Chapter 5

A stationary observer exposed to a rotating scene experiences self-rotation, or circularvection in the opposite direction and perceives the scene as stationary (Fischer and Kornmüller 1930). It generally takes 20-30 s for vection to reach its full strength, which is ascribed to the absence of the vestibular inputs that would accompany actual self rotation during this initial period (Dichgans and Brandt 1978). Because the semicircular canals do not respond to steady rotation of the body, the relative motion of the visual surroundings is the only cue to steady passive rotation about a vertical axis. Therefore, during circularvection about a vertical axis there is a visual-vestibular conflict only at the onset of the visual motion. Accordingly, the latency of circularvection can be shortened by an impulsive rotation of the body in the same direction (Brandt et al. 1974).

Circularvection about a horizontal axis is more complicated, because it affects the perceived orientation of the body with respect to gravity. This not only involves inputs from the semicircular canals, but also from the otolith organs. The otoliths register the steady state orientation of the head relative to gravity, and thus contradict visually induced sensations of self-tilt in a stationary observer. As a consequence, rotation of the visual surroundings about a horizontal axis typically produces a sensation of continuous self-rotation accompanied by constant illusory self-tilt with a mean value of about 20º (Dichgans et al. 1972; Held et al. 1975). Moreover, the magnitude of circularvection has been found to be smaller for rotations about a horizontal axis than for rotations about a vertical axis (Held et al. 1975; Howard et al. 1988). Because of the persisting visual-otolith conflict, it is questionable whether passive body rotation at the onset of the visual motion can enhance visually-induced sensations of self-rotation and self-tilt about a horizontal axis, as it does for circularvection about a vertical axis. In the present study we addressed this question for rotation about the line of sight (roll axis) in an upright observer.

In flight simulators with a motion base, vestibular onset cues are applied to enhance sensations of self-motion. A motion base has limited travel and must be returned to its original position after a maneuver without disturbing the sensation of self-motion. This is done by an acceleration below the vestibular threshold, a procedure known as washout. We examined the effects of washout temporally separated from the onset by delaying the restoration of the body position. The effective vestibular stimulus to body roll can be decomposed into a rotation component and a tilt component. It was hypothesized that the magnitude of circularvection primarily depends on initial body acceleration in the washout maneuver, whereas the extent of perceived self-tilt primarily depends on the angle of body tilt. To investigate this hypothesis we exposed subjects to whole-field visual rotation about the horizontal roll axis and superimposed two acceleration levels of body roll, one above the vestibular threshold and one in the threshold region.

The subject sat inside a 9-foot diameter sphere lined with black dots on a white surface. The seat was padded with "slow recovery" plastic foam and the subject was constrained by a five-point harness and head rest with the head upright and close to the center of the sphere. The sphere was illuminated by a 60 watt light bulb placed above and behind the subject's head. The sphere and chair could be rotated independently about the same earth-horizontal axis. The rotation axis passed through the center of the sphere and was parallel to the subject’s roll axis centered on the subject’s head. There were seven conditions (Table 5.1). The effect of visual motion itself was investigated in one condition in which the sphere alone rotated ("visual motion"). The effect of body motion was investigated in two conditions, in which the subject alone rotated ("subject motion"). The effect of washout was investigated in three "washout conditions", in which both the sphere and the subject rotated. In two washout conditions the washout was delayed for 40s ("delayed-washout"), and in the third condition the washout immediately followed the onset ("immediate-washout"). Finally, the effect of body tilt without washout was investigated in a condition in which the subject was tilted before the sphere was accelerated, and restored to upright after deceleration of the sphere ("pre-tilt"). The sphere accelerated at 3º/s² until it reached a terminal velocity of 18º/s which was maintained for 80 s before deceleration. The subject was accelerated at either 0.7º/s² ("subthreshold") or 3º/s² ("suprathreshold"). We selected an acceleration of 0.7º/s² because it is in the threshold region for detection of self rotation about the roll axis and takes over 1 s to detect. An acceleration of 3º/s² is detected in under 0.1 s (Gundry 1978).

Table 5.1. Acceleration profile of chair and sphere in onset-phase and washout-phase in the concordant motion conditions. Chair tilt was achieved by constant acceleration up to half the final tilt angle of 15º, followed by constant deceleration of equal magnitude.

The subject was accelerated and then decelerated to a tilt angle of 15º. This is designated the onset phase of the washout trial. In the onset phase the sphere accelerated in the opposite direction so that the acceleration of the sphere relative to the subject was always 3º/s². In the delayed-washout conditions, the body tilt of 15º was maintained for 40s while the sphere rotated at a constant velocity of 18º/s. This is the post-onset phase of the trial. After 40s the subject returned to the vertical with the same acceleration/deceleration sequence as in the onset phase. Simultaneously, the sphere was accelerated/decelerated to keep its velocity relative to the subject constant. This is the washout phase of a trial. After the subject came to a stop in the upright position, the sphere continued to rotate at 18º/s for a further 40s. This is the post-washout phase of the trial. In the immediate-washout condition, the suprathreshold onset phase was immediately followed by a subthreshold washout phase, so that the subject was upright for the remaining 80s of sphere rotation at 18º/s. This condition most closely resembles washout sequences used in flight simulators. Finally, in the "pre-tilt" condition the subject was tilted to 15º about the roll axis with the sphere stationary. After 20s the sphere motion was initiated and maintained at 18º/s for 40s and stopped while the subject remained tilted.

In the above washout conditions, the sphere and the subject rotated in opposite directions. This was designated "concordant motion", because both motions produced sensations of self-motion in the same direction. We added two "discordant motion" conditions, in which the sphere and the body moved in the same direction. To keep the relative acceleration between the subject and the sphere comparable to that in the concordant conditions, the sphere accelerated faster than the chair, which created self-tilt in the opposite direction to the subject motion. In one discordant condition the body was tilted at subthreshold acceleration, and in the other at suprathreshold acceleration.

The velocity profiles of the subject and sphere for the concordant conditions are set out in Figure 5.1. The corresponding position profiles of the subject are shown in Figure 5.2. All conditions were performed with the light on. Each condition was run twice, once for clockwise and once for anti-clockwise motion of the sphere. The order of conditions was counterbalanced over subjects.

Responses

Subjects were asked to report verbally the magnitude of perceived body tilt and vection. The reports were taken 20 s after each acceleration-deceleration onset phase or restoration phase during the time of constant body posture and constant sphere rotation. It was assumed that the post-acceleration effects of the semicircular canals had faded after 20 s. The angle of self-tilt was expressed in degrees by all but two subjects, who preferred to use a scale based on half-hour

Figure 5.1. Velocity profiles of sphere (thin line) and chair (bold line) for concordant conditions. To indicate relative velocity, the same sign is used for sphere and chair velocity, although they rotated in opposite directions. The supra- and subthreshold subject-motion conditions are not shown, but the velocity profiles of the chair were the same as in the supra- and subthreshold delayed-washout conditions, respectively. The arrows indicate when the subject reported perceived self-tilt and vection. In all but the pre-tilt condition, there were two test phases.

Figure 5.2. Position profiles of the chair. In the supra- and subthreshold conditions, the chair was tilted to 15º and kept in that position throughout the 40 s post-onset phase, before the upright position was restored. In the immediate-washout condition the chair was tilted at a suprathreshold acceleration, and immediately tilted back to upright at a subthreshold acceleration.

intervals of the hour hand of a clock. Vection was expressed by a five-point scale as shown in Table 5.2. Subjects were also asked to keep a tactile rod aligned with the perceived vertical. The rod was 20 cm long and placed to the side of the subject’s right knee. The rod was not quite out of sight but subjects were asked not to pay visual attention to it. The first movement of the rod indicated the latency of illusory self-tilt at the beginning of each trial and the magnitude of rod tilt provided a measure of the perceived vertical which could be related to the verbal estimates.

Table 5.2. The five-point vection scale

Twelve subjects participated in the experiment, seven men and five women between ages of 24 and 60. One subject had to withdraw because of developing nausea. The data from another subject were excluded from the analysis because she experienced full self-rotation in all conditions, probably because she confused vection with illusory self-tilt. A third subject felt supine in about half of the conditions, and the perceived tilt data were entered as missing. Her vection and latency data were used.

Latency

The mean values and standard errors of the 10 subjects are shown in Figure 5.3 for the latency of perceived tilt, as indicated by settings of the tactile rod. The latency in the visual-motion condition was significantly longer than in the washout conditions (within subjects design ANOVA; F=9.91; df=3,6; p<0.01). A post-hoc Tukey test revealed a difference between the visual-motion condition and suprathreshold delayed-washout condition and between the visual-motion condition and the immediate-washout condition (p<0.05), but not between the visual-motion and the subthreshold delayed-washout condition. There was no difference in latency between the visual motion and the pre-tilt condition. The latency was longer in the subthreshold subject-motion and washout conditions than in the suprathreshold subject-motion and washout conditions (F=47.13; df=1,4; p<0.01). There was no difference between the suprathreshold delayed-washout condition and the immediate-washout condition.

There was no main effect of acceleration level of the subject on the angle of perceived self-tilt in the post-onset phase of both subject-motion and delayed-washout conditions. The magnitude of perceived self-tilt was greater in a tilted body position than in an upright body position, according to a significant difference between the delayed-washout conditions and the visual-motion condition (tactile rod: F=6.37; df=2,18; p<0.01; verbal estimate: F=7.22; df=2,18; p<0.01). In the pre-tilt condition the angle of perceived tilt as indicated by the tactile rod (but not the verbal estimate) was significantly greater than in the visual-motion condition (F=20.15; df=1,9; p<0.01) and also greater than in the delayed-washout conditions (F=4.47; df=2,18; p<0.05). However, in neither the delayed-washout conditions nor in the pre-tilt condition, was perceived self-tilt greater than the sum of the visual-motion condition and the suprathreshold or subthreshold subject-motion condition, respectively. Perceived tilt in the immediate-washout condition was not different from that in the visual-motion condition, indicating that the effect of actual body tilt disappeared after washout.

Figure 5.3. Mean latency of perceived self-tilt in the concordant-motion conditions. The bars indicate standard errors of the means (n=10).

Perceived self-tilt

The results for the magnitude of perceived self-tilt, as indicated by settings of the tactile rod and by the verbal estimate, are shown in Figs. 5.4a and b, respectively. Verbal estimates were consistently higher than estimates derived from settings of the tactile rod. For instance, in the post-onset phase of the subject-motion conditions the mean perceived tilt indicated by the settings of the tactile rod was 10.3º, while the verbal estimate amounted to 21.9º (F=15.12; df=1,9; p<0.01). In general, the two measures varied in a similar way with experimental conditions.

In both subject-motion conditions, restoration of the subject to vertical resulted in a significant decrease in perceived self-tilt to about zero (tactile rod: F=17.44; df=1,9; p<0.01; verbal: F=27.96; df=1,9; p<0.001). In the delayed-washout conditions, the perceived tilt was smaller in the post-washout phase than in the post-onset phase for both verbal (F=20.0; df=1,8; p<0.01) and tactile judgments (F=6.14; df=1,9; p<0.05), and was no longer different from the judgments in the visual-motion condition. There was no effect of acceleration level on the perceived tilt in the post-washout phase.

Magnitude of vection

The mean vection magnitudes are shown in Figure 5.5. In the post-onset phase there were no differences between the various conditions. On average, vection magnitude was about 2.5, indicating that the subjects experienced slightly more self motion than sphere motion. Washout seemed to reduce the vection magnitude: in the suprathreshold and subthreshold delayed-washout condition, the mean values before washout were 2.4 and 2.7, respectively, and became 1.8 and 1.9 after washout. This difference, however, did not reach significance.

Discordant conditions

Subjects were very confused in the discordant conditions. According to the verbal reports, the direction of perceived body tilt was initially determined by the tilting chair, but then quickly changed sign and became determined by the visual stimulus. But even during the constant velocity phase, subjects did not have much

confidence in their judgments. They felt tilted to one side (visually induced) and pressure to the other side. As a consequence, the subjects indicated that they were guessing when they set the tactile rod to the vertical, and they often failed to move it. Therefore, only verbal estimates were analyzed.

The magnitude of estimated self-tilt in the post-onset phase varied around zero (1.9 ±3.5 in suprathreshold condition; -2.4 ±7.0 in subthreshold condition). This suggests that the visual stimulus roughly canceled the self-tilt induced by the body

Figure 5.4. Mean angle of perceived self-tilt during the post-onset and post-washout phases in the concordant-motion conditions. Settings of the tactile rod (upper) and Verbal estimates of tilt (lower). Bars are standard errors of the means (n=9).

tilt, which is about 20º according to the verbal estimate in the subject-motion conditions. This value is similar to the experienced self-tilt in the visual-motion condition. Thus, as in the concordant washout conditions, the combination of body tilt and visual motion induced an effect no greater than the sum of effects produced by the corresponding visual-motion and the subject-motion conditions. After washout of the subject to the vertical, estimated self-tilt changed in the same direction and was no longer different from that in the visual-motion condition (9.7 ±3.2 in suprathreshold condition; 19.1 ±4.2 in subthreshold condition).

Figure 5.5. Mean vection magnitude in the post-onset and post-washout phases of the concordant-motion conditions. Vection was scored on a scale of zero to four. The bars are standard errors of the means (n=10).

There were no significant differences in vection magnitude between the concordant and discordant combined-motion conditions. In the post-onset phase the mean vection magnitude was 2.5±0.3 in the discordant suprathreshold condition, and 2.3±0.3 in the discordant subthreshold condition. After washout these values were 2.0±0.3 and 2.35±0.3, respectively, which is not significantly different from the post-onset values.

Discussion

In this study we investigated whether passive body motion accompanying motion of a visual scene about the horizontal roll axis affects the latency and magnitude of sensations of self-tilt and self-motion (vection). The magnitude of perceived tilt was greater when visual motion was accompanied by concordant tilt of the subject than when there was only visual motion or only subject tilt. This is what one would expect from addition of the response to body tilt and that to visual motion. Because the magnitude of perceived tilt did not depend on the level of subject acceleration, we conclude that the increase in perceived tilt was merely due to the tilt component of the subject roll stimulus - and not to the rotation component. Sensations of self-motion, as opposed to self-tilt, were not much affected by whether visual motion was or was not accompanied by real self-motion or real self-tilt.

There seem to have been few studies on the additivity of visual and vestibular contributions to perceived self-tilt. It has been shown that the effects of a rotating or tilted display on sensations of vection and self-tilt are larger when the head is in a tilted position (Asch and Witkin 1948; Held et al. 1975; Bishof 1978; Young et al. 1975). This non-linearity is attributed to the sensitivity of the utricles being higher when the head is erect than when it is tilted. As a result, otolith inputs would be less reliable when the head is tilted, resulting in a lower weighting of otolith inputs relative to visual inputs. We did not find that actual tilt facilitated visually-induced self-tilt. When corrected for the response to real body tilt, perceived tilt in the washout conditions was not different from that in the visual-motion condition. It is likely that the angle of 15º was too small to produce sufficient change in otolith sensitivity.

For rotation about the vertical yaw axis Zacharias and Young (1981) found that the threshold for detection of self-motion was reduced when visual motion was accompanied by real self-motion. With vertical yaw motion, however, semicircular canal inputs are involved but not otolith inputs. They proposed a model in which visual and vestibular inputs are combined with weightings which depend on whether visual inputs are consistent with vestibular inputs. Inconsistent vestibular inputs were assigned higher weightings than consistent vestibular inputs. Our results showed signs of linear addition when visual and vestibular inputs were consistent. However, when the two inputs were in conflict, the results were so variable that the additive model could not be tested.

When the body was rotated along with the visual stimulus and then returned to vertical, perceived self-tilt in the erect observer was the same as it would have been if the body had not moved. Thus the effects of real self-tilt on the magnitude of sensations of self-tilt persist when the body remains in a tilted position but do not persist after the body has returned to the vertical. In the typical washout sequence in a motion-base flight simulator, the body is returned to the vertical at near threshold acceleration immediately after being accelerated with the visual scene, as in our immediate-washout condition. Under these conditions, washout produces only a momentary potentiation of sensations of self-tilt while the body is tilted, but the effect does not outlast the washout procedure. From these results we expect that when a motion base executes a series of washout movements in close succession, the magnitude of illusory self-tilt will be larger than that produced by visual motion alone.

The main effect of body roll was a reduction of the latency of detection of self-motion. The latency of tilt sensations was shorter when suprathreshold body motion and visual motion were combined compared with when the visual scene alone was rotated, and was similar to the latency of indicating real body motion alone. A reduction in latency in a flight simulator means that the subject will show less phase lag in his response to the simulated motion. Gundry (1977) distinguished between "maneuver motion" arising from the pilot’s control of the aircraft, and "disturbance motion" arising from turbulence or aircraft failure. There has been conflicting evidence about whether motion bases in flight simulators improve pilot training, as reflected in transfer from simulator to actual flight. In general, performance has been found to improve under disturbance motion, but not under maneuver motion. When performance improved, pilot reactions were more rapid and accurate with motion-based training than without motion-based training (Caro 1979). In addition, learning time and error rate of a visual tracking task improved when visual motion was accompanied by concordant body motion. Since disturbance motion is imposed on the subject without warning, responses should have longer latency than those to maneuver motion. In maneuver motion, subjects can anticipate the motion of the aircraft and, under these circumstances, latency will be short without a motion base. Also, with disturbance motion, reaction time should be longer in the absence of real self motion than in the presence of real self motion.

We used body roll as the vestibular onset cue for self-tilt. This paradigm is not appropriate to simulate the vestibular stimulation arising from banked turns. A banked turn of a real aircraft generates a centrifugal force in excess of 1g to which the pilot’s body and the otolith organs remain aligned (Gilson et al. 1973). In real tilt of a motion base the otolith organs are displaced relative to gravity and there is no g-excess. Similar differences between real flight and motion-base simulation exist in pitch maneuvers (Cramer and Wolfe 1970). Differences of this type probably account for why the pilot’s ratings of fidelity of motion sensations in a flight simulator were not related to large changes in the motion of a motion base (Bussolari et al. 1987).

The visual display used in the present study was lined with dots which gave no indication of "up" and "down" (visual polarity). Visual motion alone of a whole-field stimulus does not completely override the contradictory information from the otolith organs in an erect observer, resulting in combined sensations of continuous vection and limited self-tilt (Howard et al. 1988). To override the otolith inputs completely, one needs a visual stimulus containing clear horizontal and vertical lines (visual frame) and features with a top and bottom, such as chairs and tables. Howard and Childerson (1994) showed that more than 60% of erect observers reported sensations of head-over-heels rotation (360º of self-tilt) when placed inside an 7-foot cubic furnished room with abundant visual polarity cues, rotating about the line of sight. However, when exposed to static tilt of the furnished room up to 120º, the mean perceived self-tilt was at most 15º. The larger effects produced by the moving room are presumably due to the combined effects of motion, the visual frame and visually polarized objects. It therefore seems that the most effective visual stimulus to induce apparent self-tilt in an upright observer is a moving polarized scene. We conclude that apparent self-tilt can be enhanced by a typical washout sequence of real body tilt, but only during the actual tilt. More importantly, the latency of perceived self-tilt is greatly reduced when visual motion is accompanied by suprathreshold body rotation. We found no effects of actual body tilt on sensations of self-rotation.

This work was supported by Defence and Civil Institute for Environmental Medicine (DCIEM) contract No. W7711-5-7256 awarded to Dr. Ian Howard.