding inverse nystagmus to the original OT recordings, we obtained corrected OT plots (dashed lines in OT plots) that are highly correlated for both subjects with the theoretical neural integrator responses (dash-dotted lines in AFR plots). A step response of the SPV can be easily detected for subject SG, but this is more difficult for subject AH. SPV, slow phase velocity; GVS, galvanic vestibular stimulation; aVOR, angular vestibuloocular reflex.

Nystagmus analysis

We compared the nystagmus intensity, the SPV, the slope of the SPV, the corrected OT, the beating field of the original OT, and the time constant of the neural integrator dynamics for GVS and aVOR. There was no significant difference at a 5% level of confidence between the two paradigms for any of the given variables. The beating field, however, was greater during GVS than during aVOR, and the difference almost reached significance level [F(1,5) = 6.32, P = 0.054].

Although the nystagmus parameters differed from subject to subject, the correlation between GVS and aVOR was significant. To calculate the correlation coefficient for, e.g., the nystagmus intensity, we correlated the values for GVS with the values for aVOR from the column denoted with "Int" in Table 1. For nystagmus intensity, the correlation coefficient was 0.93 (P < 0.01) and for SPV 0.97 (P < 0.005). The correlation coefficient for the SPV slope was 0.88 (P < 0.05). The beating field of the original OT and the corrected OT correlated with 0.86 (P < 0.05) and 0.93 (P < 0.01), respectively. These findings imply an intrasubjective similarity between the OT patterns elicited by GVS and by aVOR (Fig. 4). The addition of artificial inverse nystagmus to the original OT recordings eliminated the influence of nystagmus beats. There was a significant correlation across time of 0.92 (P < 0.001) between the corrected OT and the theoretical response to a step input expected from the torsional neural integrator. This leaky neural integrator is known to have a time constant of around 2 s (Seidman et al. 1994). These findings are illustrated by the dashed and dash-dotted plots of Fig. 4.

Blink analysis

Eye movements following blinks consistently showed the same pattern as nystagmus: they consisted of a quick phase toward the resting position of the eye followed by a slow phase with an exponential trajectory in the opposite direction. This pattern can be observed in both the GVS and the aVOR paradigms. The compensatory algorithm worked equally well for both nystagmus and blink-related eye movements (Fig. 2, bottom row). Thus the dynamics of torsional eye movements following a blink are similar to the dynamics of torsional nystagmus. The mean SPV following a blink (3.41 ± 2.48°/s) was significantly higher [F(1,11) = 25.99, P < 0.0003] than the mean nystagmus SPV (0.95 ± 0.58°/s). Similarly, the average torsional quick phase amplitudes induced by blinks (1.99 ± 1.15°) were significantly larger [F(1,11) = 76.70, P < 0.0001] than the amplitudes of torsional nystagmus beats (0.33 ± 0.21°). The blink-induced quick phase amplitudes were not significantly different between the two paradigms [F(1,5) = 5.53, P > 0.05]. It is worth mentioning that even small blinks, which did not occlude the pupil, were observed to trigger quick phases (Fig. 2, right column).

Analysis of gains and sensitivities

For both stimulation conditions, the response of the corrected OT to the stimulus onset consisted of an exponential trajectory that saturated with time constants in the range of 2 s. The corrected OT showed a sensitivity of 1.76 ± 0.85°/mA for GVS and a gain of 0.39 ± 0.16 for aVOR.¹ The SPV sensitivity to GVS was 0.45 ± 0.26°/s/mA, and the SPV gain for aVOR was 0.22 ± 0.11. From the two slow phase velocities p_GVS and p_aVOR the galvanically induced increase in afferent spike frequency f_GVS can be estimated by f_GVS = f_aVOR · p_GVS/p_aVOR. The naturally induced spike frequency f_aVOR can be estimated on the assumption that the sensitivity of canal afferents to head rotations in humans is similar to the sensitivity of s_SCC = 0.55 Hz/deg/s measured in monkeys (Fernández and Goldberg 1971). The initial step velocity  of the rotational stimulus thus leads to an afferent activation of f_aVOR = s_SCC · /3 (Table 1, last column). For simplicity, the last equation assumes the geometrical alignment of SCCs given in METHODS (Fig. 1). Since  in this approximated configuration acts along the vectorial sum of all activation vectors, the projection of the total activation caused by  onto a single SCC vector can be calculated with /3. By inserting the equation for f_aVOR into the equation for f_GVS, the mean sensitivity of the SCC afferents to transmastoidal currents was roughly estimated to be 1.00 ± 0.46 Hz/mA if the irregular afferent fibers are not involved in these eye movement responses. Alternatively, the galvanic sensitivity can be calculated from the corrected OT, instead of the SPV, by using the same set of equations. With this procedure a value of 1.26 ± 0.34 Hz/mA was obtained, which was not significantly different from the SPV deduced value [F(1,5) = 4.52, P > 0.05].

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION APPENDIX REFERENCES

Relative otolith and SCC contributions

The hypothesis behind our study was that head rotations around a combined roll-yaw axis should be able to elicit torsional eye movements with properties similar to those seen during GVS. The statistical analysis of the OT recordings indeed revealed no significant differences between the two conditions. This result supports our hypothesis. It should be stressed that these head rotations, which are known to activate only canal pathways, were able to cause a prolonged tonic OT. The presence of this tonic component was the basis for the conclusion that the activation of otolith pathways must play a significant role during GVS (Watson et al. 1998; Zink et al. 1997, 1998). Similarly, posturographic studies have revealed a maintained body tilt during GVS, an observation from which the same conclusion regarding an otolith contribution was drawn (Day et al. 1997; Inglis et al. 1995). However, there is evidence even in the older literature that prolonged tilt reactions can be induced by pure rotational stimuli as well. For example, optokinetic patterns that constantly rotate around the roll axis have been shown to induce both a deviation of perceived gravity by 40° and a maintained postural tilt (Dichgans et al. 1972). These patterns cannot only induce offset positions of OT, as has recently been shown by Lee et al. (2000), but also torsional eye movements with similar properties as those known from GVS (Romberg and Ohm 1944). In addition, it has long been known that SCCs contribute to the perception of the visual vertical (Stockwell and Guedry 1970). More recent model-based studies indicate that (post-) rotational stimuli have a systematic influence on the internal estimate of the gravito-inertial force and hence on the properties of the VOR (Merfeld et al. 1999, 2001; Zupan et al. 2000). Specifically, the VOR model of Merfeld (1995) predicts a deviation of the gravity estimate in the presence of pure SCC stimulation as applied in our study. This deviation may cause a tonic OT. Thus the observation of a tonic OT does not necessarily justify the view that otolith pathways must be involved in the GVS-induced eye movements.

Pure SCC stimulations and GVS induced OT patterns with similar phasic and tonic parameters. Although these parameters vary among subjects, a statistical analysis has shown that there is no intrasubjective difference between the two paradigms. From a statistical point of view, this result may appear trivial. Since the stimulation velocity was individually adjusted to approximately obtain the same eye movements as those seen during GVS, the independent sample assumption required for a statistical test was violated for mean SPV and mean nystagmus amplitude. Not surprisingly, the differences between these parameters did not reach significance level. The surprising result of this study, however, is that it was possible to adjust the rotational stimulus by a single scaling factor so that GVS and aVOR-induced eye movements became statistically indistinguishable for all analyzed parameters. In the context of the GVS-related literature, which consistently attributed the tonic components of GVS responses to otolith activations (Day et al. 1997; Inglis et al. 1995; Kleine et al. 1999; Watson et al. 1998; Zink et al. 1997, 1998), it was surprising to measure eye movements that we were familiar with from GVS, although only canals and no otoliths were stimulated. The statistical analysis of the data obtained from aVOR stimulation with intrasubjectively adjusted amplitudes shows that with appropriately chosen angular head velocities it was possible to generate eye movements that are statistically indistinguishable from GVS-induced eye movements.

These similarities lead to the conclusion that during GVS the activation of SCC afferents alone is sufficient to induce the observed OT patterns; an otolith contribution is not necessary for either the so-called tonic OT (Watson et al. 1998; Zink et al. 1997, 1998) or the variability of tonic versus phasic OT seen among subjects (Kleine et al. 1999). However, if utricular afferents contributed to the OT during GVS, this effect should become apparent in a larger corrected OT and a higher beating field of the original OT, i.e., in a higher tonic component. This increase can be expected since a galvanic stimulation of the utricular nerve branch alone is known to induce OT in the same direction as a galvanic stimulation of the afferents originating from vertical SCCs (Suzuki et al. 1969). Since transmastoidal GVS resembles a combined utricular and SCC activation (Goldberg et al. 1984; Kleine and Grüsser 1996), the otolithic and SCC effects might be additive and thus lead to a higher OT than during a pure SCC stimulation. Consequently, we observed slightly increased values for both the beating field and the corrected OT during GVS. Similarly, the afferent galvanic sensitivity estimated from the corrected OT was increased compared with the sensitivity deduced from the SPV. The increases were on the order of 10-26%, but did not reach significance at a confidence level of 5%. This may be the amount by which otolith pathways contributed to the OT during GVS. At first glance it might appear contradictory that, on the one hand, otolith and SCC fibers are activated equally by GVS (Goldberg et al. 1984; Kleine and Grüsser 1996) but that, on the other hand, OT is dominated by SCC effects. This apparent contradiction is resolved if we take into account that the known gains and sensitivities of the utricular (Bucher et al. 1992; Clarke et al. 1999; Fernández and Goldberg 1976) and SCC (Fernández and Goldberg 1971; Seidman et al. 1994; Tweed et al. 1994b) pathways differ by a factor of more than 3 for the torsional VOR. A change of 1 Hz in the SCC fiber activities will roll the eye by an amount of 0.95°, while the same change in the activation of utricular fibers will roll the eye by an amount of only 0.27°. The observation of eye movements dominated by SCC inputs is thus compatible with the notion of equal otolith and SCC fiber activation during GVS. In an earlier study (Schneider et al. 2000a) we estimated from these theoretical considerations that the relative contribution of otolith afferents to the observable GVS-induced OT amounts to a maximum of 22% if the fibers of all vestibular end organs are activated equally and if there is no cross striolar inhibition between utricular fibers. However, such an inhibition (Uchino et al. 1999) may lead to an even smaller contribution of the utricles, but this cannot be resolved with the current data.

Variable nystagmus patterns

On the basis of our observation that during the two different stimulation paradigms the subjects exhibit similar OT patterns, we conclude that sustained GVS steps with amplitudes of 2 mA constitute a stimulus that is similar to an aVOR around a combined roll-yaw axis with initial step amplitudes ranging from 4 to 12°/s. We measured a torsional SPV gain of 0.22 at these stimulation levels. This value is significantly smaller (P < 0.05, t-test) than the torsional VOR gain of 0.37 reported earlier at higher angular velocities of 37.5°/s (Tweed et al. 1994b). In contrast, the gain of 0.39 of the corrected OT is comparable to known values for the gain of the torsional VOR (Seidman et al. 1994; Tweed et al. 1994b). The mechanism that may lead to a decreased SPV gain is best illustrated in the left OT plot of subject AH (Fig. 4). During the first 10 s of stimulation, only three nystagmus beats were detected. Without these quick phases the OT would theoretically show a pure low-pass response with a saturated period after about 6 s. Once saturation is reached, the derivative of this theoretical OT, i.e., the SPV, asymptotically approaches zero. If we calculated the SPV gain only for the saturated period of the theoretical OT, it would approach zero. In contrast, subject SG seems to have a totally different nystagmus pattern. His more frequently triggered quick phases keep the OT from becoming saturated and thus the SPV gain always remains different from zero. There are even cases where both patterns can be observed in the same subject or patient (APPENDIX, Fig. 5). The data from Fig. 5 clearly show that without nystagmus the SPV gain approaches zero. In contrast, after quick phases with high amplitudes, e.g., after blinks, the SPV showed significantly increased values compared with the SPV following "normal" nystagmus beats.

View larger version (20K): [in this window] [in a new window]   Fig. 5. Example for the processing involved in the nystagmus compensation algorithm. The original OT data (A) was obtained from a patient with a vestibular imbalance during GVS steps of ±3 mA (A, bars). The amplitudes of nystagmus quick phases were detected and plotted as bursts with inverted directions (B). After the bursts were processed by a leaky integrator with a gain g and a time constant , inverse nystagmus was obtained (C). When the inverse nystagmus of C was added to the original OT of A, a corrected OT is obtained (D, continuous line). For direct comparison, the original trace from A is repeated in D (dashed line). The necessary calculations (a summation and a leaky integration) are schematically outlined on the right side of the figure. In a last step (not shown) the parameters g and  are optimally adjusted by a least-squares procedure to minimize the error between the corrected OT and a low-pass filtered version of the stimulus. OT, ocular torsion; GVS, galvanic vestibular stimulation.

These observations lead to the conclusion that the SPV gain of the torsional VOR is inherently related to both the size and the frequency of the quick phases. It is thus dependent on nystagmus intensity, which is the product of mean quick phase amplitude and frequency. The role of the quick phases in the processing of the VOR has been emphasized before (Katsarkas et al. 1991). Besides the more conventional approach of relating SPV to head velocity, the gain of the VOR can be calculated alternatively by relating eye position to head displacement (Seidman et al. 1994). In fact, our method of calculating the corrected OT (APPENDIX) is similar to the method Seidman et al. (1994) used for obtaining pure position recordings of torsional eye movements. While Seidman et al. (1994) eliminated those recordings from the analysis that were interrupted by nystagmus beats, we eliminated all nystagmus beats from all our recordings. As a result of the nystagmus elimination, the torsional VOR gains of 0.37 and 0.39 obtained by the two methods were comparable. They were even comparable to the SPV gain of 0.37 measured at higher velocities than the velocities we used (Tweed et al. 1994b). But all these gains significantly differ from both the SPV gain of 0.22 we measured in the aVOR paradigm and the SPV gain measured by Peterka (1992) at slow head velocities. While the gain deduced from torsional eye positions does not seem to depend on stimulus amplitude, the torsional SPV gain has a nonlinear characteristic with decreasing values at smaller stimulus amplitudes (Peterka 1992). Since the crucial difference between the two methods of calculating the gain is whether or not the effects of quick phases are eliminated, we conclude that the nonlinear characteristics of the SPV gain of the torsional VOR may be the result of a nonlinear nystagmus processing rather than a nonlinearity in, e.g., the direct or the neural integrator pathway of the torsional VOR (Robinson 1981; Seidman et al. 1994).

With this in mind, it seems plausible to attribute the smaller torsional SPV gain to a nystagmus pattern that is less compensatory at the used slow head rotations. Additionally, the nystagmus pattern may vary from subject to subject and thus lead to the observed interindividual variability of OT patterns. Furthermore, a variable nystagmus pattern is suspected to have a strong effect on the phase of the SPV during dynamic stimulations (Galiana 1991). According to previously published data (Kleine et al. 1999), subjects stimulated with sinusoidal GVS indeed exhibit a torsional SPV phase with a considerable amount of variability.

Nystagmus threshold

In the present study we developed a new method of calculating the OT from artificial landmarks on the sclera. This method provided torsional data with a high signal-to-noise ratio. In spite of using an inexpensive camera, we were able to detect even small quick phase amplitudes down to 0.06°, which would have gone undetected if we had used a cross-correlation of iral landmarks. In the left plot of Fig. 2, for example, three of four nystagmus beats are hidden by the high noise level in the OT trace obtained from iral landmark detection. Hence, a meaningful measure for, e.g., the nystagmus intensity or a nystagmus threshold, can be given only on the basis of a method that ensures that all or at least the majority of the occurring quick phases can be discriminated from the underlying noise. In this context, it becomes clear that the concept of individually variable thresholds for nystagmic OT responses to GVS (Zink et al. 1997, 1998) must be reconsidered.

Blink-triggered quick phases

Our findings suggest that blinks are able to trigger torsional quick phases when the SCC afferents are stimulated either by head rotations in the torsional plane or by transmastoidally applied currents. However, blinks do not have any effect during ocular counter-roll induced by a static head tilt of 15° around the nasooccipital axis (unpublished observation). Thus blink-triggered torsional quick phases resemble torsional corrections during saccadic eye movements, which are elicited when torsion is induced by, e.g., optokinetic stimulation in roll (Lee et al. 2000), but not during static head tilt in roll (Klier and Crawford 1998). This is an additional indication that the GVS-induced tonic OT primarily arises from SCC pathways. Exponentially shaped OT trajectories after blinks have been observed before (Ferman et al. 1987; Schneider et al. 1999). It has recently been shown that blinks may trigger saccades by inhibiting the omnipause neurons (Goossens and Van Opstal 2000). Similarly, the effect of blinks on pathological saccadic eye movements has been attributed to the modulation of saccade-related pause cells by blink activity (Hain et al. 1986). The pathways responsible for these effects might be common projections of the extraocular muscle and levator palpebrae efferents to rostral mesencephalic regions (Horn et al. 2000). A blink-triggered torsional quick phase could be physiologically meaningful, since rapid reorientation after a blink can be crucial during head movements. One possible strategy would be to generate a quick phase, e.g., at the time the eyelid is opened. Then the following compensatory slow phase keeps the image stable on the retina. In addition, the saccadic visual suppression mechanism may be used by the blink system during the time the eyelid is closed (Ridder and Tomlinson 1993).

Galvanic sensitivity of vestibular afferents

In an earlier study (Schneider et al. 2000a) we estimated the effect of GVS on vestibular afferents. We calculated that a current of 1 mA triggers an increased firing rate of 0.76 Hz. There was no significant difference at a 5% level of confidence (t-test) between this value and the value of 1 Hz/mA estimated in the present study from the SPV. To calculate this sensitivity we have used the approximations outlined in Fig. 1 to obtain the simple equations given in RESULTS section. In view of the departure of these approximations from the true anatomical SCC orientations (Blanks et al. 1975; Rabbit 1999), this value can be regarded only as a very rough estimate of the magnitude of the afferent sensitivity to transmastoidally applied currents. An additional error, which might have been introduced by an inaccurately estimated cupular time constant, can be rejected, since the SPV slopes were not significantly different for the two stimulation conditions.

In a further approximation we considered only regular afferents, although it is known that irregular afferents contribute to the horizontal VOR (Angelaki and Perachio 1993) with a sixfold increased galvanic sensitivity compared with the sensitivity of regular afferents (Goldberg et al. 1984). Angelaki and Perachio (1993) showed that functional ablation of irregular afferents leads to a significant decrease in the gain of the horizontal VOR if angular velocity steps were applied. A change in horizontal VOR parameters, however, has not been observed during sinusoidal stimulations with frequencies of 0.5-4 Hz (Angelaki and Perachio 1993; Minor and Goldberg 1991). A tentative explanation of these rather contradictory observations was that regular afferents drive the direct pathway of the horizontal VOR, while irregular afferents have an influence on the indirect pathway of the velocity storage system, which (by resembling the characteristics of a low-pass filter) artificially increases the cupular time constant (Angelaki and Perachio 1993). Since velocity storage has not been observed for the torsional VOR (Tweed et al. 1994a), we believe that it is justified to disregard irregular afferents in our approach to estimate the galvanic sensitivity from torsional eye movement data.

Conclusion

Our study shows that the ocular torsion-related effects of constant galvanic vestibular stimulation with tolerable currents are similar to an accelerated head rotation at small amplitudes, which stimulates all semicircular canal afferents. The applied method of corrected ocular torsion provides comparable results showing a similar time course for all subjects even though the raw ocular torsion data may differ markedly. Thus the present study may be a basis for the use of galvanic stimulation at the mastoid level as a tool for the clinical examination of vestibular function.