II. VOR and perceptual responses during combined Tilt&Translation. To compare and contrast the neural mechanisms that contribute to vestibular perception and action, we measured vestibuloocular reflexes (VOR) and perceptions of tilt and translation. We took advantage of the well-known ambiguity that the otolith organs respond to both linear acceleration and tilt with respect to gravity and investigated the mechanisms by which this ambiguity is resolved. A new motion paradigm that combined roll tilt with inter-aural translation (“Tilt&Translation”) was used; subjects were sinusoidally (0.8 Hz) roll tilted but with their ears above or below the rotation axis. This paradigm provided sinusoidal roll canal cues that were the same across trials while providing otolith cues that varied linearly with ear position relative to the earth-horizontal rotation axis. We found that perceived tilt and translation depended on canal cues, with substantial roll tilt and inter-aural translation perceptions reported even when the otolith organs measured no inter-aural force. These findings match internal model predictions that rotational cues from the canals influence the neural processing of otolith cues. We also found horizontal translational VORs that varied linearly with radius; a minimal response was measured when the otolith organs transduced little or no inter-aural force. Hence, the horizontal translational VOR was dependent on otolith cues but independent of canal cues. These findings match predictions that translational VORs are elicited by simple filtering of otolith signals. We conclude that internal models govern human perception of tilt and translation at 0.8 Hz and that high-pass filtering governs the human translational VOR at this same frequency.
In the study presented in the companion paper (Merfeld et al. 2005), we simultaneously recorded vestibuloocular reflexes (VORs), perceived tilt, and perceived translation in human subjects to determine what neural mechanisms humans use to resolve the ambiguous otolith transduction of gravito-inertial force (GIF). We found that internal models contributed to perceived tilt and perceived translation and also found that simple high-pass filtering of otolith signals contributed to the human translational VOR at frequencies >0.2 Hz. The finding that the human translational VOR included a substantial component elicited by simple high-pass filtering surprised us for several reasons. First, these findings seemed to contradict previously published findings from the monkey using similar motion paradigms (Angelaki et al. 1999). Moreover, at first glance, these findings even appeared to contradict some of our own earlier reports demonstrating contributions of sensory fusion via internal models to the VOR in both humans (Merfeld et al. 1999, 2001; Peterka et al. 2004; Wall et al. 1999; Zupan and Merfeld 2003; Zupan et al. 2000) and monkeys (Angelaki et al. 2001; Merfeld and Young 1995). Finally, we were surprised by the fact that the perceptual responses were governed by qualitatively (not just quantitatively) different mechanisms than the VOR.
Given these unexpected findings, we designed and utilized a novel motion paradigm that we will refer to as the Tilt& Translation paradigm because this paradigm combined simultaneous tilt and translation of the head. In brief, we roll tilted the subjects about an earth-horizontal axis using the same tilt device described in the earlier paper, but instead of always keeping the ears near the center of rotation, we varied the location of the ears with respect to the center of rotation (Fig. 1). With proper choice of parameters (ear location and frequency), the tangential acceleration, which was always aligned with the inter-aural axis, could null the inter-aural gravitational cue transduced by the otolith organs such that little or no inter-aural otolith cue was provided even though the head was both tilting and translating (Fig. 2). An important feature of this new motion paradigm is that the roll rotation measured by the canals was maintained the same across trials, while the inter-aural otolith signal varied linearly with the placement of the head with respect to the center of rotation. As in the companion study, we simultaneously recorded the VOR, perceived tilt, and perceived translation so that we could directly compare and contrast measures of “perception” (perceived roll tilt and perceived inter-aural translation) and “action” (VOR). To test a published model (Merfeld and Zupan 2002), we include modeling predictions simulated prior to performance of the experimental investigations.
Methodologies were nearly identical to those used for the companion study (Merfeld et al. 2005) except that a novel motion paradigm, which combined tilt and translation, was used. A full description of the new motion paradigm follows. As for the companion study, we used proximal vergence (Jaschinski-Kruza 1990; Schor et al. 1992; Wick and Bedell 1989) to maintain vergence. The effectiveness of this proximal vergence technique to help maintain vergence is demonstrated in Fig. 7 of the companion paper (Merfeld et al. 2005). Detailed descriptions of other methodologies were included in the companion paper (Merfeld et al. 2005).
Tilt&Translation motion paradigm
The subjects sat in the same device as for the Tilt paradigm described in the companion paper (Merfeld et al. 2005), and the same head and body restraints were used. This device passively tilted upright subjects in roll (toward left-ear-down or right-ear-down) by rotating about an earth-horizontal axis. Initially, the subject was upright, and the head was placed such that the subject's ear level was at the same level as the earth-horizontal rotation axis. The only difference between this Tilt&Translation paradigm and the earlier Tilt paradigm was that the subject's head was displaced from the center of rotation for most trials (Fig. 1). The subject's ear level (R) was varied between 20 cm above and 40–50 cm below the center of rotation at 10 cm intervals with the lowest testing position dependent on the subject's height. Changes in radial position always occurred between trials; the radius never varied during a trial.
For these Tilt&Translation trials, the frequency of the sinusoidal roll-tilt motion was 0.8 Hz, a frequency at which the canals accurately sense angular velocity, and the maximum tilt amplitude was 5, 10, or 20°. Acceleration discontinuities were eliminated by linearly increasing the sinusoidal velocity over three cycles. Twenty steady-state cycles of stimulation were provided. The steady-state sinusoidal roll stimulus was left/right symmetric about the upright position. Trial order was randomized for each subject.
For this paradigm, the steady-state, inter-aural, gravitational force measured by the otolith organs can be represented mathematically as gy(t) = −G sin [Θmax sin (2πft)] ≈ −G sin (Θmax) sin (2πft), where G is 9.8 m/s2, Θmax is the maximum tilt angle (5°, 10°, or 20°), and f is the stimulation frequency (0.8 Hz). In addition, a tangential acceleration, which varies with radius and is always aligned with the inter-aural axis, is present, ay(t) = −Rθ̈ = RΘmax(2πf)2 sin (2πft), where R is the radial distance between the rotation axis and ear level. Therefore the total, specific inter-aural (y axis) GIF is fy(t) = gy(t) − ay(t) ≈ −[G sin (Θmax) + RΘmax(2πf)2] sin (2πft). Specifically, sinusoidal roll tilts yield sinusoidal variations in both inter-aural gravitational force and tangential acceleration, which combine to yield the inter-aural GIF measured by the otolith organs. The inertial force due to tangential acceleration augments the inter-aural component of gravity when the ears are located above the center of rotation and decrements from the inter-aural component of gravity when the ears are located below the center of rotation (Fig. 2, A and B). With a stimulus frequency of 0.8 Hz, gravity and linear acceleration nearly cancel one another at a radial position 38 cm below the center of rotation [i.e., when R = −G sin (Θmax)/Θmax(2πf)2, Fig. 2, A and B]. As the radius continues to increase downward, the inter-aural GIF component begins to grow but the phase reverses (Fig. 2C).
For the fore-described Tilt&Translation paradigm, the inter-aural otolith cues varied proportional to radius, while identical dynamic roll rotation cues were maintained across trials.1 This complements the Translation and Tilt paradigms used in the companion paper, where the otolith cues were matched across conditions, but the dynamic roll rotation cues varied across conditions (roll rotation cues were present during Tilt but not present during Translation).
Subjects and instructions
Nine healthy subjects were prescreened as “normal” via several, standard, clinical, vestibular tests. All subjects signed an informed consent, consistent with institutional procedures prior to participation. The subjects (8 males and 1 female) were between 22 and 60 yr of age. Eight of the nine subjects were also subjects for the companion study (Merfeld et al. 2005).
The inter-aural GIF measured by the otolith organs during the Tilt&Translation paradigm is shown in Fig. 2 as a function of radius. The sinusoidal variations in the inter-aural GIF are minimal when the ears are located 38 cm below the rotation axis, because the tangential acceleration nulls the inter-aural gravitational cue at this radius.
Frequency segregation predictions—simple filter model
Predicted responses of simple filters during the Tilt&Translation paradigm at 0.8 Hz are shown as a function of radius (Fig. 3, A and B). Both the low-pass (flp = 0.07 Hz) and high-pass (fhp = 0.07 Hz) filters have outputs that scale linearly with the otolith cues (Fig. 3). Low-pass filters always pass a small component at high frequencies, and high-pass filters always pass a small component at low frequencies; the cut-off frequency of the filters affects the magnitude of the filter output but does not affect the characteristic V shape of the filter output; the minimal response of a simple filter will always occur when the filter input (inter-aural force, Fig. 2) is minimal.
Both high- and low-pass filters predict that the amplitude of the accompanying response (translation and tilt, respectively) will scale linearly with radius, with minimal amplitude when the ears are 38 cm below the rotation axis (i.e., when no inter-aural force is measured by the otolith organs). While not shown, phase of both high- and low-pass filter outputs demonstrates an abrupt 180° phase shift at a radius of –38 cm. This occurs because of the 180° phase shift in the GIF input to the filters at this radius (Fig. 2C).
Internal model predictions—sensory fusion model
Predicted responses of the model during combined Tilt&Translation at 0.8 Hz are shown as a function of radius in Fig. 3, C and D.2 The model predicts (Fig. 3C) that perceived roll tilt will be independent of radius (i.e., constant) and that the translation response amplitude (Fig. 3D) will be proportional to radius with a “minimum” when the ears are ∼10 cm below the rotation axis (i.e., V-shaped, with a minimum near −10 cm).
The predicted inter-aural estimate of gravity is maintained constant by the hypothesized neural pathways that use rotational cues from the canals to help predict the relative orientation of gravity during rotation. These are the modeled neurons and neural pathways shown with thick black lines in the appendix of the companion paper (Merfeld et al. 2005). As a direct effect, the model predicts that tilt perception will be estimated more or less accurately at frequencies where the canal receptors respond, even during the Tilt&Translation paradigm where the inter-aural otolith cues vary with head position. If the nervous system correctly estimates that an otolith cue is due to a change in orientation with respect to gravity, the nervous system can then determine that little or no linear acceleration is present by simply taking a difference between the otolith measure of GIF and the estimate of gravity. [These linear acceleration calculations are performed by the neural pathways represented by thick gray lines in the appendix of the companion paper (Merfeld et al. 2005).]
To allow direct comparisons across data sets, eye-movement data were analyzed for the same five subjects presented in the companion paper (Merfeld et al. 2005). During Tilt&Translation at 0.8 Hz, the horizontal VOR during 20° tilts was about twice as large as for 10° tilts, which was, in turn, about twice as large as for 5° tilts (Fig. 4A). A substantial horizontal VOR existed when the ears were at the same level as the rotation axis (0 cm). The horizontal response amplitude increased linearly with radius as the head moved upward relative to the center of rotation (radius: >0 cm). The horizontal response amplitude decreased as the head moved downward relative to the rotation axis. The horizontal eye response was smallest when the ears were ∼40–50 cm below the center of rotation (Fig. 4A).
For most trials, the phase of the horizontal response showed an average lag of ∼15° relative to the roll angular velocity where 0° would be perfectly compensatory. At around –40 to –50 cm, the phase showed an abrupt transition for two of three peak angular displacements (5 and 10°). A phase shift of ∼180° would be expected for a simple filter since the inter-aural otolith cue reverses at this point. The amount of the observed phase shift appears just short of this theoretical expectation, but the transition was not complete because physical device limitations prohibited us from testing at radii beyond –50 cm. Because the magnitude of the inter-aural otolith cue goes to zero and there is evidence of a phase reversal at around –40 cm, these results show that simple filtering (i.e., Fig. 3B) contributes to the human translational VOR at 0.8 Hz.
The amplitude (Fig. 4C) of the torsional VOR, which is primarily an angular VOR elicited in response to the roll stimulus velocity, was more or less constant for a given peak tilt amplitude. The slow phase velocity of the torsional VOR for the 20° peak amplitude (48.7°/s) was a little less than twice that for the 10° peak amplitude (30.0°/s), which was, in turn, about twice that for the 5° peak amplitude (15.2°/s). This was expected because the stimulus angular velocity was independent of radius and doubled with each doubling of the tilt amplitude (5–10°, 10–20°). The phase of the torsional VOR (not shown) was always about –5°, roughly compensatory for the rotational roll motion. Vertical responses (not shown) were small.
In response to Tilt&Translation stimulation, the perception of head roll tilt (Fig. 5A) was roughly constant and veridical with the actual tilt. The mean tilt responses were 5.4, 9.8, and 18.8°, which were not substantially different from the actual peak tilt angles of 5, 10, and 20°, respectively. The slopes of the responses as a function of ear location were not significantly different from zero. Thus the perception of roll tilt was not dependent on changes in the inter-aural otolith cues (Fig. 2). Particularly interesting was the finding that the amount of head roll tilt was accurately reported even when no substantial inter-aural otolith cues were available (i.e., with the ears placed ∼40 cm below the axis of rotation). These findings are consistent with internal model predictions (Fig. 3C) but are not consistent with simple filtering predictions (Fig. 3A).
In response to the Tilt&Translation stimulation, the perception of head translation was consistent with the actual head translation (Fig. 5B). The reported translation perception was minimal when the ears were at the level of the rotation axis. When the head was placed downward or upward relative to the rotation axis, the amount of reported head translation increased. The magnitude of the translational response was proportional to the tilt angle (5, 10, or 20°) and linearly proportional to the magnitude of ear location (radius) relative to the rotation axis. This is consistent with the physics because both the tangential acceleration and the associated inter-aural translation are proportional to both the radius of rotation and the tilt angle. As for tilt perception, the reported perception of inter-aural translation was not proportional to the inter-aural otolith cue (Fig. 2B). Thus the data do not provide any suggestion that the perception of translation varies with the otolith stimulation as it would if there were a high-pass filtered response component (Fig. 3B) at 0.8 Hz. These perceptual reports of inter-aural translation are more consistent with internal model predictions (Fig. 3D) than simple filtering predictions (Fig. 3B).
Simple filtering contributes to human translational VOR
We found a horizontal VOR during Tilt&Translation (Fig. 4A) that mimics the amplitude of the total inter-aural GIF at the level of the ears (shown in Fig. 2). This horizontal VOR does not match true inter-aural linear acceleration of the ears, which varied linearly with radial displacement. These results are consistent with simple filtering of otolith signals (Fig. 3B) and inconsistent with sensory fusion (Fig. 3D).
Because earlier studies have shown that graviceptors other than the otolith organs contribute to data obtained using a body posture matching task (e.g., Mittelstaedt 1996), it is worth noting that our VOR data do not provide any evidence that these alternate graviceptors contribute to translational VOR responses via simple filtering. This is not surprising because the VOR helps compensate for head motion not motion of the trunk.
The finding that the human translational VOR is governed by simple filtering does not appear consistent with reports showing that monkey translational VORs are primarily governed by internal models (e.g., Angelaki et al. 1999; Green and Angelaki 2003; Merfeld and Young 1995). However, a number of differences between how humans and non-human primates process gravito-inertial cues have previously been noted. For example, humans and monkeys show large differences in their postrotational tilt responses. Monkey eye movements change so that the axis of eye rotation rapidly aligns with gravity (Angelaki and Hess 1994; Merfeld et al. 1993b), but human responses show little or no shift in the axis of eye rotation after postrotational tilt (“dumping”) (Fetter et al. 1992; Zupan et al. 2000). Responses to centrifugation show similar differences. Both rhesus monkeys (Wearne et al. 1999) and squirrel monkeys (Merfeld and Young 1995) show rapid shifts in the axis of eye rotation such that the axis of eye rotation aligns with the resultant GIF, with little (squirrel monkey) or no (macaques) difference in the horizontal VOR in the facing-motion and back-to-motion orientations. Contrary to these monkey responses, humans have only small shifts in the axis of eye rotation but show large differences in the horizontal responses in the facing-motion and back-to-motion orientations (Merfeld et al. 2001; Wearne 1993). Related differences are also apparent in eye responses during off-vertical axis rotations (OVAR). Monkeys demonstrate a large “bias” component and a small sinusoidal response at the frequency of rotation (e.g., Goldberg and Fernandez 1982). Human responses show a small bias and a large sinusoidal component (e.g., Benson and Bodin 1966; Correia and Guedry 1966). Are these documented differences due to an inherent species difference, to a difference in the motion experience of the humans and non-humans, or to some other “lifestyle” difference between the human and non-human primates? Our data do not allow us to answer this question.
It is also possible that a subtle difference in the motion paradigms could explain the difference between our horizontal eye response findings and those reported previously (Angelaki et al. 1999). However, as shown in Fig. 6, the forces recorded by the otolith organs in the two conditions are quite similar.3 Although some small quantitative differences are shown in the figure, the inter-aural forces are nearly zero for both paradigms, demonstrating tiny force components both at the tilt frequency and twice the tilt frequency. The dorsoventral (z axis) forces for the two motion paradigms are qualitatively similar; both include a constant value along with a modulation at twice the tilt frequency.
It is worth noting that neither simple filtering nor the existing canal-otolith interaction models (Angelaki et al. 1999; Droulez and Darlot 1989; Glasauer 1992; Glasauer and Merfeld 1997; Green and Angelaki 2003; Merfeld 1995b; Merfeld and Zupan 2002; Zupan et al. 2002) predict that such subtle quantitative differences in the stimuli should yield qualitatively different VOR response patterns.
Simple filtering does not contribute to tilt perception
The magnitude of reported tilt perception was constant, independent of radius, and veridical for all three, peak, tilt angles. Because the inter-aural otolith cue varied with radius, the absence of any variation dependent on radius demonstrates that simple filters do not contribute to tilt perception (or, at best, contribute in some small manner not observable in our data). As one specific example, subjects reported roll tilt perceptions that were veridical to the motion even when their ears were placed 40 cm below the rotation axis, even though this radial location resulted in little or no inter-aural otolith cue. As another specific example, when a subject’s ears were placed above the rotation axis and the inter-aural otolith cue increased, there was no apparent effect on tilt perception. Therefore these tilt data provide no indication or hint that simple filtering mechanisms contribute, even minimally, to human roll tilt perception. We do not think that the z-axis forces (Fig. 6) have a substantial influence on these roll tilt responses, but if they do have such an influence, this would remain inconsistent with the simple filtering model because use of the z-axis force cues by the brain clearly involves more than simple filtering.
As shown in Fig. 3C, the sensory fusion model predicted that tilt would be underestimated during Tilt&Translation at 0.8 Hz. As shown in the companion paper (Fig. 6A, Merfeld et al. 2005), a similar underestimation of predicted tilt was achieved at all frequencies above ∼0.05 Hz. It seems reasonable to consider that the nervous system might scale-up the tilt estimate to yield veridical tilt perception at physiologic frequencies >0.05 Hz. Such increased scaling would result in tilt being overestimated at low frequencies. This is exactly what was reported in the companion paper (Fig. 6A, Merfeld et al. 2005). Such an overestimation at low frequencies would not seem to pose a substantial behavioral problem because visual orientation cues are generally available at these lower frequencies. A similar fusion of low-frequency visual and mid- to high-frequency canal influences has previously been reported in the vestibular nuclei (Henn et al. 1974).
Simple filtering does not contribute to translation perception
The magnitude of the actual peak-to-peak translational nose displacement experienced during the Tilt&Translation paradigm is At a radius of 20 cm, the maximal upward displacement, this means that the actual peak-to-peak translation was 4, 7, and 14 cm, respectively, for 5, 10, and 20° tilts, which matches the mean reported translation (Fig. 5B) of 5.7, 8.9, and 13.0 cm. At a radius of –40 cm, the minimal radius tested for all subjects, the actual peak-to-peak translation was 7, 14, and 28 cm, respectively. Again this matches the mean reported translation (Fig. 5B) fairly well, though the reported translation (10.6, 20.0, and 33.7, respectively) somewhat exceeded the actual translation.
Minimal translation perception was reported when the ear-level was at (or very near) the center of rotation. Therefore subjects reported veridical (or nearly veridical) inter-aural translation during this combined Tilt&Translation paradigm. Again, this is not consistent with the frequency segregation hypothesis, which predicts that inter-aural translation perception should be elicited by high-pass filtering of inter-aural otolith cues. With the ear-level at the rotation axis (0 cm), there is a substantial inter-aural otolith cue due to tilt with respect to gravity. Yet, little or no translation perception was reported. At the same time, when ear level was 40 cm downward from the rotation axis, the inter-aural otolith cues were negligible. Yet substantial translation perception was reported. Consequently, these translation data provide no indication that simple filtering mechanisms contribute, even minimally, to human translation perception.
The findings demonstrate that simple filtering does not contribute substantially to human perception of tilt or translation. The measured eye responses show that simple filtering governs the human translational VOR at 0.8 Hz, and the perceptual data show that internal models govern perceptions of roll tilt and inter-aural translation at this same frequency. These findings lead to the conclusion that qualitatively different mechanisms contribute to human translational VOR and perceptual responses. Because the neural substrate for the VOR is fairly well understood, this difference between perception and action emphasizes the importance of developing a better understanding of the neural substrates underlying perception.
The authors acknowledge support from National Institute on Deafness and Other Communications Disorders Grants DC-004158 to D. M. Merfeld and S. Park and DC-00205 to C. Gianna-Poulin, S. Wood, and F. O. Black as well as National Aeronautic and Space Administration Grants NNJ04HB01G to D. M. Merfeld and NAW9-1254 to F. O. Black.
The authors thank T. Bennett and V. Stallings for technical contributions, M. Marsden and P. Cunningham for administrative assistance, and Drs. Lionel Zupan and Rick Lewis for commenting on early versions of the manuscript. Experiments were performed at the Legacy Neurotology Research Laboratory.
↵1 Secondary “practical” benefits of this new motion paradigm are that it is relatively inexpensive (because only a single dynamic motion actuator is needed) and simple (because synchronization of 2 or more actuators is not required).
↵2 For these 0.8 Hz data, we only show the model predictions obtained from the exact model, including model parameters, published earlier (Merfeld and Zupan 2002). We do not show simulated results using the revised parameters described in the companion paper because the simulations obtained from the revised model are barely distinguishable from the modeling predictions shown herein in Fig. 3, C and D.
↵3 Figure 6 shows the forces for a 20° Tilt&Translation at 0.8 Hz with the ears displaced 38 cm downward, which reduces the peak inter-aural force from 3.35 m/s2 (Fig. 6A), which is the peak inter-aural gravitational force with the head-centered, to 0.03 m/s2 (Fig. 6E). This is below the human threshold for detection of the direction of translational motion (e.g., Benson et al. 1986; Melvill Jones and Young 1978). The dorsoventral (z axis) force, due to both centrifugal force and gravity, had a mean of −10.09 m/s2 (Fig. 6F), and the peak deviations from the mean were 0.88 m/s2 (Fig. 6F). For the paradigm used by Angelaki et al., the forces using a tilt device mounted on a translation device at 0.8 Hz are shown. For a 20° tilt, nulling of the peak inter-aural force is achieved with a sinusoidal translation having a peak acceleration of 3.57 m/s2 (Fig. 6C), yielding inter-aural, peak, force deviations of ∼0.06 m/s2 (Fig. 6E), which is near the lower limit of human perception. The mean z-axis force, due to both the applied horizontal linear acceleration while tilted and gravity, was −8.89 m/s2 (Fig. 6F), and the peak deviations from the mean were 0.91 m/s2 (Fig. 6F).
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2005 by the American Physiological Society