Chapter 4

Rotation of a visual display about an earth-horizontal axis induces a sensation of continuous self-motion accompanied by a sensation of self-tilt in the opposite direction, or equivalently an apparent tilt of the subjective vertical in the same direction (Dichgans and Brandt 1978). The degree of illusory self-tilt is determined by several factors. First, the effect increases with the area of the rotating display, especially when the stimulus motion expands to the retinal periphery (Held et al. 1975). Second, illusory self-tilt is a function of stimulus velocity, but only up to about 40º/s (Held et al. 1975; Howard et al. 1988). Third, the strength of the effect depends on the orientation of the head relative to gravity. The effect is stronger when the head is tilted 90º to the side, or inverted, as compared to erect (Young et al. 1975; Dichgans et al. 1975). Additionally, the axis of rotation relative to the subject may be important too. It was found that illusory self-inclination is larger for backward rotation of the scene about the subjects’ y-axis (pitch) compared to forward pitch motion (Young et al. 1975). In general, the average maximum illusory body inclination does not exceed 20º, which is attributed to the restraining influence of the otolith organs.

Furthermore, illusory self-tilt depends on the type of visual information in the display. Three types of visual cues contribute to the sense of body orientation with respect to gravity: visual motion, visual frame, and visual polarity. A visual frame is a set of lines and surfaces that constitute a frame which is normally aligned with gravity. Visual polarity comprises objects which show a identifiable "up" and "down". Visual motion alone can be obtained by rotation of a display with a texture lacking visual frame and visual polarity cues. Most studies on visually induced self-tilt have applied rotating displays textured with non-polarized dot patterns. Recently, Howard and Childerson (1994) used a 7 feet cubic room, which contained distinct cues for visual frame and visual polarity, and which rotated about the subjects’ roll-axis. In total, 60% of the upright subjects reported a compelling sensation of complete self-rotation, meaning that the information of the graviceptive sensors was completely overridden by the dominant visual stimulus. The other 40% of the subjects experienced a limited degree of self-tilt, either constant or alternating. During the latter, the subjects perceived self-tilt with the body becoming erect each time a surface became horizontal. Some subjects noted the feeling that they were supine and experienced self-rotation about a vertical roll-axis.

The so-called "Tumbling Room" used by Howard and Childerson was rebuilt to make it more realistic and provide it with a richer variety of polarized objects. The study described here was performed using this new Tumbling Room, with the following objectives: (1) Determine the percentage of subjects that perceive full self-rotation when immersed in a highly polarized scene. (2) Examine the effects of body posture on illusory self-tilt produced by rotation of the room (in the previous study the subjects sat in an erect posture). (3) Compare the effects of rotation of the room for different viewing directions of the subject relative to the room (roll, pitch, or yaw). Because the results showed that the rotating Tumbling Room was very effective in producing sensations of complete self-rotation, it was of interest to know whether (4) static tilt of the room could induce sensations of self-tilt which exceeded previously reported illusory self-tilt in response to static tilt of visual scenes. This question was addressed in an additional experiment by testing erect subjects with rotation about the roll axis only.

Methods

The furnished room consisted of an 8 foot cubic frame lined with plywood on the floor and foam plastic sheets on the walls and ceiling (Figure 4.1). One side wall had a window with a lighted outdoor scene, a valance curtain and plants. The opposite wall had an interior house door that allowed access to the room. The walls were papered and decorated with pictures, hanging objects, and a bookcase holding books, a teddy bear and three cups hanging from cup hooks. There was a small kitchen table screwed to the floor, set with cutlery, plates, cups and saucers, a bread basket and a box of tissues. Two wicker chairs either side of the table completed the scene. The floor was carpeted and there were baseboards against the walls, and additional floor details included waste baskets and slippers. All objects were carefully held in place by either hidden bolts, glue or Velcro, so that they stayed in place when the room rotated. In the center of the ceiling was a simple 60W fluorescent light fixture with a plastic cover. Thus, the room provided a clear visual frame and a rich variety of visual information about which direction was up and which direction was down.

Figure 4.1. The interior of the furnished Tumbling Room (upper picture) tilted 180º. A variety of objects was fastened to the walls and floor. The room rotated about a fixed earth-horizontal axis (lower picture). By changing the orientation of the chair, the effective rotation axis relative to the subject could be varied between roll, pitch, or yaw.

Bearings were placed on the centers of an opposite pair of walls and the whole structure was mounted on external brackets so that it could be rotated about a horizontal axis passing through the center of the room. Room rotation was generated by a computer controlled DC servomotor. The subject was strapped into a seat supported on a 10 inch square boom which protruded through one wall along the axis of rotation of the room. The chair could be removed and replaced by a bed, which was suspended between the boom on one side and a shaft on the other side of the room.

Constant velocity rotation of room

Profile

The room rotated about an earth-horizontal axis at a constant velocity of 30º/s for 4 cycles over a period of 48 s. The initial acceleration and final deceleration were 5 º/s² applied over a period of 6 s. The lights were on all the time and the subjects looked straight ahead at the center of the wall in front of them.

The first experimental variable was the rotation axis with respect to the subject. The subject was placed in each of three postures so that the room rotated about the subject’s roll, pitch, or yaw axis. For roll, the subject was looking along the axis of room rotation. For pitch, the chair was turned so that the subject was facing the side wall with the door. For yaw, the chair was replaced by the bed with the subject lying along the axis of rotation of the room. The second variable was body orientation. In roll, the subject was seated upright, tilted left ear down or right ear down. In yaw, the subject was lying either in supine position, or with left ear down or right ear down. In pitch, the subjects were seated upright or pitched backwards 90º into supine posture, but they were not pitched forward, since this would have been an awkward position. Finally, the direction of room rotation was either clockwise or anti-clockwise. The 16 conditions were arranged in blocks per rotation axis. The order of conditions within each block and the order of blocks was counterbalanced over subjects. Additionally, conditions were alternated for direction.

Thirty-two subjects (of whom 12 were familiar with the purpose of the experiment) were tested in each of 16 conditions. Subjects were asked to report their sensations of self-tilt according to the following categories (Howard and Childerson 1994): (a) "Full tumbling", subjects felt as if they had rotated completely through 360º. (b) "Alternating self-tilt", subjects felt that they were rotating opposite to the room to a certain angle of self-tilt, then felt suddenly upright, and then began tilting again. (c) "Constant self-tilt", subjects felt inclined at a more or less constant angle in the opposite direction of room motion. (d) "Supine", subjects felt as if they were lying supine, looking up at the rotating display that now appeared to be rotating in a horizontal plane. (e) "No effect", subjects did not experience self-tilt but felt veridically upright.

The subjects were also asked to rate the magnitude of their sensations of self-motion (vection) on a seven point scale (0 - only room motion; 1 - much more room motion than self; 2 - more room than self; 3 - equal room and self; 4 - more self than room; 5 - much more self than room; 6 - only self-motion).

Static tilt of room

The subject sat erect in the chair with the head near the center of the furnished room. The room rotated about the subject’s roll axis and brought to rest at one of four angles with respect to the vertical. While the room was moving, subjects kept their eyes closed for a fixed time so that the duration of eye closure could not indicate how far the room had moved. The angles of scene tilt were 0, 40, 80, and 120º, either clockwise or anti-clockwise. There were thus seven conditions, each repeated four times per subject.

Response

Subjects indicated their subjective vertical by aligning two black one-inch diameter discs ("indicator") with the apparent vertical. The discs were placed on a transparent circular disc about 1 foot in diameter which was mounted at its center on a horizontal shaft protruding from the axis of room rotation in the facing wall, about five foot in front of the subject. The setting of two dots was preferred to the setting of a line because of possible tilt contrast effects between a line and the room. The subject could set the disc to the desired angle by means of a servo device which controlled a motor attached to the disc shaft. There was some overshoot each time the disc rotated. However, subjects quickly learned to handle the rotating device with a minimum of overshoot. Six subjects were tested in each condition, making 168 randomized and counterbalanced trials.

Results

Constant velocity rotation

The rotating room induced compelling sensations of self-rotation about an earth-horizontal axis. Subjects were aware that the pressure sensations acting on the five-point harness were not what they expected from full self-rotation but this did not prevent them from feeling that they had rotated through 360º. After completing the experiment, several subjects found it difficult to believe that they had remained stationary and the room had been rotated. The percentages of the different kinds of sensations produced by room rotation are set out in Figure 4.2 for each axis of rotation. There were no significant effects of the factors rotation axis, body orientation, or direction of rotation on the type of self-rotation (tested with Pearson’s Chi-square). In total, 83.8% of the 512 responses were "full tumbling"; 8.4% were "alternating self-tilt"; 5.5% were "constant self-tilt"; and 2.3% were "no effect". The number of subjects that accounted for these percentages are shown above the bars in Figure 4.2 (in fact, these numbers indicate how many subjects experienced the regarding sensation in at least one of the trials). There were no "supine" responses.

Figure 4.2. Percentage of all 512 responses (32 subjects in 16 conditions) for each category of illusory self-tilt in the tumbling room for each axis of rotation with respect of the subject’s body for a constant room velocity of 30º/s. The numbers above the bars indicate the number of subjects who experienced the corresponding sensation at least once.

Mean vection magnitude over the 32 subjects for room rotation about the subject’s roll, pitch, and yaw axes are shown in Fig. 4.3. Vection data were subjected to an ANOVA (within subjects design). There were no significant effects of rotation axis, body orientation, or direction of rotation. Vection was on average relatively well saturated with an average of 4.6 (st. dev.1.8), which means that - except for some residual sensation of room motion - most of the visual motion was attributed to self-motion.

Figure 4.3. Mean vection magnitude in the tumbling room for each axis of rotation with respect to the subject’s body during constant room velocity of 30º/s. Bars are standard errors of the mean (n=32).

Static tilt

The mean subjective vertical (constant error from true vertical) and the mean variable errors (standard deviation) are shown in Figure 4.4. There was a significant main effect of scene tilt angle on the subjective vertical (F=4.99; df=3; p<0.05), but only between 0º and 40º (post-hoc Tukey: p<0.05). There was a non-significant trend for the effect of room tilt on the variable error between the four repetitions per condition (F=3.0; df=3; p=0.06). The data from the condition in which the room was vertical show that the accuracy of the settings was quite good. On average, the visual stimulus was set at -1.1º, that is, slightly anti-clockwise from the true vertical. The precision was 1.5º (mean standard deviation).

Figure 4.4. Mean constant error (a) of the vertical setting of the indicator (subjective vertical) for each angle of room tilt (six subjects with four measures per condition). Corresponding mean variable errors (b) of the subjective vertical for each angle of room tilt. Bars are the standard errors of the mean. (Click to enlarge graphs)

Figure 4.5 shows the distribution of all settings of the indicator to vertical. For 40º of room tilt, some subjects aligned the indicator with the room, while others showed only a tendency to align it with the room. In only one trial of one subject was the indicator set in the opposite direction of the room. For larger angles of room tilt, 80º and 120º, the response varied greatly between subjects. One subject (S6) aligned the two dots almost exactly with the tilted room in all conditions. Another subject (S4) sometimes reported that the indicator appeared vertical when it was close to being aligned with the room and sometimes when it was close to the actual vertical. There was no time limit in setting the indicator, and this subject kept rotating it over large angles before making a decision. As a result, his settings for 80º and 120º scene tilt were rather scattered. Two other subjects (S1, S3) consistently set the indicator close to the true vertical. Finally, two other subjects (S2, S5) seemed to use either the real floor or the surface that was closest to horizontal as the apparent floor. This resulted in a bimodal distribution of their responses which is reflected in the mean distribution settings over subjects in Figure 4.6. The settings of the indicator for each subject are shown in Figure 4.5.

Figure 4.5. Settings of the indicator to the apparent vertical for each subject.

Figure 4.6. Distribution of vertical setting of the indicator for each angle of room tilt, collected in 5º intervals. Data from clockwise and anti-clockwise trials collapsed (n=6; 4 repetitions for each angle of room tilt). Positive angles indicate that the stimulus was rotated from the true vertical in the direction of the room (with the exception of the upright room condition, where the sign is not related to the room tilt, but illustrates the accuracy of the indicator setting).

Discussion

Constant velocity rotation

All subjects experienced complete self-rotation (self-tilt through 360º) in most of the trials when exposed to rotation of the furnished room. Over 80% of the responses were "full tumbling". Subjects also experienced strong sensations of self-motion (vection) whether or not they experienced full self-rotation. In the furnished room used by Howard and Childerson (1994), 60% of subjects reported full self-rotation. This room was smaller and contained less polarized visual features. There have been no other scientific reports of sensations of full self-rotation produced by visual stimulation alone in the presence of restraining information from the otolith and somatosensory organs. People experienced full tumbling in the "haunted swing", a fairground device that was built towards the end of the last century (Wood 1895). The room rotated about the pitch axis, while the observers sat in a stationary gondola with their heads several feet below the axis of rotation. In this case, the information of the otolith organs was not necessarily contradictory with the simulated motion, since the resultant G-force would have stayed approximately in line with the body axis when the gondola were to have actually rotated. In the present study the observer’s head was near the axis of rotation, so that during actual body rotation the otolith organs would have indicated self-tilt. The high rate of full tumbling therefore demonstrates that immersion in a rotating polarized scene can fully override the restraining otolith organ and semicircular canal inputs.

There was no difference in incidence of self-rotation sensations between the three axes of rotation: roll, pitch, and yaw. Nor was there a measurable effect of body orientation with respect to gravity. Previous reports showed that vection magnitude is larger for yaw rotation than for pitch, which in turn is larger than for roll (Howard et al. 1988). Moreover, visually induced self-tilt and self-rotation normally increase when the body is inclined to gravity (Young et al. 1975). The reason that these effects were not found in the present study must be due to the general effectiveness of the furnished room. Presumably the response was already saturated in an upright body posture, and tilting the body could not add to this.

Static tilt

Subjects were most consistent in setting the visual display to the apparent vertical when the furnished room was upright. The variable errors were comparable to those reported by Witkin and Asch (1948). When the room was tilted 40º the position of the apparent vertical shifted in the direction of room tilt and was equal to room tilt for some subjects. For tilt angles of 80º and 120º, responses became highly variable. Subjects became confused and their disorientation sensations were less stable. Some subjects fluctuated between two criteria of verticality: the room, as indicated by the visual polarity, and the actual direction of gravity as signaled from the otolith and somatosensory sense organs.

Asch and Witkin (1948a) had subjects stand in the laboratory and look into a furnished room tilted 22º. Settings of a rod inside the room to the apparent vertical were displaced on average 15º in the direction of room tilt. A simple tilted frame produced an apparent tilt of only 6º. In the present experiment subjects were immersed in the tilted room, and the angles of scene tilt were larger. This resulted in larger effects. In a similar study, Howard and Childerson (1994) observed on average only 15º at 40º of room tilt, with a maximum of 60º. At larger angles of scene tilt, there was a strong tendency of the subjects to use the surface-closest-to-horizontal as subjective floor. In the same study, results were compared to the effects of a framed room without any polarity. The results were similar, except for 40º of room tilt, where subjects in the framed room took the diagonal (a 45º tilted square frame may be seen as an erect diamond with a "vertical" at 45º intervals).

Conclusion

It is concluded that a static, richly polarized visual scene in which the subject is immersed can induce illusions of self-tilt far greater than previously reported. However, for most subjects, a static polarized room is not as effective as a rotating room in inducing strong illusory self-tilt. The full self-rotation experienced by most people when immersed in a richly polarized and moving visual environment, such as the rotating furnished room, must represent the combined effect of motion, visual frame and the polarity features in the room. With the rotating room subjects were generally consistent in their responses and did not report the ambiguity that they experienced in the static tilted room.

Acknowledgments

This study was performed at Dr. Ian Howard’s Human Performance Laboratory of the Institute for Space and Terrestrial Science (located at York University, Toronto, Canada). Financial support was provided by Dr. Bob Cheung of the Defence and Civil Institute for Environmental Medicine (DCIEM) under contract No. W7711-5-7256. I would especially like to thank Heather Jenkin for her help in running the experiments.