Chapter 3

Adapted from: Groen E, Bos JE, De Graaf B. Contribution of the otoliths to the human torsional vestibulo-ocular reflex. Submitted to Journal of Vestibular Research

It has been assumed that ocular torsion is determined exclusively by the shear force along the maculae of the utricles, and not by the "swinging" of the resultant gravito-inertial force relative to the head. Recently, however, it was suggested that the ocular torsion response to interaural oscillation in man can be divided into a component responding to linear motion and a component responding to the swinging resultant gravito-inertial force (Merfeld et al. 1996). Correspondingly, the steady state nystagmus observed during constant velocity rotation about an off-vertical axis has been recognized as a third type of OOR, which reflects the ability of the otoliths to estimate angular velocity from the rotation of the gravito-inertial vector relative to the head (Guedry 1965; Benson and Bodin 1966; Angelaki and Hess 1996a).

Basically, linear motion and constant velocity rotation about an off-vertical axis are unnatural conditions for tilt OORs such as ocular torsion. The findings obtained with these methods suggest that otolith-induced ocular torsion does exist during rotations of the head in the roll plane, when the semicircular canals are activated simultaneously. Nevertheless, there seems to be no decisive evidence that such is the case in humans. Collewijn et al. (1985) studied ocular torsion during active head roll, but did not determine the contribution of the otoliths. Vieville and Masse (1987) made an attempt to isolate the contribution of the otoliths from that of the semicircular canals, by measuring ocular torsion during voluntary head movements in the roll plane (in only one subject), once with the head erect, and once with the head prone. In the head-erect condition the motion was about an earth-horizontal axis, modulating inputs from the otoliths and the semicircular canal. In the head-prone condition the motion was orthogonal to gravity, only stimulating the semicircular canals. In a frequency range from 0.1 to 0.4 Hz, the response gain was slightly higher and the phase lead smaller when the head was erect compared to when the head was prone. Conversely, Tweed et al. (1994) did not find a difference in three-dimensional VOR dynamics for oscillations of 0.3 Hz about an earth-horizontal axis and about an earth-vertical axis. However, taking the low-frequency characteristics of tilt OORs into account, a frequency of 0.3 Hz may have been too high to reveal an otolith effect.

The goal of the present study was to investigate whether the otoliths improve the low-frequency dynamics of the torsional VOR in human subjects. We focused on the torsional VOR because it is considerably weaker than the horizontal and vertical VOR (Crawford and Vilis 1991; Tweed et al. 1994), suggesting that especially ocular torsion could benefit from assistance of the otoliths. We compared the response to passive sinusoidal body roll about an earth-horizontal axis and an about earth-vertical axis (in upright and supine body orientation, respectively) in the frequency range from 0.05 to 0.4 Hz. At very low frequencies the canal-induced VOR normally becomes less effective, showing a smaller gain and a progressive phase lead (Carpenter 1988). Our main hypothesis was that the otoliths would enhance the gain and reduce the phase lead of the response in the lower end of the frequency range.

In a previous study we measured ocular torsion during sinusoidal body roll at 0.25 Hz about an earth-horizontal and about an earth-vertical axis, and observed that the torsional nystagmus contained less anti-compensatory saccades in the former condition, which corresponded to larger amplitudes of ocular torsion (Groen et al. 1996b). A similar finding has been reported for the vertical VOR in rabbits (Barmack 1981). We believe that modulation of the saccadic activity in vestibular nystagmus offers the possibility for the otoliths to control the position of the eyes in space, and that this may be independent of effects on the slow component (velocity) of the response. In addition to the slow component dynamics, we therefore investigated the saccadic activity and the amplitude of torsional eye position in the original nystagmus.

Each of five healthy subjects, one female and four male, 28-37 years old, participated in one experimental session of an hour with their informed consent. Seated in a rotating chair, the subjects were sinusoidally rotated about their naso-occipital (roll) axis. The subjects’ head was supported by a head rest, and the body was firmly harnessed to the chair.

Stimuli

In total, there were six different stimuli of sinusoidal roll motion. Four of them varied in frequency (0.05, 0.1, 0.2, and 0.4 Hz), but had a fixed amplitude (25°), allowing for evaluation of the frequency response. Moreover, at the frequency of 0.2 Hz there were two additional stimuli with an amplitude of 12.5° and 50°, respectively, so that the linearity of the response could be examined. Responses were measured during ten cycles at most, with a maximum duration of one minute (that is 3 cycles at 0.05 Hz, 6 cycles at 0.1 Hz, and 10 cycles at 0.2 and 0.4 Hz).

The same set of stimuli was applied in two different body orientations (Figure 3.1). In upright position, the rotation axis was earth-horizontal. Here, the stimulus modulated semicircular canal and otolith inputs. In supine position, the rotation axis was earth-vertical. Here, only the canal input was modulated. Thus, there were 12 different conditions (six motion stimuli in two body orientations). To save time of rearranging the chair between the two orientations, conditions were not randomly applied. Instead, the stimuli were divided into two blocks of three, each of which was carried out in the upright orientation first, followed by the same block in the supine orientation.

With two small synchronized CCD-cameras mounted on a head set, movements of both eyes were simultaneously recorded on video with a sample frequency of 50 Hz. To avoid artifacts from movement of the cameras relative to the head, the head set was attached to a dental frame with individual fit. Except for a point source fixed to the camera, the measurements were performed in darkness. The subjects were asked to gaze straight ahead into the cameras, using a reflection spot on the camera lens for fixation. Thus "pure" torsional movements were measured. The video recordings were digitized and analyzed off-line by means of the method described in Chapter 1. The angle of ocular torsion was determined relative to a reference image, arbitrarily chosen from the recording just prior to the stimulus onset. The eye position signals were stored in separate datafiles, synchronized with the corresponding stimulus signals (the actual rotation angle of the chair).

The datafiles were analyzed using an interactive computer program. First, we determined gain and phase of the slow component velocity (SCV). For this, the position signal was differentiated digitally, and saccades were removed by a moving median filter with a window of 500 ms (the saccades were typically 100 ms of duration, or less). Subsequently, the amplitude and phase of the component with the stimulus periodicity were estimated by means of discrete Fourier transformation. The SCV gain was computed as the amplitude ratio of eye velocity and stimulus velocity. The SCV phase was determined relative to the stimulus. Positive values were used for a phase lead, where zero phase indicated perfect compensatory eye velocity.

Although the gain and phase are interdependent parameters, we examined them separately. This implies that an ambiguous phase of a small response may have a disproportionate weight in the group’s average. Anticipating the results, however, no data had to be excluded for this reason. First, all individual responses were distinct enough to allow for reliable parameter estimation. And second, the between-subject variability was small.

Analysis of eye position

In the absence of saccades, the behavior of the eye position would be directly described by the SCV parameters. But in any nystagmus the excursion range of eye position is modified by saccades. To examine the effects of saccades, we therefore determined the gain of torsional eye position, by averaging a number of cycles of theoriginal nystagmus (including saccades) and measuring the amplitude of the approximately sinusoidal result. Analogous to the determination of the SCV gain, this amplitude then was divided by the stimulus amplitude to obtain the eye position gain.

Figure 3.1. Two pictures showing the orientation of the TNO tilt chair for whole body roll about (a) an earth-horizontal axis and about (b) an earth-vertical axis. To illustrate the camera equipment for OT recordings, the hood was removed in (a). The roll axis was more or less centered between the eyes.

The saccades themselves were characterized by their amplitude and incidence (frequency). In general, the saccade amplitude is monotonously increasing with peak velocity (Bahil et al. 1975, Lebedev et al. 1996), a rule which also holds for torsional saccades (Collewijn et al. 1985). For technical reasons we took the saccade peak velocity as a measure for their amplitude. The saccade frequency was expressed as the number of saccades per second.

In general, the above analyses were performed on the entire response, with disregard of the first cycle to allow the response to build up. Incidentally, one or more cycles had to be ignored because of artifacts (e.g. eye blinks).

Assuming additive contributions of the semicircular canals and the otoliths to the slow component, we calculated a hypothetical otolith response as the difference between the upright and the supine response. Instead of actually subtracting the original signals, we calculated averages of the individual vector differences using the available gains and phases of the SCV.

Results

Figure 3.2 shows 20 s periods of torsional nystagmus of one subject at different stimulus frequencies, with a fixed amplitude of 25°. Data are shown in both the upright and supine conditions. Similarly, Figure 3.3 shows data of the same subject at different stimulus amplitudes, where the frequency is fixed at 0.2 Hz. All responses feature an obvious sinusoidal "slow" component sharing the periodicity of the stimulus. The amplitude of the nystagmus appears to increase with increasing stimulus frequency and amplitude. In the next paragraphs the contribution of the slow component and the saccades will be quantified.

Slow component velocity

In Figure 3.4a the SCV dynamics (gain and phase), averaged over all subjects, are plotted versus stimulus frequency (with an amplitude of 25° ). The data of both eyes have been combined. The SCV gain increases significantly with stimulus frequency (within subjects design ANOVA: F=34.87; df=3,12; p<0.01). Furthermore, the average gain is significantly higher in upright than in supine orientation (F=29.80; df=1,4; p<0.01). The linear relationship between the SCV gain and the logarithm of the stimulus frequency (f) is demonstrated by the two regression lines, plotted in the upper diagram of Figure 3a: y_upright = 0.50 + 0.26 log(f), and y_supine = 0.37 + 0.21 log(f).

The corresponding phase characteristics, as presented in the lower diagram of Figure 3.4a, reveal a significant interaction between body orientation and stimulus frequency (F=11.07;df=3,12; p<0.01). In the supine orientation, at 0.05 Hz the response leads the stimulus on average by 78°, which decreases slightly at 0.1 Hz, and increases again up to 101° at 0.4 Hz. In the upright orientation, the average phase lead is only 29° at 0.05 Hz, and successively increases to 99° at 0.4 Hz. From 0.2 Hz and above, the phase leads in both body orientations are identical.

The hypothetical "otolith only" response, the vector difference of the upright and supine response, is also depicted in Figure 3.4a (dotted line). The gain of this response is about 0.1 for all frequencies, where the phase changes from a lag of about 15° at 0.05 Hz into a lead of 90° at 0.4 Hz.

In Fig. 3.4b, the SCV gain and phase are plotted versus stimulus amplitude (with frequency fixed at 0.2 Hz). There was no effect of amplitude, which implies that the response behaves linearly in both body orientations for amplitudes from 12.5º to 50º. Again, there is a main effect of body orientation on the SCV gain, which is consistently higher in upright than in supine (F=59.87;df=1,4;p<0.01).

The three parameters describing eye position behavior are plotted in Fig. 3.5a for increasing stimulus frequency, and in Fig. 3.5b for increasing stimulus amplitude. The excursions of torsional eye position are invariably smaller than the corresponding SCV gains. This confirms that the saccades are oppositedly directed to the slow component, reducing the maximum amplitude of torsion. This is best seen by comparing the upper diagrams of Figures 3.4b and 3.5b: the eye position gain decreases significantly (F=64.78;df=1,4;p<0.01), whereas the gain of the SCV remains constant with increasing stimulus amplitude. Parallel to this, the average saccade frequency increases significantly with stimulus amplitude (F=15.76;df=2,8; p<0.01), as does the peak velocity (F=49.84;df=2,8; p<0.01). Similarly, the saccade frequency increases with stimulus frequency (F=29.28; df=3,12; p<0.01), as does the peak velocity (F=23.24; df=3,12; p<0.01). In this case, however, the gain of the eye position still increases, due to the increase in SCV gain. From the original recordings, as shown in Figures 3.2 and 3.3, it is evident that the saccadic activity is higher supine than upright. In fact, the saccades seem to operate around an "offset", which appears sinusoidally modulated in the upright conditions, but seems about zero in the supine conditions. Figure 3.5 shows that this can be attributed to a difference in the saccade frequency, which is significantly higher in the supine than in upright orientation (F=26.76; df=1,4; p<0.01). On the other hand, the saccade amplitude (or "beating field") is independent of body orientation, as is indicated by the average peak velocity, which does not differ between upright and supine conditions.

Figure 3.2. Twenty seconds of torsional nystagmus of one subject at four stimulus frequencies (fixed amplitude 25 °) in the upright (left) and supine (right) orientation. The small markers below each trace indicate the moment of maximum clockwise body tilt. Upward deflection of the eye position signal corresponds to counter-clockwise rotation. (Click graphs to enlarge)

Figure 3.3. Twenty seconds of torsional nystagmus at three different stimulus amplitudes (with frequency fixed at 0.2Hz) in upright (left) and supine (right). Data are from the same subject as in Figure 3.2. The small markers below the traces correspond to maximum clockwise body tilt. (Click graphs to enlarge)

Figure 3.4. The SCV dynamics, binocular data averaged over five subjects, for the upright (filled circles) and supine conditions (open circles). In the left column (a) the SCV gain and phase are depicted versus log stimulus frequency. Here, the dotted line shows the hypothetical "otolith only" response, calculated as the vector difference between upright and supine. The solid lines in the gain plot (upper left diagram) are the fitted regression lines: y_upright = 0.50 + 0.26 log(f), and y_supine = 0.37 + 0.21 log(f). In all other plots the connecting lines between data points are drawn. In the right column (b) gain and phase are shown for the log stimulus amplitude. Positive phase values indicate phase lead relative to a perfect compensatory response. Error bars indicate S.E.M.

Figure 3.5. The eye position parameters as they vary with log stimulus frequency (a, left column) and amplitude (b, right column). The mean eye position gain (upper diagrams), saccade frequency (middle diagrams), and saccade peak velocity (lower diagrams) are shown for upright (filled circles) and supine conditions (open circles). Error bars indicate S.E.M.

In this study we compared the torsional VOR during oscillations about an earth-horizontal and an earth-vertical roll axis, that is with varying and invariant otolith input, respectively. We used passive body roll in order to minimize inputs from the neck receptors. Therefore, the response differences between the two body orientations were attributed to the contribution of the otoliths.

Slow component velocity

We found that in both body orientations the gain of the slow component velocity increased with stimulus frequency. The gains observed were within the common range of the human torsional VOR. For instance, the average SCV gain of 0.37 at 0.4Hz in the upright condition is similar to the value measured by Tweed et al. (1994) at 0.3Hz. A further increase of the torsional gain up to 0.7 has been reported for active head movements of 1 Hz (Collewijn et al. 1985; Vieville and Masse 1987). The logarithmic relation between the SCV gain and the stimulus frequency, suggested by our data, confirms these results.

In the present study, the gain was consistently higher, by an amount of 0.1, in the upright orientation than in the supine orientation. In a previous study we measured the torsional response in exactly the same way at the frequency of 0.25 Hz (Groen et al. 1996b). We then observed a higher gain in the supine conditions compared to the upright conditions. Among the 11 subjects of the previous experiment were the same five subjects of the present study. Although the difference between the two conditions was significant for the entire group of 11 subjects, it was not for the subgroup of five subjects. Later we collected more (unpublished) data of the same subjects, again at the frequency of 0.25 Hz, which were consistent with the results described here. Considering these variable results within subjects and the small effects, we do not believe that the otoliths contribute significantly to improve the gain of the (torsional) VOR in humans. Other studies have also failed to show a difference in SCV gain between earth-horizontal and earth-vertical rotation (Tweed et al. 1994; Seidman et al. 1995). In this, humans differ from lateral-eyed species, such as rabbits, where the otoliths almost completely compensate for the reduced gain of the canal-induced VOR at very low stimulus frequencies (Barmack 1981).

We adhere to a different conclusion with respect to the phase of the response: in the lower frequency range, contribution of the otoliths had a significant effect on the phase lead. At the frequency of 0.05 Hz the response to earth-vertical rotation showed a phase lead of about 80°, which was only 30° in the response to earth-horizontal rotation. Similarly, Vieville and Masse (17) reported that the phase lead of ocular torsion during active head roll in their (only) subject was larger when the head was prone than when the head was erect. There is evidence that otolithic control of the phase is not restricted to the torsional VOR. The same effect was reported for the phase of the horizontal VOR in humans during barbecue spit rotation (Benson 1970). Optimizing the phase therefore seems to represent a general mechanism of the otoliths to improve the VOR dynamics.

In the present study, there are two otolith-induced responses under consideration: a (position-dependent) response to head tilt, and a (position-independent) response to the rotation of the gravito-inertial force relative to the head ("tilt velocity"). With our experimental design however, it is difficult to discriminate the two components in the calculated "otolith only" response. This may explain why this response did not show the typical low-pass filter characteristics, which have been described for tilt OORs (Paige and Tomko 1991a; Hannen et al. 1966; Lichtenberg et al. 1982; De Graaf et al. 1996; Merfeld et al. 1996). It is possible that the observed otolith effect on the phase of the slow component exclusively resulted from the response to tilt velocity, as was argued by Angelaki and Hess (1996a). Using sinusoidal rotation with an amplitude of more than 360°, these authors demonstrated that it was the "tilt velocity" information by which the otoliths improved the low-frequency dynamics of the angular VOR in monkeys.

Theoretically, the reduction of the phase lead of the slow component may be accomplished via the neural integrator (Robinson 1971), which converts eye velocity signals into position signals. In their physiological range, the canal afferents encode for angular head velocity, which is suitable to drive the movements of the eyes, but needs to be integrated to hold the eyes in their new position after cessation of the movement. At lower frequencies, the canal signal is shifted toward acceleration (Férnandez and Goldberg 1971), as becomes clear in the progressive phase lead of the VOR. To hold eye position then, the signal would have to be integrated twice. Our data suggest that the otoliths provide supplementary (tilt) velocity information that can be used to achieve this extra integration step.

Eye position

Due to anti-compensatory saccades the eye position gain was invariably smaller than the gain of the slow component. This effect was most noticeable in the response to earth-vertical rotation: at all stimulus frequencies the nystagmus consisted of more saccades in the supine condition than in the upright condition, even though the corresponding SCV gain was higher in the latter. The amplitude of the saccades was equally large in both body orientations, indicating that the beating field itself was unaffected. Instead, it was the offset of the beating field that was sinusoidally modulated during earth-horizontal rotation. This resulted in larger accumulation of ocular torsion in the upright orientation, illustrated by the higher eye position gain. This confirmed our assumption that the otoliths modulate torsional eye position by influencing the saccadic behavior. In contrast to the observed effect on the phase of the slow component, the effect on the saccade frequency reached over the entire frequency range, so that presumably the two effects are independent.

We believe that modulation of the beating field offset is an instrument for the otoliths to maintain eye position in space. It may well be that this is common to any system associated with gaze stabilization, since it has been reported that the incidence of anti-compensatory saccades in the ocular torsion response is reduced in the presence of a visual scene (Collewijn et al. 1985). By the same token, our results indicate that stabilization of the eyes-in-space is less important when the rotation is independent of gravity. Here, the system seems inclined to compensate for head velocity, irrespective of the orientation of the eyes. From this we conclude that the canals are primarily concerned with stabilization of the retinal image in head-centric coordinates, whereas the otoliths tend to anchor the oculomotor coordinate system with respect to gravity.

Listing’s law

Although our study concerned one-dimensional eye movements, the above described effects are relevant in the light of Listing’s law, which applies to three-dimensional eye movements. According to this law, the eyes usually assume positions that can be reached from a certain reference position by rotation about an axis laying in a head-fixed plane. When this reference position is the primary position, these axes lie in a plane (Listing’s plane) perpendicular to the gaze direction (Tweed et al. 1990). As a result of the strategy as described by Listing’s law, the oculomotor system tries to avoid torsional eye positions. This implies that there is a conflict between minimizing the degree of ocular torsion and stabilizing the position of the eyes in space. This raises an alternative explanation for the low torsional gain of the VOR. Originally, it has been argued that a low torsional gain is not necessarily inadequate, because torsional movements do not alter the direction of gaze and thus preserve fixation on a stationary target (Seidman et al. 1989; Ferman et al. 1987). As a result, there would be little retinal slip in the foveal area. However, these assumptions only hold when ocular torsion is regarded as rotation about the line of sight, that is relative to the eye. There is reason to believe that the torsional VOR should be regarded in head-fixed coordinates, or, more precisely, in Listing’s coordinates (Crawford and Vilis 1991; Tweed et al. 1994). In the view of these authors, the low torsional gain signifies a trade-off between perfect (co-axial) compensation for head rotations and obeying Listing’s law. The compensatory slow component tilts the rotation axes of the eyes out of Listing’s plane, which is subsequently restored by anti-compensatory saccades. Our data confirm this behavior for the supine position. During earth-horizontal rotation, however, the (now) varying otolith contribution complicates the matter. It is questionable whether this otolith-induced ocular torsion should be considered as a violation of Listing’s law. In fact, it may establish the link between the eye in its head-centric orientation (according to Listing’s law) and the eye in the orientation to the environment (i.e. to gravity). It has been hypothesized that the static ocular torsion during head tilt may result from a shift of Listing’s plane (Vilis 1993). Indeed, the rotation axes of voluntary saccades have been shown to lie in a plane shifted in torsional direction when the subject is tilted laterally (Crawford and Vilis 1991). Similar observations have been made for static tilt in pitch (Haslwanter et al. 1992). Recently, Angelaki and Hess (1996a,b) found that in monkeys the otoliths did not only influence the slow component but also the saccades of vestibular nystagmus. They argued that the orientation of primary position is dynamically modulated by the otoliths. The modulation of the beating field offset, observed in the present study, indicates that the otoliths serve this function also in humans.

In conclusion, our data showed that the dynamic contribution of the otoliths improved the human torsional VOR dynamics at low stimulus frequencies, mainly by keeping the response in phase with the stimulus motion. In addition, the otoliths reduced the number of anti-compensatory saccades in the nystagmus, increasing the excursion range of torsional eye position. Through these effects the otoliths achieve better stabilization of the eyes-in-space.

Acknowledgments

This work was supported by the Foundation for Behavioral and Educational Sciences of the Netherlands Organization for Scientific Research. The dental frames with individual fit were provided by ACTA (Academisch Centrum Tandheelkunde Amsterdam). We are grateful to Astrid de Vries for her contribution to the data analysis. We also thank Dr. Doug Crawford for his helpful discussions.