Chapter 5

Expectation and the perception of tilt during linear horizontal ego-motion.

B.S. Mesland, W. Bles, A.H. Wertheim

Introduction

The question that underlies the experiments presented in this paper is: Does expectation influence the perception of linear horizontal self-motion. Expectation, in this context, is defined as 'the knowledge that one is going to be moved or is currently being moved along a linear horizontal rail'. Such knowledge may come through visual feedback during the motion and through memory from seeing the motion device before one mounted it. Translated into experimental specifics, the question in the first experiment presented was: how well is linear horizontal self-motion perceived when one is completely ignorant of the constraints and possibilities of the motion device one is seated on, both prior and during the passive selfmotion. Is one able to correctly specify the undergone motion as linear motion, and if so, is this motion correctly perceived as horizontal or will other (illusory) percepts, such as tilt, be reported as well?

The latter question, i.e. whether or not tilt will be reported, is a particularly interesting one because, in terms of physics, the linear proprioceptive sensors (otoliths, proprioception etc) are inertial sensors. As such they cannot differentiate between linear accelerations acting on the body and gravity. What they sense is the vector sum of all linear acceleration forces that stimulate them. This resultant vector (usually called the gravito-inertial force vector) deviates from gravity as soon as linear accelerations with a horizontal component are present.

Yet it is our everyday experience that we are perfectly able to distinguish a tilt of the head from a linear displacement, i.e. to extract gravity from the sensed gravito-inertial force vector. This can be explained by the additional influence of the semi-circular canals. When we tilt our head both the linear proprioceptive sensors and the semi-circular canals are stimulated (i.e. Guedry, 1974; Mayne, 1974). This explanation is supported by experiments in which blindfolded labyrinthine defective patients could not distinguish between horizontal linear accelerations and tilt, whereas subjects with intact peripheral vestibular organs could (Jongkees & Groen, 1950; Guedry & Harris, 1963).

There are circumstances however, in which a purely horizontal acceleration creates a sensation of tilt in healthy subjects as well. This is the case for instance in blindfolded subjects during prolonged centrifugation in a human centrifuge (Clark and Graybiel, 1966; Guedry, 1974), or in aviators during constant acceleration (see e.g. Graybiel et al. 1979). In addition, ocular torsion induced by horizontal linear acceleration has been reported (Lichtenberg et al., 1982; de Graaf et al., 1995), suggesting an output from the vestibular system similar to what happens during tilt. Such perceptions are generally explained by a leak through the low pass characteristics of the vestibular system (i.e. Mayne, 1974; Bles and Bos, 1994) and should therefore appear only in the lower frequency range of the system or at high accelerations.

It is possible however that there is another (cognitive) reason why percepts of tilt during linear horizontal acceleration are not common at higher frequencies and low accelerations: such percepts may be prevented by the knowledge that one is being accelerated along a straight horizontal rail. For this knowledge may create an expectation of the kind of motion that will be perceived. When subjects are not blindfolded, the expectation simply stems from the concurrent visual flow pattern. In total darkness the expectation could arise from the memory of what one has seen (e.g. a sled on a horizontal rail) before the start of the experiment (other cognitive cues may be provided by the sound produced by the motion device, vibrations of the device or even the air flow sensed by the subject during motion). The thus formed expectation of the undergone motion may then, in combination with the lack of canal stimulation, 'attract' the interpretation of the changing gravito-inertial force vector to a percept of horizontal motion.

The idea that expectation or 'mental set' has an influence on the perception of self-motion is not new, it has often been hinted at. It is named as a factor of the so called ' internal model' and the influence of expectation on the latency of vection has been reported (i.e. Henn et al. 1980). In addition mental set has been thought to influence tilt adaptation and viual-vestibular interaction effects on tilt percepts (i.e. Guedry, 1974). To our knowledge however, expectation has never been examined as a variable in an experiment.

The hypothesis leading to the following experiment was that at low sinusoidal linear horizontal accelerations, with a frequency that lies within the bounds of natural head movements (0.2 Hz), blindfolded subjects who have no prior knowledge regarding the device on which they are being moved, and no additional clues whilst moving, will report significantly more tilt percepts than subjects who have seen the sled and its horizontal rail before being blindfolded and who do have access to other clues whilst moving (sound of the motion device etc.).


Experiment 1

Method

Eighteen paid volunteer subjects (age 20-30) participated in this experiment. The only information they were given beforehand was that they would participate in an experiment about self-motion perception. None of them had ever visited the institute before and they were all completely unaware of the existence of the linear acceleration sled (or of any other moving device in the lab). They were given instructions in a room adjacent to the experimental room. All doors to motion labs were closed and all posters and photos concerned with motion devices had been taken away. After their introduction the subjects were blindfolded and led to the experimental room and were helped onto the seat of the sled.

The subject was moved on a linear horizontal acceleration device called 'the ESA sled' (see for a tecnical description Soons et al. 1981). The sled can move sinusoidally along a linear track with a maximum peak to peak displacement of 3.20 m. When the subjects were seated, care was taken not to provide cues as to the existance of rails. The design of the sled's seat requires subjects to sit with their legs crossed. Once seated, the five point seat-belt was fastened and ear phones, placed inside sound barring ear muffs, were put over the subject's ears. A small microphone attached to the headphones allowed for communication between subject and experimenter. White noise, shaped specifically to mask the noise characteristics of the sled, was presented through the subject's headphones, effectively blocking all outside auditory information during sled motion.

The subject was seated upright facing the front end of the rail and was told that it was important not to make any head movements during the experiment. The subject's head was supported with a vacuum cushion. To prevent air flow cues during sled motion, a (black) cloth was fastened to a metal frame around the seat. Finally, a vibration device was attached to the seat, which effectively prevented recognition of tactile cues about the movement of the (plastic sheeted) wheels of the sled (fig 1).

figure 1: The ESA sled motion decice as it was used in experiment 1.

Prior to the experiment the subjects were given only the following information: They were going to be seated in a kind of fun fair attraction, capable of making any movement one could imagine. Each time the noise in their ear phones was turned off, they would be asked to describe as accurately as possible the movements they had experienced during the prior period of noise. They were notified that they should include percepts of stationarity in their reports as well.

Four sinusoidal motion profiles were used. Each profile consisted of 5 periods of forward and backward motion along the subject's x-axis (see Table 1).

Table 1: characteristics of the four motion profiles used in the main study.

Ten subjects received the profiles in random order. They were also presented with a stationary control condition in which the sled did not move (condition O). With the 8 other subjects we used a Latin square design. The latter subjects did not perform the control condition. Since the starting positions of the sled were not identical, the sled had to be repositioned between profiles. This was done at a constant velocity of 10 cm/sec.

Before starting a profile the experimenter turned on the noise through the subject's earphones and switched on the vibration device. After finishing a profile (5 periods), when the sled had stopped moving, noise and vibration were turned off and the subject was asked to report. If necessary, the experimenter would ask for clarification. Subjects who reported percepts of horizontal motion without tilt were asked to estimate their peak to peak displacement if possible. All communication between subject and experimenter was recorded on-line on a casetterecorder.

After finishing the experiment, subjects took off their blindfold and saw the sled motion device for the first time.

The results of the present experiment were compared to data obtained in a control study which had been carried out earlier (within the framework of another research project, not relevant here). In this control study 8 different subjects (aged 20-47) participated. They were moved on the sled along their x-axis, continuously and sinusoidally for 30 min. Just as in the present study, they had been blindfolded and moved in total darkness. However, in this control study we did not attempt to prevent (prior or any other) knowledge regarding the motion restraints of the sled: the subjects did see the sled before they were seated and sound, vibrations and airflow resulting from its motion were not masked. Six subjects were moved at a maximum acceleration of 0.2 g and with a frequency of 0.17 Hz, the other 2 subjects were moved according to profile C of table 1. In this study the subjects were asked specifically to report any occurrence of a hilltop sensation (that is an illusory percept of tilt during their horizontal motion).


Results

The responses in the main study could roughly be divided into five categories:

I. Linear horizontal forward and backward motion (HOR)
II. Linear horizontal forward and backward motion combined with tilt sensations at the turning points in the forward/backward-vertical plane (X-Z plane) (HOR+TILT)
III. Moving to and fro along a 'hill like' curved path in the X-Z plane: Hilltop (HILL)
IV. Angular swing sensation in the X-Z plane (SWING)
V. Moving along a linear but tilted path in the X-Z plane, either upward or downward (SLOPE)

Only one subject experienced motion outside the X-Z plane, reporting ego motion slanted to the left and right. Since the other subjects all reported motion within the X-Z plane, we doubted the reliability of this subject's data and excluded them from statistical anaysis.

The results of the remaining 17 subjects are summarized in table 2

Table 2: number of subjects per perceptual response category for profiles A to D.

Note that the total of 0.5 HILL and 1.5 SWING in subcondition D regard a subject who scored HILL during forward motion and SWING during backward motion.

Also note that SLOPE percepts should not be interpreted simply as HOR percepts with a vertical deviation, because whenever SLOPE was perceived, a difference in slope between backward and forward motion was reported. In most cases the downward slope in forward direction was experienced as stronger than the upward slope in backward direction. In some cases a slope was only perceived during forward motion, the backward motion path being perceived as horizontal.

In condition O none of the 10 subjects that performed this condition reported motion sensations.

In the control study none of the 8 subjects reported anything but a horizontal to and fro motion during the first 5 peroids ( which is the time span that compares to the presented main study). First reports of tilt percepts did not happen until after 53 sinuoids at the lowest and 250 sinusoids at the highest.

Figure 2 is a histogram depicting the results of the main study and the first five periods of the control study. It shows, for each sled motion profile, the number of subjects that reported a HOR percept compared to the number of subjects that reported one of the 4 tilt percepts ( HOR+TILT, HILL, SWING or SLOPE).

Figure 2: Results of the main study and the first five periods of the control study of experiment 1 are depicted as a histogram. For each sled motion profile (X-axis), the number of subjects (Y-axis) that reported a horizontal (HOR) percept (shaded pillars), or a tilt (HOR+TILT, HILL, SWING and SLOPE percepts are all added together) percept (white pillars) are presented.

Most of the subjects in the main study who reported HOR motion were also able to make an estimate of their displacement amplitude. A few subjects occasionally perceived a difference between forward and backward displacement amplitude ( most of these subjects perceived forward displacement as larger than backward displacement). For them the average of the forward and the backward displacement estimates was taken as their estimate.

One subject who reported HOR motion in all conditions never noticed that he was moving to and fro, and reported only forward motion with profiles A, B and D and backward motion with profile C. This subject was excluded from our dataset, because his displacement estimates were too extreme (varying between 100 and 400 meter).
The data from subjects who reported HOR motion with 3 or all profiles are presented in Figure 3 as a function of profile (ordered along the x-axis in terms of their maximum sled velocity - see table 1).

Figure 3: Peak to peak displacement (Y-axis) estimates, of those subjects that reported HOR percepts in minimally 3 of the presented profiles, are presented for each motion profile (X-axis). The estimates are depicted as symbols, the group means are depicted as dotted lines and the actual peak to peak displacements are depicted as solid lines.

With a maximum sled velocity of 0.4 m/s, the average estimated displacement was higher than the actual displacement (mean: 1.63 , sd: 0.54, p = 0.05), whereas with a maximum sled velocity of 1.6 m/s the average displacement estimate was smaller than the actual displacement (mean: 2.48 , sd: 0.87, p = 0.05). In between these two sled velocity extremes average displacement did not differ significantly from the actual displacement (with peak sled velocity of 0.8 : mean 1.79, sd 0.80, p > 0.1; with peak sled velocity of 1.1: mean 2.59, sd 1.32, p > 0.1).


Discussion

This experiment was carried out with subjects who had no knowledge about their movement on the basis of any kind of visual or other clues apart from linear proprioceptive sensor stimulation. The data show that when such subjects are brought into sinusoidal horizontal linear motion in the X direction at low accelerations (0.04 g to 0.16 g), pure linear motion in the X-Z plane was perceived by all of them with the exception of one. In addition tilt percepts were reported almost immediately (within 5 periods) by about half of them (46%). In the control study - in which knowledge of the undergone motion was present in the form of prior visual or other sensory information - tilt percepts were much harder to obtain and happened only after prolonged oscillation (53-250 periods). It may be added that the sled motion profiles used in this experiment have been used many times before in experiments in our lab, in which subjects oscillated with prior knowledge of the constraints of the motion device, in complete darkness. In none of these experiments have tilt illusions ever been reported.

We may thus conclude that - besides the vestibular aparatus, proprioception and vision - expectations about one's motion have a clear effect on perceived linear self-motion. At the low accelerations and frequencies used in the present study, such cognitions appear to 'attract' the possible interpretations of the changing gravito-inertial force vector (horizontal linear acceleration, tilt of the head or both) towards a percept of horizontal linear oscillation. In other words one can say that thresholds for perceiving tilt can be lowered by depriving the system of knowledge about one's motion.

The average displacement estimates of subjects with a pure horizontal motion senstation seemed to relate to the maximum velocity of the passive sinusoidal motion, which agrees with data from Mittelstaedt and Glasauer (1991). These authors showed that, with blindfolded passively moved subjects, displacement is overestimated when motion is below normal walking speed, and underestimated with motion above normal walking speed. An underestimation of horizontal displacement at maximum velocities in the vicinity of walking speed was also reported by Israel et al. (1993).

The following experiment was designed to measure the size of the perceived tilt created under circumstances similar to those in experiment 1.


Experiment 2

In this experiment situation was again created in which subjects could not know whether their motion was horizontal or tilted. The sled's chair was rebuilt so that it could be sinusoidally tilted in the X-Z plane and accelerated horizontally in the X direction at the same time. The amplitude of actual sinusoidal tilt that needed to be combined with a sinusoidal linear horizontal motion, in order nullify illusory percepts of tilt, was measured. This was done by determining the thresholds for perceiving tilt in 2 directions: tilt in the same direction as sled motion and tilt in the direction opposite to sled motion. The midpoint of the 2 thresholds was taken as the amplitude of sinusoidal tilt needed to create a percept of horizontal oscillation.

Subjects were made aware of the possibilities of the motion device prior to the experiment. Thus they knew that either horizontal accelerations, or tilt of the chair, or a combination of both was possible. Therefore their expectation was more structured than in experiment 1. But since the possibilities of the set-up were equal to the percepts reported in experiment 1, and the subject had no way of knowing what combination of tilt and horizontal motion was presented, the reported sensations should be similar to those in experiment 1.

The described set-up implies concurrent stimulation of the linear proprioceptive sensors (otoliths, proprioception etc) and the semi-circular canals. Therefore a control condition was added consisting of an experimental set up in which the same combined semicircular canal and linear proprioceptive sensor stimulation happens in the absense of any linear horizontal ego motion. This was not done with the tilting chair on the sled, because even with a stationary sled, chair tilting always implied a small horizontal displacement of the head. Hence, the control condition was carried out with a different motion device: a 3-D rotatable chair. The rotation axis coinciding with the subject’s inter-aural axis, was either horizontal or vertical. In the last condition only the canals were stimulated.


Method

Thirteen paid volunteer subjects, aged between 21 and 30, participated in this experiment.

The same motion device (ESA space-sled) that we used in experiment 1 was used again, but this time an extra motion possibility was added. A device was built around the chair, such that apart from being accelerated linearly along the horizontal rail, the chair could also be tilted in the X-Z plane about an axis underneath the seat.
Subject's were seated on the sled's chair in exactly the same manner as in experiment I (that is, they were strapped in the seat belt, blindfolded, used the same head phone communication device; and the same vibration device and air flow shielding was used). The noise in the head phones was again adapted, to also mask the noise from the chair tilting motor.

Just as in experiment 1 the subject was seated facing the front end of the rail and was told that it was important not to make any voluntary head movements during the experiment. The subject's head was again supported with a vacuum cushion (see fig 4).

In both the sled and the two rotating chair control conditions, the subjects were presented with four sinusoidal motion profiles. The profiles were chosen such that two profiles (F and H) were of equal frequency but different in maximum acceleration and thus in the angle of the resultant gravito-inertial force vector relative to the body. The two other profiles (E and G) differed in frequency but had equal maximum accelerations and thus equal gravito-inertial force vectors. In addition, profiles F and G had equal maximum displacement amplitudes, but differed with respect to all other variables (see table 3).

Table 3: Linear sled motion profiles at which thresholds for sensing chair tilt were measured.

figure 4: The adapted ‘ESA sled’ motion device as it was used in experiment 2.

Whilst the subject was moving on the sled along the X-axis according to one of the profiles of table 3, the sled's chair could be tilted over a pre determined angle, set by the experimenter. This tilting motion of the sled's chair was always sinusoidal and exactly in phase with the sled's horizontal motion. Tilt thresholds were measured for two directions. They were labeled the + (plus) and - (minus) tilt thresholds. During +tilt the subject was tilted from a backward angle to a forward angle during forward sled motion, or tilted from a forward angle to a backward angle during backward sled motion. During -tilt the subject was tilted from a forward angle to a backward angle during forward sled motion and from a backward angle to a forward angle during backward sled motion. Thus during +tilt the horizontal displacement of the subject's head was slightly larger than sled displacement, and during -tilt the horizontal head displacement was slightly smaller than sled displacement. However, since the differences between the actual head motion magnitudes and sled motion magnitudes were very small (approximately 1% of sled motion magnitude), we considered them as negligible, and defined the horizontal vestibular stimulus only in terms of the horizontal sled motion characteristics.

Within each profile the thresholds for + and -tilt were measured using a staircase method with an in between trial stepsize of 1 deg. A trial consisted of 7 sled periods. During the first quarter of the first sinus and during the final quarter of the last sinus the sled's chair was kept at 0 degrees tilt (to prevent the subject from sensing tilt of the chair before it moved or after it had stopped moving). Before starting a trial the experimenter switched on the noise in the subject's earphones. The first trial of a staircase measurement sequence was always carried out with a -4 deg tilt of the chair. During each trial, while the noise was kept on, the subject was asked to continuously report whether she sensed forward or backward tilt. After each trial, the noise was turned off and the subject was asked to provide a magnitude estimate of the sensed tilt angle (if there had been a percept of tilt). In addition, subjects were asked to give an estimate of their horizontal displacement. Depending on whether a + or a - threshold was measured, if at least 3 out of the 5 middle periods of a trial yielded a + or -tilt sensation respectively, this response was scored on line as a + response or a - response respectively. In the next trial tilt amplitude was then reduced by 1 degree. If 3 out of the 5 middle periods yielded no tilt sensation, this was scored as a 0 response, and in that case tilt was increased by 1 degree in the next trial. Presentation of trials continued until the sixth turning point was reached. The average of those turningpoints was taken as the threshold measure.

In the control conditions with the 3D-rotating chair the thresholds for perceiving sinusoidal tilt were similarly measured, but this time the threshold was calculated by averaging the second to fifth turning points. It should be noted that since there was no linear displacement in this control study, tilt didn't have a sign (+ or -) like in the sled conditions (where the sign denoted whether tilt was in the same direction as, or in the direction opposite to that of the linear horizontal sled motion).

The profiles were presented in random order, both in the sled conditions and in the control conditions. Half of the subjects performed the sled conditions before the control conditions, the other half did it the other way around.


Results

Figure 5a and b depict the maximum angles of the tilt sine at the individual thresholds for sensing tilt (a) and the individual estimates of the maximum tilt angle at these thresholds (b), for sled conditions E to H. In other words, fig 5a depicts the actual tilt angle at the thresholds whereas fig 5b depicts the perceived tilt angle at the thresholds. Upright triangles depict the thresholds for perceiving a '+' tilt angle. Upside down triangles depict the thresholds for perceiving a '-' tilt angle. Theoretically tilts above the '+' threshold should yield 'hill-top' sensations and tilts below the '-' threshold should yield 'angular swing' sensations.

The midpoints between the thresholds for '-' and '+' angles in fig 5a (filled dots) reflect the tilt amplitude that is needed to create a feeling of pure horizontal motion. The filled dots in fig 5b illustrate that whereas the chair is tilting to compensate for illusory tilt percepts (5a) the subjects perceive themselves as moving horizontal.

Figure 5a: Actual chair tilt amplitude (Y-axis) at the thresholds for sensing +tilt (triangles pointing upwards) and –tilt (triangles pointing downwards) are depicted, at each motion profile (X-axis) for each subject (seperate graphs). The black dots represent the chair tilt amplitude that is needed to compensate illusory tilt percepts and create a percept of pure horizontal oscillation.

Figure 5b: Figure 5b is similar to figure 5a, except that the triangles represent the perceived tilt amplitude at the thresholds for sensing +tilt (triangles pointing upwards) and –tilt (triangles pointing downwards). Accordingly, the black dots represent perceived tilt as well.

Figure 6a: The group averages of actual chair tilt amplitude (Y-axis) at the thresholds for sensing +tilt (triangles pointing upwards) and –tilt (triangles pointing downwards), at 4 motion profiles (X-axis). Black dots represent the chair tilt amplitude needed to compensate for illusory tilt percepts and create a percept of pure horizontal oscillation.
Figure 6b: Figure 6b is similar to figure 6a except that in this case perceived tilt amplitude is represented.

The group averages of the same data are shown in figure 6a (actual tilt angles) and 6b (perceived tilt angles). At group level, subjects perceived themselves as moving horizontally when they were actually tilting slightly. The difference in tilt amplitude between the two data sets (actual tilt amplitude (a) and perceived tilt amplitude (b)) was significant ( ANOVA: p = 0.01), there was no significant effect of motion profile (p = 0.2), nor was there a significant interaction between data set and motion profiles (p = 0.09).

Table 4: means and standard deviations of actual and perceived tilt amplitude data

The area between the two opposite thresholds in figs 5a an 6a reflects the extent to which the amplitude of chair tilt may vary around the midpoint without being noticed. Thus the difference between the + and - thresholds divided by 2 may be compared to the thresholds measured in the rotating chair conditions. When comparing both control (rotating chair) conditions and the sled condition a significant effect of condition (MANOVA: p = 0.002) and motion profile (p = 0.04) was found. The motion profile effect seems to be caused mainly by the sled conditions in which the average threshold increases with the velocity amplitude (see fig 7). The interaction of condition with profile was found to be just not significant (p =0.07). A Tukey test revealed a significant difference between the chair upright conditions and the chair sideways conditions (p=0.04), and between the chair upright conditions and the sled conditions (p=0.002) but not between the chair sideways conditions and the sled conditions (p=0.4) (see fig 7 for average threshold data of all conditions).

Figure 7: Average threshold data (Y-axis) for the four motion profiles (X-axis) and three conditions of experiment 2.

For all treshold staircase sequences taken together there was a total of 130 trials during which chair tilt happened to be 0, i.e. the chair remained in a vertical position (no tilt). Across the whole experiment, these trials were more or less evenly distributed among the four profiles (28 for profile E; 41 for profile F; 37 for profile G; and 24 for profile H). In 62.3% of these instances the subjects responded that they experienced tilt (64% for profile E; 56% for profile F; 62% for profile G; and 67% for profile H).

When we divide the means of table 4 by the amplitude of the gravito-inertial force vector of the respective motion profiles (5.7 deg for profiles E, F and G and 11.3 deg for profile H), we find the fractions 0.22, 0.24, 0.21 and 0.24 respectively. Thus, little less than one fourth of the tilt amplitude of the changing GIF vector seems to be interpreted as tilt, independent of frequency and motion amplitude.


Discussion and Conclusions

First of all the results of experiment 1 were replicated in experiment 2 in the 130 instances in which the sled moved purely horizontally. In more than half of these instances a tilt percept was reported. The amount of tilt percepts was actually slightly higher in experiment 2, which may have been due to the fact that subjects knew that tilt was one of the options of the motion device, as opposed to the subjects of experiment 1, who had no idea at all what was going to happen.

To measure the size of this tilt percept, in experiment 2 we measured the amplitude of sinusoidal tilt that needed to be combined with a horizontal oscillation in order to create a percept of pure horizontal motion. On average this amplitude proved to be slightly under one fourth of the amplitude of the induced change in the gravito-inertial force vector. The result seems to be independent of the frequencies used and of displacement amplitude. These results are in accordance with a model on the perceived vertical, proposed by Bles and Bos (1994).

A significant difference between amplitude of perceived tilt motion thresholds was found between the two rotating chair conditions. As mentioned earlier, with rotation about a vertical axis only the semi circular canals are stimulated whilst with rotation about the horizontal axis, the linear proprioceptive sensors are stimulated as well. Consequently the results indicate that when linear proprioceptive sensor stimuli are added to semi-circular canal stimulation, sensitivity to rotation around the inter-aural axis is significantly increased.
However, the results of the sled conditions show that sensitivity is decreased again when a linear horizontal acceleration is added to the tilt. It was found that the chair upright condition differed significantly from the sled condition whereas the chair sideways condition did not. Apparently when a sinusoidal linear horizontal acceleration is added to a sinusoidal tilt in the X-Z plane, thresholds for perceiving tilt are increased. In addition, in the sled condition the thresholds for perceiving tilt of the chair seem to increase with the velocity amplitude of the profiles used. Thus the higher the velocity of horizontal linear motion, the higher the threshold for perceiving tilt of the chair.

The thresholds we are referring to in the sled conditions are not situated symmetrically around true horizontal motion but around perceived horizontal motion. As discussed above, the latter implies a low amplitude sinusoidal tilt motion of the sled's chair. This means the canals are stimulated, which is apparently not noticed by the system (even though the –thresholds were clearly above the thresholds obtained with stimulation of the canals only {rotation about a vertical axis}).
In addition when the threshold for a '+' angle is negative (e.g. as in all conditions of subject 7, fig 5b), this means that the canals are stimulated in the direction opposite to the linear proprioceptive sensors. The linear sensors are stimulated in the + direction whereas the canals are stimulated in the - direction. Therefore it seems fair to conclude that in the set-up used in this experiment the linear proprioceptive sensors dominate the self-motion percept.

It is possible that the increased threshold levels of the sled study, in relation to those of the upright chair condition of the control study, reflect an increased noise level within the inertial perceptual system, due to missing some of the necessary information (prior knowledge and vision) to form a correct percept. Passive linear horizontal motion is a rather unnatural kind of motion, which has been invented by human kind in the form of buggies, cars, trains etc). The possibility then exists that the capability of human's to perform path integration tasks during passive linear motion (i.e. Mittelstaedt and Glasauer, 1991; Israel et al., 1993; Klatzky et al.,1990) has appeared thanks to cognitive abbilities such as the ones discussed in this paper. When prior knowledge and vision are not present the final percept may depend on the thresholds and on past experience of the perceptual system (i.e. memory of similar sensations). This idea may explain the variety of different sensations that were reported in experiment 1 and the individual differences in experiment 2. A strong influence of cognition may also explain the fact that no differences in self-motion percept were found between linear horizontal oscillation on earth and linear horizontal oscillation in space (Arrot and Young, 1986).

The above described idea of a strong influence of cognition (prior knowledge and past experience) may be illustrated by a study on homing behaviour with gerbils (Mittelstaedt & Mittelstaedt 1982). The results of this study showed that when in complete darkness gerbils are linearly and horizontally accelerated away from a certain spot in relation to their nest (in the temporal range of their normal movements) they seem completely unaware of the fact that they have been displaced. As a result, in searching their nest they miss it by a linear component equal to the distance and direction over which they have been displaced. In the same study it is shown that angular accelerations under the same conditions are clearly noticed by the gerbil and corrected for when searching the nest (see also Etienne et al.1988). If, as argued above, passive linear proprioceptive motion percepts without the aid of vision are influenced by cognitive abilities, such as prior knowledge and past experience, this behaviour of the gerbil is understandable. Since gerbils are not usually moved around passively, they have no memory of such a sensation and the percept that goes with it. They have never learned to use their linear motion system for path integration purposes. Humans however have a wide experience of linear passive motion and thus come up with a variety of linear motion percepts. Apparently for angular motion perception the story is a different one. This can be understood when it is realized that with the horizontal canals the similarity in sensation between horizontal movement and tilt does not exist. As a result angular motion without vision or prior knowledge is less ambiguous than linear motion and thus the influence of past experience in the forming of a percept may be less important.

References

Arrot A.P. and Young L.R. (1986) Perception of linear acceleration in weightlessness. Proceedings of the Norderney Symposium on Scientific Results of the German Spacelab Mission D1 (Sahm, Jansen en Keller eds.) Wissenschaftliche Projectführung D1 c/o DFVLR, 5000 Köln, Germany, p. 490-499

Bos J.E. and Bles W. (1997) A vestibular based motion sickness incidence model. TNO Report TM-97-A032, TNO Human Factors Research Institute, Soesterberg, the Netherlands.

Clark B. and Graybiel A. (1966a) Factors contributing to the delay in the perception of the oculogravic illusion. Amer. J. Psychology 79, 377-388.

De Graaf B., Bos E.B., Tielemans W., Rameckers F., Rupert A. and Guedry F. (1995) Otolith contribution to ocular torsion and spatial orientation during acceleration. NAMRL Technical memorandum 96-3, Naval Aerospace Medical Research Laboratory, Pensacola, Florida.

Etienne A.S., Maurer R. and Saucy (1988) Limitations in the assessment of path dependent information. Behavior 106, 81-111.

Graybiel A., Johnson W.H., Money K.E., Malcolm R.E. and Jennings G.L. (1979). Oculogravic illusions in response to straight-ahead acceleration of a CF-104 aircraft. Aviat. Sp. and Environm. Med. 50, 383-386.

Guedry (1974) Psychophysics of vestibular sensation. In: Handbook of Sensory Physiology (Kornhuber H. H. ed.), Vol. VI/2, Springer Verlag Berlin Heidelberg, New York, 3-154.

Guedry F.E. and Harris C.S. (1963) Labyrinthine function related to experiments on the parallel swing. NSAM-874. Pensacola, Florida: Naval School of Aviation Medicine.

Henn V., Cohen B. and Young L.R. (1980) Visual-vestibular interaction in motion perception and the generation of nystagmus. Neurosciences Research Program Bulletin. Published by the MIT press, ISSN 0028-3967, Vol.18, no.4, p. 558.

Israël I., Chapuis N., Glasauer S., Charade O. and Berthoz A. (1993) Estimation of passive linear whole-body displacement in humans. Journal of Neurophysiology, Vol.70, no.3, 1270-1273.

Jongkees L.B.W. and Groen J.J. (1950) A quantitative analysis of the reactions of a person after loss of the function of both inner ears. J. Laryng. 64, 135-140.

Klatzky R.L., Loomis J.M., Golledge R.G., Cicinelli J.G., Doherty S. and Pellegrino J.W. (1990) Acquisition of route and survey knowledge in the absence of vision. Journal of Motor Behavior, Vol.22, no.1, 19-43.

Lichtenberg B.K., Young L.R. and Arrot A.P. (1982) Human ocular counterrolling induced by varying linear accelerations. Experimental Brain Research 48, 127-136.

Mayne R. (1974) A systems concept of the vestibular organs. In: Handbook of Sensory Perception (Kornhuber H.H. ed.), Vol. VI/2 Springer Verlag Berlin Heidelberg, New York, 493-580.

Mittelstaedt M. and Glasauer S. (1991) Idiothetic navigation in gerbils and humans. Zool. Jb. Physiol. 95, 427-435.

Mittelstaedt H. and Mittelstaedt M.L. (1982) Homing by path integration. In: Avian Navigation (ed. by Papi/Wallraff), Springer Verlag Berlin Heidelberg, 290-297.

Soons A.F.L., Burden D.F., Garvin M.J. and Wyn-Roberts D. (1981) The Space Sled – A European facility for life-science experiments in Spacelab. ESA Journal, Vol.5, 55-108.