Experimental Brain Research

, Volume 185, Issue 4, pp 553–562

Vestibulo-ocular responses to vertical translation in normal human subjects


  • Ke Liao
    • Department of Biomedical EngineeringVeterans Affairs Medical Center and University Hospitals, Case Western Reserve University
  • Mark F. Walker
    • Department of NeurologyVeterans Affairs Medical Center and University Hospitals, Case Western Reserve University
  • Anand Joshi
    • Department of Biomedical EngineeringVeterans Affairs Medical Center and University Hospitals, Case Western Reserve University
  • Millard Reschke
    • Neurosciences LaboratoriesJohnson Space Center
    • Department of NeurologyVeterans Affairs Medical Center and University Hospitals, Case Western Reserve University
    • Department of Biomedical EngineeringVeterans Affairs Medical Center and University Hospitals, Case Western Reserve University
Research Article

DOI: 10.1007/s00221-007-1181-z

Cite this article as:
Liao, K., Walker, M.F., Joshi, A. et al. Exp Brain Res (2008) 185: 553. doi:10.1007/s00221-007-1181-z


Prior studies of the human translational vestibulo-ocular reflex (tVOR) report that eye rotations amount to less than 60% of those required to keep the eyes pointed at a stationary visual target, unlike the angular VOR (aVOR) which is optimized to maintain stable gaze. Our first goal was to determine if the performance of the tVOR improves when head translations are combined with head rotations in ambient lighting. A second goal was to measure tVOR during vertical head translations (bob), which has not received systematic study. We measured tVOR alone and in combination with the aVOR in 20 normal human subjects, aged 25–72 years, as they sat on a moving platform that bobbed at 2.0 Hz while rotating horizontally (yaw) at 1.0 Hz. When subjects viewed a visual target at 2 m, median “compensation gain” (eye rotational velocity/required eye rotational velocity to maintain foveal target fixation) was 0.52 during pure bob and 0.59 during combined bob–yaw; during viewing of a near target at ∼17 cm, compensation gain was 0.58 for pure bob and 0.60 for combined bob–yaw. Mean phase lag of eye-in-head velocity for the tVOR was ∼19° with respect to the ideal compensatory response, irrespective of whether translation was accompanied by rotation. Thus, the tVOR changed only slightly during translation–rotation versus pure translation, and our subjects’ ocular rotations remained at about 60% of those required to point the eyes at the target. Comparison of response during binocular or monocular viewing, and ambient or reduced illumination, indicated that relative image motion between the target and background was an important determinant of tVOR behavior. We postulate that tVOR evolved not to stabilize the image of the target on the fovea, but rather to minimize retinal image motion between objects lying in different planes, in order to optimize motion parallax information.


Vestibulo-ocular reflexLocomotionMoving platform


During natural activities, such as locomotion, head perturbations occur that have both rotational and linear (translational) components (Grossman et al. 1988; Pozzo et al. 1990). The vestibulo-ocular reflexes generate eye rotations to compensate for such head perturbations at short latency. The angular vestibulo-ocular reflex (aVOR) has been shown to produce eye rotations that compensate for angular head perturbations, thereby guaranteeing clear vision during natural activities (Grossman et al. 1989; Moore et al. 1999). The translational vestibulo-ocular reflex (tVOR) has also received study, mainly using fore-aft and side-to-side head perturbations (Israël and Berthoz 1989; Schwarz and Miles 1991; Gianna et al. 1997; Ramat and Zee 2003; Ramat et al. 2005; Angelaki 2004). Although its latency is short—about 19 ms in humans (Ramat and Zee 2003)—prior studies have reported that the tVOR typically generates eye rotations that are less than 60% of that required to hold the eye on target (Hirasaki et al. 1999; Ramat and Zee 2003). This behavior predicts that vision will be degraded during linear head motion, particularly when viewing near objects. Yet why this should be the case is poorly understood.

It is possible that the performance of the tVOR is better during locomotion and other natural head movements than has been measured in the laboratory. First, in most prior studies, the tVOR has been measured in darkness or in a very limited visual environment. In contrast, daylight vision is rich in information about the three-dimensional (3-D) location of objects. It might be that this additional visual information can be used by the brain to generate a more nearly compensatory tVOR. Second, during locomotion, pure head translation is rare. It has been suggested that under natural conditions, combining head translations with rotations may improve gaze stability (Ramat and Zee 2003).

Thus, one goal of this study was to determine whether full ambient illumination and simultaneous rotation causes the tVOR to hold the eyes on target. A second important goal was to characterize in detail responses to vertical (bob) translations, because these are prominent during locomotion, due to the straight-legged gait of humans (Massaad et al. 2007), and because they have only rarely been studied in humans (Paige 1989). For the rotational stimulus, we chose horizontal head rotations (yaw) because they were easier to control with our apparatus. We selected frequencies of bob (2 Hz) and yaw (1 Hz) that were similar to those reported to occur naturally during walking (Grossman et al. 1988; Pozzo et al. 1990). We found that neither a full visual environment, nor simultaneous head rotation improved the vertical tVOR: it still only generated eye rotations that were about 60% of those required to hold the eye on target. This finding led us to our final goal: to investigate the nature of the visual cues used to set the magnitude of tVOR responses and to re-evaluate the role of tVOR during natural activities. Preliminary findings are reported in an abstract (Liao et al. 2007b).



We studied 20 healthy human subjects (8 female) age range 25–72 years (median 55 years). No subject was taking medicines with effects on the nervous system or wore a refractive correction greater than four diopters. Refractive corrections were not worn during testing, and all subjects reported that they could easily see the visual targets, including the near targets. Experiments were performed in ambient light, so that natural visual cues, such as motion parallax and relative size, were available, and the safety of the subject on the platform could be monitored by one of the investigators who stood by the platform with an emergency stop switch. All gave informed, written consent, in accordance with the Declaration of Helsinki and the Institutional Review Board of the Cleveland Veterans Affairs Medical Center.

Vestibular stimuli

Subjects sat in a chair on a Moog 6DOF2000E electric motion platform (East Aurora, New York) that could move with 6° of rotational and translational freedom through a range of ±20° and ±20 cm, with peak rotational acceleration of 400°/s2 and peak linear acceleration of 5 m/s2 (0.5 g). Belts were used to secure the subject’s torso and a snugly fitting skate-board helmet, inlaid with foam, was used to stabilize the subject’s head. Any head movements that were decoupled from chair or platform motion were measured, as described below.

Visual stimuli

There were two main visual conditions to compare tVOR performance during bob versus combined bob–yaw under ambient illumination. (1) Subjects binocularly viewed a laser spot projected on a wall at a distance of 2 m (“far target”). (2) Subjects binocularly viewed a “near target” (reflective ball, diameter 1 cm) suspended at a distance of ∼17 cm in front of their left eye. All 20 of our subjects, including more elderly individuals, were easily able to view these visual stimuli without refractive correction. The actual positions of the near target, for each subject, were measured directly as described below.

In order to investigate visual factors that might determine tVOR behavior, 20 subjects also viewed the near target monocularly (right eye occluded). Since convergence decreased during monocular viewing, six subjects viewed targets binocularly at 2 m, 40 and 17 cm first directly and then with a 15- or 10-diopter base-out prism placed before the right eye (prism power selection was based on each subject’s ability to fuse the visual stimulus). Thus, each stimulus was viewed binocularly at one distance with two different vergence angles. In two subjects, we turned the room lights out for periods of 2–4 s, as subjects attempted to fix upon the remembered location of the near target, which they had previously viewed binocularly. These two subjects also binocularly viewed the near target under conditions of strobe illumination, in order to minimize retinal image slip information. The strobe illumination was achieved using an array of bright light-emitting diodes, which were illuminated at a flash rate of 4 Hz, with a 30 ms flash duration. Finally, to determine whether tVOR behavior in ambient lighting could be explained simply by the contribution of smooth visual tracking, 13 subjects followed a moving visual stimulus (Amsler grid), subtending 25.6° horizontally and 18.6° vertically with a central dot, at a target distance of 110 cm. The stimulus moved sinusoidally, in the vertical plane: (A) through ±9.0° at 0.2 Hz (peak velocity 11°/s); (B) through ±5.6° at 2.0 Hz (peak velocity 70°/s); (C) through ±2.8° at 2.0 Hz (peak velocity 35°/s). The first two of these moving visual stimuli imposed the same requirements on eye movements as those imposed by the translation stimuli (see next section), if there were no vestibulo-ocular responses; the last stimulus corresponded to the remaining visual motion if tVOR compensated for half of that required to hold the eye on target.

Experimental paradigms

Each experimental run, which lasted 90 s, started with three cycles of bob at 0.2 Hz (typical amplitude ±5.6 cm) followed, after a pause of 3 s, by three cycles of yaw at 0.2 Hz (typical amplitude ±6°). We assumed that our normal subjects could continuously view the visual target during these 0.2 Hz stimuli (due to normal smooth pursuit) and used their eye movements as one index of “ideal” responses. Then, we applied bob translations at 2 Hz (typical amplitude ±1.5 cm) for 12 s to test tVOR. After a 3-s pause, we applied yaw rotations at 1.0 Hz (typical amplitude ±5°) for 12 s, to test aVOR. Finally, after a 3-s pause, we applied combined bob at 2 Hz and yaw rotation at 1 Hz (starting at zero phase difference) for 12 s.

Measurement of eye and head movements

Three-dimensional eye rotations were measured using the magnetic search coil technique. Three orthogonal magnetic fields, top/bottom, left/right, front/back oscillating, respectively, at 60, 90, and 135 KHz, were implemented in a 76 cm cube (CNC Engineering, Seattle, WA) rigidly attached to the chair mounted on the platform. Dual scleral search coils capable of measuring 3-D rotations (Skalar, Delft, The Netherlands) were calibrated prior to each experimental session. First, signal offsets were nulled with the coil inside a metal tube that shielded it from the magnetic fields. Next, the relative gains of each channel were determined by aligning the coil with each of the three field and recording the corresponding maximum signal. These gains are used to normalize the raw coil signals when calculating rotation vectors. Following calibration, a scleral coil was placed on each eye following application of topical anesthesia. A coil was also taped to the subject’s forehead to detect any rotations due to incomplete head stabilization. Linear and rotational movements of the chair frame and subject’s head were monitored by an infrared reflection system (Vicon Motion Systems, Los Angeles, CA). Six reflective markers were attached by adhesive tape to the subject’s forehead and skin over the zygomatic malar processes (cheeks). Rotational and translational movements of the coil frame were monitored by attaching the four reflective markers on the coil frame. For each subject, before experiments were started, two extra reflective markers were attached over the subject’s eyelids so that we could calibrate the geometric relationship of the subject’s eyes to the facial markers. In addition, the position of the near target was measured. Six cameras allowed head and coil frame movements to be measured with a resolution of 2 mm and 0.1°.

Data analysis

Coil signals were digitized at 500 Hz with 16-bit precision after Butterworth filtering (0–150 Hz) to avoid aliasing and were saved on computer disk for subsequent analysis. Raw coil signals were normalized by the recorded gains and converted to rotation matrices and then 3-D rotation vectors in degrees (Haustein 1989), using a straight-ahead reference position at a distance of 2 m; 3-D angular velocity vectors were calculated from the rotation vectors (Hepp 1990). Both eye and head rotations were expressed in the same earth-fixed coordinate system (see Appendix in Electronic supplementary material). The horizontal vergence angle was calculated as the difference in the horizontal components (left eye–right eye) of the eye orientation vector; because the Haustein (1989) correction was applied, this measure took into account the effect on horizontal gaze position of an ocular rotation about the head-fixed torsional axis, when the eye was looking up or down. Because reference positions were recorded at a target distance of 2 m, not at optical infinity, the vergence angle corresponding to this distance and each subject’s interpupillary distance was added to the calculated values. Positive values correspond to leftward, downward, and clockwise rotations from the subject’s viewpoint (Steffen et al. 2000; Straumann et al. 2003), and divergence. Position signals from the infrared reflection system were digitized at 120 Hz, and used to calculate the movements of the subject’s head, the coil frame, and “required eye rotations” to hold gaze (corresponding to the line of sight) on the visual target (see Appendix in Electronic supplementary material). A linear accelerometer (Crossbow Technologies) was mounted on the platform to measure the vertical acceleration induced by a step movement at the beginning of each trial, and thereby synchronize coil system signals with those from the Vicon system.

Measurement of subjects’ responses

Complete head stabilization during these experiments was difficult to attain (even with a bite bar which, in preliminary studies, subjects could not tolerate during the head perturbations). Therefore, we used the infrared motion detection system to measure actual head perturbations. Most subjects’ head movements were small with respect to the coil frame (typically <1° rotations and <2 mm translation in each plane). We computed eye-in-head movements as rotation vectors, and desaccaded records using a manually selected velocity threshold for each eye movement session. We computed head rotations and translations in space using the infrared motion detection system (see Appendix in Electronic supplementary material). We then carried out Fourier transforms of eye and head velocity, measuring the response at the frequency of the stimulus. The responses around the stimulated frequency were also examined, but no significant values were found, i.e., responses were limited to the frequency of the applied stimulus.

We quantified the responses in two ways. First we measured gain of aVOR as eye-in-head rotational velocity/head rotational velocity, and responsivity of tVOR as eye rotational velocity/head translational acceleration. Note that aVOR gain has no units but tVOR responsivity has units of degrees/second of eye rotation per meters/second2 of head translation (hereafter stated as degrees × seconds/meter). The utility of aVOR gain and tVOR responsivity are that they provide a direct measure of changes in the absolute magnitude of the responses as a function of target distance. We also calculated the ratio: eye rotational velocity/required eye rotational velocity to maintain foveal fixation of the visual target (far or near), hereafter referred to as compensation gain, similar to prior usage (Ramat and Zee 2003; Ramat et al. 2005). This measurement allowed us to relate measured responses to the ideal response (1.0), assuming that the goal of tVOR is to hold the eye on target. The gain and phase lag of smooth tracking with respect to visual target motion were calculated by desaccading eye velocity data, and computing Fourier transforms.


tVOR responses during binocular viewing in ambient illumination

Representative records from one subject while viewing the near or far targets during either bob or combined bob–yaw are shown in Fig. 1; note that, apart from vergence, individual traces have been offset to aid clarity of display. Both tVOR responsivity (indicated at bottom) and aVOR gain (indicated at top) increased during viewing of the near target (17 cm) versus the far target (2 m). In addition, tVOR responsivity increased slightly during combined bob–yaw compared with during bob.
Fig. 1

Representative records from one subject showing responses to bob or combined bob–yaw during viewing of far and near targets; note that, apart from vergence, individual traces have been offset in position to aid clarity of display. Both tVOR responsivity (indicated at bottom of panels in units of degree × seconds/meter) and aVOR gain (indicated at top of panels) increase during near versus far viewing. In addition, modest increases in tVOR responsivity occur during combined bob–yaw versus bob. Positive values correspond to leftward, downward, and divergence movements. Required eye movements were calculated (see text and Appendix in Electronic supplementary material)

Results from all 20 subjects are summarized in Fig. 2. This shows a large increase of tVOR responsivity during binocular viewing of the near target at 17 cm (diamonds) versus the far target at 2 m (circles), although less than the ideal calculated response (gray line). The aVOR also showed gain increases during binocular near viewing. Paired comparisons of all data (Wilcoxon signed rank test) indicated a statistically significant increase of tVOR responsivity (P = 0.003) during combined bob–yaw versus bob (12% median increase in responsivity). There was also a small but significant (P = 0.001) decrease in aVOR gain during yaw–bob versus yaw (2% median decrease in gain).
Fig. 2

Summary of tVOR responsivity and aVOR gain from all 20 subjects, during binocular viewing of the target at 2 m (Far), or the target at 17 cm binocularly (NearBinocular) or monocularly (NearMonocular). Responsivity of both aVOR and tVOR increased with vergence angle; solid lines are linear regressions (intercept A, slope B, in bottom right corners) and dashed lines are 5 and 95% prediction intervals. The gray lines, labeled “ideal” are based on geometric predictions. The ideal response for aVOR was calculated from: \( \frac{{D(D + R)}} {{(D^2 + I^2 /4)}}, \) for which D is the distance between eye and target, R is the distance from eye to the center of rotation (assumed average value of 9 cm), I is the inter-pupillary distance (mean 6.4 cm for our subjects). The ideal response for tVOR was calculated from: \( \frac{1} {{D*2*\pi *f}}, \) for which D is the distance between eye and target, and f is the frequency of the stimulus in Hz. Also plotted are data from several trials from two subjects who attempted to view the target at 17 cm either during strobe illumination (NearStrobe) or after switching to darkness (Darkness). See text for details

Phase lags for tVOR (with 0° being ideal) were similar for different viewing conditions being 18.9° (±12.7) during far viewing and 18.7° (±10.9) during binocular near viewing. Phase lags of aVOR with respect to ideal response were small, being 1.6° (±4.8) during binocular far viewing and 2.7° (±4.2) during binocular near viewing. Phase lags of tVOR for combined translation–rotation versus pure translation were similar, with median differences being <2.6° for each viewing condition.

How much did the increases in tVOR responsivity during combined bob–yaw contribute to generating eye rotations to hold the eyes on target? When subjects viewed the target at 2 m, compensation gain increased from a median of 0.52 during bob to 0.59 during combined bob–yaw (P < 0.01). When subjects viewed the target at 17 cm, compensation gain increased from a median of 0.58 during bob to 0.60 during bob–yaw (difference not significant; average value: 0.57). Thus, the tVOR changed only slightly during translation–rotation versus pure translation, and our subjects’ ocular rotations remained at about 60% of those required to point the eyes at the target. The compensation gain of aVOR or tVOR for the ten subjects aged 25–55 years did not differ statistically, for any visual test condition, from subjects aged 56–72 years; furthermore, vergence angles were similar under the same viewing condition in each age group.

Since the magnitude of the increase in responsivity from far to near was large (a median factor of 8.7 for combined bob–yaw), this suggested that the brain was controlling tVOR behavior so that eye velocities were about 60% of those required to hold the eyes on target. This prompted us to approach tVOR differently, and investigate which visual factors contributed to this response.

Comparison of tVOR responses during different viewing conditions

First, we asked whether monocular viewing conditions would affect tVOR behavior during viewing of the near target. Figure 2 summarizes responses of all 20 subjects (squares) and shows that both tVOR responsivity and aVOR gain were reduced compared with during binocular viewing. As a group, the subjects showed a significant drop of tVOR responsivity (P = 0.019, Wilcoxon signed rank test), but not aVOR responsivity, from binocular to monocular viewing of the target at 17 cm. However, also apparent in Fig. 2 is a decrease in vergence angle from 17.7° ± 2.8° during binocular viewing to 12.2° ± 7.8° during monocular viewing, which was significant (paired t test, P < 0.005). Accordingly, we compared six subjects’ responses when they viewed targets at 2 m, 40, or 17 cm either directly or with a base-out prism before their right eye; data are summarized in Fig. 3a. For each of the three target distances, tVOR responsivity values clustered in a discrete range, even though vergence angle varied according to whether viewing was direct or through a prism (Fig. 3a). Paired comparison (Wilcoxon signed rank test) of tVOR responsivity for each target distance of each subject, either directly or through a prism, showed no significant difference. This result suggested that the effects of monocular viewing on tVOR could not be simply attributed to change of vergence angle, but might reflect attenuation of visual cues.
Fig. 3

a Comparison of direct and prism viewing on tVOR responsivity. Data from six subjects are plotted, showing that for each of the three target distances, tVOR responsivity clustered in a discrete range, even though vergence angle varied according to viewing conditions. b Comparison of vertical smooth-tracking and tVOR during binocular viewing of a near target. In this polar plot, numbers on the circumference indicate phase shift and numbers on radii indicate gain for visual tracking and compensation gain for tVOR. Gain values were greater, and phase lags of eye velocity with respect to ideal eye velocity required to follow the target were smaller, during tVOR than during visual tracking, with no overlap of data (indicated by enclosing lines)

We studied this possibility further by measuring the effects of switching to darkness, and strobe illumination (to minimize retinal slip information) in two subjects as they attempted to view binocularly the near target. After switching to darkness, both subjects gave consistent and similar responses that are summarized in Fig. 2 (inverted black triangles); tVOR and aVOR both declined substantially. The same was the case during strobe illumination (Fig. 2, upright gray triangles). Thus, either switching to darkness, or minimizing retinal slip information by strobe illumination reduced aVOR and tVOR responses during attempted viewing of a near target, even though vergence angle remained greater than 18°.

Finally, we addressed the possibility that greater tVOR responsivity in ambient light was due simply to improved smooth visual tracking. Mean (±SD) of tracking gain for smooth pursuit for the full amplitude stimulus was 0.25 (±0.08) with phase lag of 58.6° (±15.6) for measured eye velocity with respect to ideal eye velocity required to follow the target; for the half amplitude stimulus, gain was 0.16 (±0.06), with a phase lag of 56.4° (±13.7). We compared these responses with tVOR behavior when responses were largest—during binocular viewing of the near target. Figure 3b is a polar plot comparing smooth tracking gain and tVOR compensation gain. Note how phase lags were smaller, and gain values larger, during tVOR than during visual tracking, with no overlap of data. Thus, smooth visual tracking could not account for tVOR during near viewing, indicating that other mechanisms must modulate tVOR during these visual test conditions.


We set out to determine how well tVOR responded to vertical head translations during combined bob–yaw movements under conditions of ambient illumination. Our test frequencies corresponded to those of head perturbations that occur during locomotion (Grossman et al. 1988; Pozzo et al. 1990; Moore et al. 1999). In 20 subjects spanning almost five decades in age, we found that tVOR generated eye movements that were only about 60% of those required to hold the line of sight on the target, whether it be located far (2 m) or near (17 cm). In contrast, aVOR generally compensated for head rotations and held the eyes on far and near targets. This led us to re-evaluate the purpose of tVOR, noting that geometry makes it impossible for tVOR to adequately stabilize images of both near and far objects within the visual scene during head translations. We found that the best responses occurred during binocular vision in ambient illumination, when natural visual cues, such as motion parallax, were available. Our findings raise a number of issues. First, how do our results compare with prior studies of tVOR? Second, what factors seem most important for adjusting tVOR responses? Third, what could account for the apparent inadequacy of tVOR to compensate for translational head perturbations?

Comparison with prior studies of tVOR

Prior studies have thoroughly documented how aVOR compensates for yaw head perturbations during viewing of distant and near targets (Viirre et al. 1986; Crane and Demer 1997; Han et al. 2001). However, to the best of our knowledge, only G. D. Paige has previously studied tVOR in bob in normal human subjects. He studied two subjects who bounced themselves up and down on a spring-suspended stool at ∼2.7 Hz with a peak excursion of 3.2 cm (Paige 1989), motion that mainly stimulates the sacculus (Fernandez and Goldberg 1976). Movements of only one eye were measured using the magnetic search coil technique (vergence angle was not monitored). The two subjects viewed Snellen optotypes at 424, 142, and 36 cm in dim illumination or in darkness with spot illumination of the optotype; thus viewing conditions were dissimilar from the ambient lighting that we employed. Similar to the present study, Paige found that eye rotation was almost 180° phase-shifted with respect to head displacement, i.e., compensatory in direction, and increased as target distance decreased, but fell short of what was required to maintain target fixation. When his two subjects viewed a head-fixed target at 36 cm, tVOR decreased by 56 and 80% of values with an earth-fixed target, and Paige concluded that “no visual following or motion detection mechanism exists which could have accounted for the major proportion of eye movement responses observed during vertical linear oscillations, even when visual inputs were available”.

Our present study essentially confirms Paige’s finding in a large number of subjects. In addition, by making binocular eye movement recordings, we were able to show (at least under our experimental conditions) that target distance was an important determinant of tVOR responses. Furthermore, by also testing smooth tracking of a large visual display moving at a frequency and amplitude corresponding to the visual demands imposed by platform motion, we confirm that visual tracking mechanisms contribute only modestly to the response to bob motion during near viewing. Thus, similar to visual modulation of aVOR (Huebner et al. 1992; Das et al. 1998), mechanisms other than superposition of visual tracking appear to contribute.

Other studies have measured human tVOR in response to transient (Gianna et al. 1997; Ramat et al. 2003, 2005), or sinusoidal (Paige et al. 1998), interaural motion. At 2 Hz, the interaural tVOR shows a positive slope with increasing vergence angle and a positive intercept (similar to Fig. 2). Studies employing transient stimuli have reported responses that are influenced by vergence, accommodation, visual cues such as motion parallax, as well as anticipation of whether the visual target will remain still or move.

Our study combined rotational and translational head movements. A prior study on rotation while translating concerned rabbits who translated across a rotating platform so that Coriolis acceleration was induced due to translation within the rotating frame (Maruta et al. 2005). Thus, this stimulus was unlike that employed in our study, in which head rotations occurred around the vertical axis in which head translated. Other studies of humans and squirrel monkeys have displaced the subject’s head eccentrically from the axis of rotation in order to induce simultaneous aVOR and tVOR (Bronstein and Gresty 1991; Anastasopoulos et al. 1996; Telford et al. 1998). When the subject’s head is positioned nose-up and one side of the head is directed outward, then a torsional aVOR and vertical tVOR are stimulated. In squirrel monkey, such orthogonal aVOR and tVOR components were similar to those measured during independent stimulation. In our subjects, both tVOR and aVOR responses showed small changes during combined rotation–translation versus pure yaw or bob. In the case of the aVOR, the decrease in responses during combined rotation–translation might have been due, at least in part, to the small increase in target distance at the extremes of vertical platform translation. On the other hand, the increase in tVOR responses during combined rotation–translation cannot be readily attributed to geometric factors and may represent another example of how vestibular responses are enhanced when coupled with other types of eye movements (Das et al. 1999). Enhancement of tVOR during combined canal-otolith stimulation induced by eccentric rotation has been previously reported (Anastasopoulos et al. 1996).

What mechanisms determine tVOR responses?

Our present results suggest that the brain’s estimate of target distance, based on multiple factors including motion parallax and vergence is the key factor that determines tVOR responses. We performed our experiments in ambient light and presented real near stimuli, which provided 3-D visual information as well as relative motion between the near stimulus and background during testing. During binocular viewing, our present responses, measured in terms of percentage of ocular rotations required to hold the eye on target, were as large as reported in any prior study. However, whenever this percept was eroded—during monocular viewing or during strobe illumination—tVOR responses declined (Fig. 2), even though some convergence was maintained. Furthermore, during binocular viewing with a prism placed before one eye to induce convergence, responses were similar to during direct viewing, indicating the relative importance of binocular information in setting tVOR responses, although convergence may also contribute.

Another important finding was that there was no significant change of the phase lag of tVOR during the far and near viewing conditions; mean values were ∼18° from that required for an ideal response. For our 2 Hz stimuli, this phase lag corresponds to a delay (latency to onset) of about 20 ms for tVOR, which agrees with reported values from studies that have employed transient stimuli (Ramat and Zee 2003). If visual tracking eye movements were supplementing otolith-ocular responses, substantial phase lags would be expected, and one implication of the constant phase of tVOR is that the response represents mainly vestibular drives. Thus, it seems that although natural visual cues are required to modulate tVOR and, specifically, to increase the magnitude of the response during viewing of near targets, visual tracking eye movements do not contribute substantially to this increase.

Possible role of tVOR during natural activities

There are differences between the head perturbations that occur during natural locomotion and those that we applied in these experiments (aside from our subjects not moving forward through their environment). For example, during walking, active and passive vertical head translations are accompanied by pitch rotations (Pozzo et al. 1990; Moore et al. 1999; Bloomberg et al. 1992), which we intentionally minimized in our experiments. It follows that hypotheses that invoke interactions between head rotations and translations to explain why tVOR does not adequately compensate for translations during near viewing (Ramat and Zee 2003) do not readily account our findings. If tVOR compensates for only about 60% of translational head perturbations during far or near viewing (Israël and Berthoz 1989; Paige 1989; Gianna et al. 1997; Ramat and Zee 2003; Moore et al. 1999), what are the potential visual consequences?

During our testing, subjects noted that the near stimulus appeared to bounce up and down (oscillopsia), probably due to excessive retinal image motion. Similar perceptions were reported in subjects during treadmill walking as they viewed a near target (Crane and Demer 1997; Moore et al. 1999). A simple experiment may convince the reader that tVOR fails to stabilize images of near objects. Place a visual acuity test card on a shelf at eye level at a target distance of ∼17 cm. First rotate the head in yaw at 1–2 Hz; visual acuity will remain about the same (aVOR). Second, bob up and down by bending the knees at a frequency of 1–2 Hz; visual acuity will deteriorate by several lines, and oscillopsia may result (tVOR). Thus, aVOR guarantees clear vision of near objects during head rotations, but tVOR does not guarantee clear vision during head translations.

If one takes the view that tVOR must compensate for head translations sufficiently to safeguard some aspect of vision then, since the target distance appears to be the main determinant of the response, at what distance are objects located for which tVOR will provide clear vision? Knowing the amplitude of head translations, it is possible to calculate the peak retinal image velocity that will occur as subjects view targets over a range of target distances (Schwarz and Miles 1991), and then scale this curve by a factor of (1-compensation gain). Figure 4 provides a comparison of such a curve with measured retinal image speeds in three subjects, whose mean compensation gain was 0.6. In general, the curve predicts the data well, with smaller values for peak retinal image speed for visual targets at 40 cm (for which their compensation gain values were greater). Also shown in Fig. 4 is a dashed line corresponding to 5°/s, which is required for clear vision of objects with higher spatial frequencies (Carpenter 1991; Demer and Amjadi 1993). It is evident that tVOR with a compensation gain of 0.6 holds peak retinal image speed below 5°/s for target distances greater than 90 cm and, even within arm’s reach (∼30–50 cm), peak image slip is <10°/s; a similar range for optimal operation of tVOR has been previously noted (Paige et al. 1998). Only for very close viewing does retinal image velocity increase to levels that degrade vision and cause oscillopsia. In Fig. 4, we also plot the peak retinal image speed of the background lying at 200 cm that, although defocused, might provide motion parallax information; its speed is similar to that of the image of the fixation target (although opposite in direction). Finally, we asked what motion of the background image would be expected if tVOR did perfectly compensate for head translations; it is evident in Fig. 4 that during near viewing, background image motion would be expected to exceed 50°/s. Thus, taken with evidence from prior studies (Paige et al. 1998), it seems possible that tVOR is set to optimize viewing of objects that fall between the distance of the target being viewed and the background (Miles 1998). More specifically, minimizing the velocity of retinal image motion of both near and distance objects might aid detection of motion parallax signals that are important for detecting relative distances of objects in the environment. Thus, the human tVOR may have evolved to maintain compensation gain at a value of ∼0.6 not to keep the eyes on target, but rather to optimize motion parallax estimates of the locations of objects in the path of locomotion. Further experiments, for example, moving the visual background with respect to the stationary near visual target, could be conducted to test this hypothesis. This reinterpretation of the purpose of tVOR could also be extended to patients with abnormal vestibular responses. (Liao et al. 2007a).
Fig. 4

Comparison of geometric prediction of peak retinal image speed (RIS) as a function of target distance for three subjects versus their measured peak retinal image speeds. The mean bob head displacement in these three subjects was ±1.5 cm. Their mean compensation gain was 0.6, and the curve defined by the equation shown was accordingly scaled by a factor of 0.4. Squares indicate measured values of RIS of the fixation target at each of the target distances for each subject for either pure bob or combined bob–yaw. There is generally good agreement, except that during viewing targets at 40 cm, peak RIS is lower than predicted due to greater compensation gain values. Circles indicate RIS of the background at 200 cm; although the direction of background image motion is opposite to that of the target image motion, the magnitude is similar. Inverted triangles indicate calculated RIS values of the background if tVOR compensation gain = 1.0; this is substantially increased during near viewing. The dashed horizontal line corresponds to a retinal image speed of 5°, above which visual acuity for high spatial frequencies will decline


This research was supported by NASA/NSBRI NA00208, Office of Research and Development, Medical Research Service, Department of Veterans Affairs, NIH grant EY06717, and the Evenor Armington Fund.We are grateful to the subjects who volunteered to serve in this study and to Drs. Harold Bedell, John Stahl, David Zee, Gary Paige, Matthew Thurtell, Louis Dell’Osso, Robert Kirsch, and Miklos Gratzl for their helpful advice. The work reported in this paper constitutes research performed by Ke Liao as part of the requirements for his Doctoral Dissertation.

Supplementary material

221_2007_1181_MOESM1_ESM.doc (42 kb)
Appendix A (DOC 42.0 kb)

Copyright information

© Springer-Verlag 2007