Underestimation of self-tilt increases in reduced gravity conditions

Abstract

BACKGROUND:

During large angles of self-tilt in the roll plane on Earth, measurements of the subjective visual vertical (SVV) in the dark show a bias towards the longitudinal body axis, reflecting a systematic underestimation of self-tilt.

OBJECTIVE:

This study tested the hypothesis that self-tilt is underestimated in partial gravity conditions, and more so at lower gravity levels.

METHODS:

The SVV was measured in parabolic flight at three partial gravity levels: 0.25, 0.50, and 0.75 g. Self-tilt was varied amongst 0, 15, 30, and 45 deg, using a tiltable seat. The participants indicated their SVV by setting a linear array of dots projected inside a head mounted display to the perceived vertical. The angles of participants’ body and head roll tilt relative to the gravito-inertial vertical were measured by two separate inertial measurement units.

RESULTS:

Data on six participants were collected. Per G-level, a regression analysis was performed with SVV setting as dependent variable and head tilt as independent variable. The latter was used instead of chair tilt, because not all the participants’ heads were aligned with their bodies. The estimated regression slopes significantly decreased with smaller G-levels, reflecting an increased bias of the SVV towards the longitudinal body axis. On average, the regression slopes were 0.95 (±0.38) at 0.75 g; 0.84 (±0.22) at 0.5 g; and 0.63 (±0.33) at 0.25 g.

CONCLUSIONS:

The results of this study show that reduced gravity conditions lead to increased underestimation of roll self-tilt.

Keywords

Perception orientation A-effect space microgravity subjective vertical

1 Introduction

According to early observations by Aubert [1], roll body tilt in darkness induces apparent tilt of a truly vertical luminous line in the opposite direction of the body tilt itself. The consequence of this so-called Aubert-, or “A-effect”, is that the subjective visual vertical (SVV), measured by aligning a line with the perceived vertical, is systematically biased towards the longitudinal body axis [2 , 25]. This egocentric bias is indicative of an underestimation of body tilt, and increases with tilt angle to reach a maximum at about 130 degrees of body tilt [26 –28]. At body tilt angles below 60 deg, the A-effect is usually less pronounced, and sometimes even a bias in the opposite direction (designated “E-effect”) is observed (see [18], for a review). In its most simple form, the A-effect can be mathematically described by considering the subjective vertical as the sum of two vectors [10 , 30]. First, a gravity vector derived from somatosensory and vestibular cues (in particular the otolith organs), which sense the direction of the gravitational acceleration in body and head coordinates, respectively. When the head is aligned with the body, the vestibular cues are also in body coordinates. Second, an “idiotropic” vector, defining the above-mentioned egocentric bias [23]. In the context of optimal cue integration, it has been compared with the a priori prediction that the head is upright [5].

The vector model has been derived from laboratory experiments under normal gravity conditions of 1 g. As illustrated in Fig. 1, this model predicts that the idiotropic vector becomes more dominant in reduced gravity conditions, which would lead to a larger A-effect (at equal angles of self-tilt). Indeed, observations during space flight suggest that in the absence of a gravitational acceleration the subjective vertical becomes entirely dependent on the longitudinal body axis [12]. However, few studies have investigated the perception of verticality at intermediate gravity levels, i.e., between 0 g and 1 g, and, when doing so, showed inconsistent results. Harris et al. [16] found no effect of lunar gravity (0.16 g) on the oriented character recognition test (OCHART) compared to normal gravity. In contrast, De Winkel et al. [9] found that the SVV was predominantly aligned with the longitudinal body axis under lunar gravity, whereas under Martian gravity (0.38 g) the SVV was predominantly aligned with the gravito-inertial acceleration vector. In both studies, the angle of body tilt itself was not manipulated: the participants were either seated upright [16], or lying on their side [9]. The current study was performed in partial gravity during parabolic flight with the objective to test the vector model shown in Fig. 1 by investigating the SVV at various angles of roll body tilt under different levels of partial gravity.

Fig. 1

The subjective vertical can be represented by a vector SV that results from summation of the gravity vector g, which is derived from vestibular and somatosensory cues, and an idiotropic vector i, which is defined by the longitudinal body axis. The shift of the SV vector towards the idiotropic vector represents an underestimation of self-tilt. The three gray-coded g-SV vector pairs illustrate the hypothesis that, for a given tilt angle, the underestimation of self-tilt increases when the gravity vector amplitude is reduced, leading to a larger relative weighing of the idiotropic vector.

2 Methods

2.1 Participants

In each of three parabolic flights, two participants could be tested, so that six subjects were anticipated. The participants did not take anti-motion sickness medication, which could affect their vestibular response. Two participants dropped out half-way through the flight due to nausea. Two participants from other research teams on the flight took their place. Anticipating such a situation, these extra participants had already been checked and trained before the flight. Hence, usable datasets were collected in six people, four males and two females. Their aged ranged from 28 to 62 years with an average of about 39 and a standard deviation of 12.4.

All participants gave their written informed consent for inclusion before they participated in the study. The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved in advance by the Comité de Protection des Personnes Nord Ouest III (Caen, France) and the European Space Agency Medical Board.

2.2 G-level

The study was performed during a partial gravity parabolic flight campaign organized in 2018 by the International Space Life Sciences Working Group (ISLSWG) comprising space agencies representing Italy (ASI), Canada (CSA), France (CNES), Germany (DLR), Europe (ESA), Japan (JAXA), and the U.S. (NASA). The flights were performed by Novespace (Bordeaux, France) in an Airbus A310. During each flight 30 parabolas were flown in six series of five parabolas, separated by a 5 min or a 10 min break. Gravity levels of 0.25, 0.50, and 0.75 g were flown for 10 parabolas each. The durations of the parabolas were on average about 22, 32, and 45 sec for the 0.25, 0.50, and 0.75 g levels, respectively. The order of the gravity levels was varied each day.

2.3 Self-tilt

Self-tilt was achieved by a tilt device, which was directly bolted to the cabin-floor, and consisted of a metal frame carrying two race car seats in a back-to-back configuration (Fig. 2). This allowed for testing two participants simultaneously. Because of the seats’ configuration, tilt was always clockwise for one participant, and anti-clockwise for the other. The participants were told in advance in which direction they were tilted. Tilt was applied using a rack and pinion jack which was manually adjusted by an operator and calibrated to four angles of self-tilt (0, 15, 30, and 45 deg). The seats were always brought into the desired tilt position at the beginning of the partial gravity phase (“injection”), immediately after the “pull-up” phase. This was done to prevent participants from assessing their tilt angle under the familiar 1 g condition of level flight. After the “pull-out” phase, back in level flight, the seats were returned to upright again. Seat tilting took about 3-4 seconds for the 45 deg angle (and less for smaller tilt angles), corresponding to a tilt rate on the order of 10–15 deg/s. Shortly after the seats were in the desired position, the SVV measurement started. Although the tilting of the seats may theoretically have induced a post-rotatory sensation in the semi-circular canals, we expect that this was negligible.

Fig. 2

Experimental set-up on board the aircraft. A subject is wearing the head mounted display for executing the SVV task, while the seat is being repositioned by one of the operators.

To minimize body movement in tilted positions, extra cushioning of the seats provided full upper body and head support, and additional supports kept the feet and knees in place. No straps were used to further fixate the head against the headrest, so that the subject would be free to move their head in case of emergency. To limit misalignment of the head relative to the trunk, participants were instructed to rest their head to an individually sized cushion.

The actual head tilt of each participant was continuously recorded in-flight with small sensors containing 3-axis accelerometers (Shimmer Sensing), mounted to the helmet mounted display (HMD) used for the psychophysical task. Two other sensors were mounted on the tilt device: one on its tiltable part, and the other on its base of the tilt device, i.e., directly connected to the cabin floor. The individual sensors weighed about 24 grams, were battery powered, and acted as stand-alone devices that continuously recorded data locally during the flight.

2.4 Subjective Visual Vertical (SVV)

The SVV task was administered on the HMD (Oculus Rift), featuring one OLED panel for each eye. Both panels had a resolution of 1080×1200 pixels; a refresh rate of 90 Hz; and a field of view of 100 deg horizontal×80 deg vertical. The HMD prevented the detection of any external cues that could serve as visual reference, such as external light penetrating the circumference. The HMD had a weight of 460 grams and was interfaced with a laptop through a HDMI cable for video and a USB cable for power and data transport. The USB port provided 2.5 W (5 V@500 mA). The participants were only allowed to take off their HMD during level flight to prevent estimation of the tilt angle during the parabolas.

The participants’ task was to align a luminous bar projected on the HMD with the perceived direction of the gravitational vertical. To prevent aliasing in the pixelated display, the bar was composed of a linear array of red dots with a larger one denoting “up”. The stimulus distance was set to 20 m, and the length of the line was arbitrarily chosen during experiment design. The bar was presented against a dark background. Its initial orientation was set in a semi-random fashion in each trial, alternating between clockwise and anti-clockwise starting positions, with a maximum offset of±40 deg relative to the actual tilt angle. The bar could be manipulated using a finger mouse connected to the computer that controlled the HMD. On deflection of the finger mouse (using the thumb), the rod started to tilt at a slow rate which gradually increased until a pre-set maximum rotation rate. On release of the mouse, rotation of the rod stopped immediately. The tilt angle of the rod could be adjusted with an accuracy of less than one degree. Once the participant felt that the SVV line was upright, the response was entered by means of a button press on the finger mouse (using the index finger). Then a black screen was presented for about 0.5 sec before the next measurement started.

The SVV was computed as the response angle over which the bar was rotated back. Perfect performance would correspond to the bar being rotated over the same angle as the chair tilt (but in opposite direction), which would make the SVV parallel to the gravito-inertial acceleration vector, hence the aircraft’s top-axis. Any additional rotation of the head would induce a rotation of the bar that is larger than the chair tilt. For this reason, we also measured the angle of head roll tilt using the inertial measurements units on the HMD.

2.5 Experimental design

A within-subjects design was used with 12 conditions: three gravity levels (0.25, 0.50, and 0.75 g) and four tilt stimuli (0, 15, 30, and 45 deg). Half of the participants were tilted clockwise, while the other half were tilted counter-clockwise. Each participant performed two trials, both in the same direction, so that 24 parabolas were used for testing per flight. The six remaining parabolas were available back-ups. Per parabola the participants made six settings of the SVV. Altogether this set-up resulted in 12 SVV measurements per condition for participants who completed all parabolas. To compensate for any learning or adaptation effects, the tilt angles were presented in different orders to the participants (using a Latin Square).

The duration of a parabola varied for each gravity level, having the longest duration at 0.75 g and the shortest at 0.25 g. The timespan for a 0.25 g parabola proved adequate to perform six measurements; therefore this number of measurements was used for all gravity levels to prevent rushing and making mistakes due to time pressure. The pace at which the measurements had to be performed was practiced by each participant during an instructional session on the day before the flight.

2.6 Statistical analysis

Four participants successfully completed all parabolas, resulting in a double dataset of 12 SVV settings for each condition. The two back-up participants who replaced the participants who had abandoned the experiment due to nausea, each produced a single dataset of six SVV settings for each condition. Hence, the final dataset included useful data of six participants.

The statistical analysis was performed in the IBM SPSS Statistics software package (version 25). In the analyses, positive values of the SVV always correspond to rotation in an opposite direction of the chair tilt (irrespective of whether the participant was tilted in clockwise, or anti-clockwise direction). We originally planned to use a repeated measures ANOVA with two factors (G-level and chair tilt). However, it was observed that the participants were not able to keep their heads perfectly aligned with their upper bodies, indicating that head tilt was not independent of chair tilt. For this reason, a regression analysis seemed more appropriate with head tilt as a continuous variable (rather than the categorical variable chair tilt). With head tilt we mean the angle between the head and the gravito-inertial acceleration vector (‘head in space’). In the data presentation, head tilt angles in the same direction as seat tilt (either clockwise or anti-clockwise) are shown as positive values.

Although the SVV data was not homoscedastic, indicating that the variance increased with head tilt, this is not considered a problem for the estimation of regression parameters [11]. For each participant we performed three separate regression analyses: one for each G-level (0.25, 0.50, 0.75 g). The estimated regression slopes are comparable to the ‘gain’ determined by Clark and Young [4] in a similar parabolic flight study. As a final step in the analysis, we performed a repeated measures ANOVA on the SVV regression slopes, with G-level as the within-subject factor. As a one-factor ANOVA does not allow for a post-hoc analysis, any differences between the regression slopes found at the three G-levels were tested by means of three separate t-tests. As we had a clear hypothesis that partial gravity results in an underestimation of self-tilt, we used one-tailed t-tests for this. The Mauchly’s test of sphericity indicated that the variances between G-levels can be assumed equal (χ ²(2) = 0.419, p = .811). Consequently, the results did not have to be corrected for sphericity.

3 Results

3.1 G-level and duration

As reported by Lee et al. [21], the mean (±standard deviation) G-levels measured across the 3 flight days were 0.25±0.02 g, 0.50±0.02 g, and 0.75±0.02 g. The G-level during straight-and-level flight preceding each parabola was 1.00±0.05 g. The duration of the reduced gravity periods depended on the gravity level, with mean parabola durations of 22.1±2.3 s, 32.1±1.8 s, and 45.2±4.1 s for 0.25 g, 0.50 g, and 0.75 g, respectively.

3.2 Subjective visual vertical

For each participant separately, the six panels A–F in Fig. 3 show the raw SVV settings, as well as the estimated regression lines, as function of head tilt, measured at the three G-levels. The horizontal scatter in the data clearly shows that head tilt was more variable than seat tilt, which was fixed at 0, 15, 30, and 45 deg. It also appears that for most participants (particularly B, D, E, F) the head tilt angles near 0 deg show a slight positive bias, indicating that their head was not exactly upright when the seat was.

Fig. 3

The panels A –F show the raw SVV settings of each participant, as well as the estimated regression lines, as function of head tilt, measured at the three G-levels. The dotted line (y = x) denotes perfect performance.

The results of the regression analyses underlying the regression lines in Fig. 3 are listed in Table 1, organized per participant and G-level. Regression lines with a smaller, or larger slope than the unity line represent SVV settings towards, or away from the longitudinal body axis, respectively. As can be seen in Fig. 3, three participants (A, B, and E) seemed to systematically underestimate their head tilt at each G-level, while another participant (D) only showed this at 0.25 g. The two remaining participants (C, F) showed a tendency to overestimate the head tilt. We confirmed these observations by testing whether the unity slope fell above, or below the 95% confidence interval of each regression line. The results of this test are included in the last column of Table 1.

Table 1

Summary of results of the regression analyses: df = degrees of freedom; B = unstandardized regression coefficient (slope); 95% CI = 95% confidence interval; SE B = Standard Error of B; β= standardized regression coefficient ^*overestimation of SVV

ID	G-level	Df total	Intercept B	Slope B	Beta	A-effect
			(95% CI)	(95% CI)
A	0.25	47	–1.45 (–2.96 –0.06)	0.26 (0.21 –0.32)	0.81	Yes
	0.50	47	–3.88 (–5.45 - –2.31)	0.46 (0.38 –0.54)	0.87	Yes
	0.75	47	1.25 (0.24 –2.26)	0.37 (0.33 –0.41)	0.93	Yes
B	0.25	23	–2.25 (–4.56 –0.06)	0.48 (0.42 –0.55)	0.96	Yes
	0.50	29	–2.79 (–5.47 - –0.11)	0.79 (0.69 –0.89)	0.95	Yes
	0.75	47	–0.56 (–4.80 –3.68)	0.73 (0.60 –0.86)	0.86	Yes
C	0.25	47	–5.23 (–11.53 –1.07)	1.20 (0.97 –1.43)	0.84	No
	0.50	47	–2.47 (–6.80 –1.86)	1.03 (0.88 –1.18)	0.90	No
	0.75	41	–4.97 (–11.09 –1.15)	1.43 (1.19 –1.68)	0.88	No^*
D	0.25	23	–2.21 (–5.72 –1.31)	0.52 (0.41 –0.64)	0.90	Yes
	0.50	23	–3.23 (–8.26 –1.80)	0.99 (0.83 –1.16)	0.94	No
	0.75	23	–5.61 (–9.42 - –1.80)	1.10 (0.96 –1.24)	0.96	No
E	0.25	47	–4.07 (–7.69 - –0.45)	0.50 (0.39 –0.60)	0.81	Yes
	0.50	47	–5.75 (–8.56 - –2.94)	0.75 (0.67 –0.83)	0.94	Yes
	0.75	47	–2.71 (–5.38 - –0.03)	0.86 (0.76 –0.96)	0.93	Yes
F	0.25	47	2.02 (–3.76 –7.79)	0.82 (0.64 –1.01)	0.80	No
	0.50	47	–1.09 (–3.72 –1.53)	1.01 (0.92 –1.09)	0.96	No
	0.75	47	–0.75 (–2.65 –1.14)	1.23 (1.15 –1.31)	0.98	No^*

The repeated measures ANOVA on the estimated regression slopes showed a main effect for G-level, F(2, 10) = 9.110, p = 0.006. The average slope values (±standard deviation) were 0.95±0.38 at 0.75 g; 0.84±0.22 at 0.5 g; and 0.63±0.33 at 0.25 g. Thus, the average slope decreased with smaller G-levels, indicative of an increasing underestimation of self-tilt. This is illustrated in Fig. 4. According to the paired sample t-tests the average slope at 0.25 g was significantly smaller than the slope at 0.50 g, t(5) = 2.380, p(one-tailed) = 0.032, and also smaller than the slope at 0.75 g, t(5) = 4.858, p = 0.005. The difference between the average slopes at 0.5 g and 0.75 g showed a trend, with the slope at 0.5 g seeming smaller than the slope at 0.75 g, t(5) = 1.540, p(one-tailed) = 0.092.

Fig. 4

This plot shows the mean (and standard deviation) of the regression slopes as function of G-level, together with the slopes measured by Clark & Young (2017) at 0.16 g, 0.38 g, and 1 g. The two horizontal dashed lines indicate the expected slope when the SVV would be aligned with gravity, or the body, respectively.

4 Discussion

The main finding of this study is that self-tilt is being underestimated in partial gravity, indicative of an A-effect. The underestimation increases with decreasing gravity levels. Recently, Clark and Young [4] performed an explorative study with a similar experimental set-up, varying the angle of roll head tilt under different levels of reduced gravity in parabolic flight. Their results suggested an interaction effect between gravity level and head tilt, where the slope of the regression lines found at 0.16 g (0.53) and 0.38 g (0.49) were significantly smaller than the slope found at 1 g (0.80), but not different from each other. However, statistical power was small because data of only one participant were collected. The current study with six participants provided statistical significance for the differences in regression slopes found at 0.75 g (0.95), 0.50 g (0.84), and 0.25 g (0.63). To allow visual comparison between both studies, the slopes reported in [5] are also plotted in Fig. 4.

SVV studies on Earth usually do not show a substantial A-effect over the range of tilt angles considered in this study. Sometimes the SVV is even more subject to the E-effect than the A-effect. Unfortunately, we did not measure the SVV in the 1 g condition, because the period of straight and level flight between parabolas was used for repositioning the tilt device. It would have been valuable to compare the results between partial gravity and normal gravity in our participants during the flight. It is interesting to see that in the 0.75 g condition, i.e., closest to Earth gravity, the regression slopes obtained in two participants (C and F) were higher than 1, indicative of the E-effect. Similar to the other participants, the slopes of these two participants also decreased at smaller gravity levels, so that there was no longer an E-effect at 0.50 g and 0.25 g. Thus the significant differences between the results obtained at the three partial gravity levels confirm that self-tilt becomes progressively underestimated with decreasing gravity level.

For safety reasons, the participants’ head was not firmly coupled to the tilt device, which led to fluctuations of the orientation of the head relative to the trunk. Therefore, the data analysis used the measured angles of head tilt, rather than the angles of chair tilt. This choice is supported by a previously reported observation that the SVV depends more on head tilt than on whole-body tilt [15]. The lack of a firm head fixation may also explain the positive bias in head tilt observed in some participants near the upright position. Because we assigned positive values to head tilt angles into the direction of seat tilt (irrespective whether this was clockwise or anti-clockwise), this bias indicates that during the experiment the participants leaned their head into the direction of seat tilt. Maybe they found comfort by leaning against the cushion, even after the seat returned back to upright. As the head tilt sensor was mounted to the HMD, imprecise placement of the HMD on the head may also have introduced some variability in the measured head tilt. However, one would not expect that misplacements of the HMD would always be in the same direction, so this does not seem to explain the observed bias.

The pressure on the subject’s side in contact with the chair during tilt was a potential source of information for evaluating self-tilt. Psychophysical studies have shown that proprioceptive and somatosensory signals play an integral role in the appreciation of body orientation and configuration [6 , 20]. This role is supported by neurophysiological studies that have shown that the activity of vestibular nucleus neurons integrates information from multiple sensory sources such as the labyrinth, the neck, and the spinal cord [19]. When vestibular information becomes unreliable, supplemental information such as proprioception and haptic information is up-weighted to maintain control of spatial orientation, posture, and locomotion [3, 29]. Therefore, the alterations in the perception of upright seen during partial gravity in our study may have been impacted by alterations in both vestibular information from the otolith organs and somatosensory information from the trunk.

It is interesting to compare the results of our parabolic flight experiment with an experiment performed on Neurolab [8]. They used an on-board centrifuge to produce interaural centrifugation in a microgravity environment. They collected verbal SV responses of roll tilt from four astronauts during constant centrifugation at 0.5 g and 1 g. Although the astronauts were exposed to a longer period of microgravity than our subjects at the time of the experiment, the A-effect was larger in 0.5 g than in 1 g during spaceflight, suggesting that the relative weight of the 0.5 g vector was less than that of the 1 g vector, analogous to the vector model in Fig. 1.

The results of this study might explain why the Apollo astronauts tended to underestimate their self-tilt when on the Moon surface [13]. In fact, the vector model would predict that the A-effect at 0.16 g is larger than the A-effect we observed at 0.25 g. However, as shown in a previous parabolic flight experiment, astronauts may no longer use Lunar gravity as a reference for verticality [9]. Indeed, at 0.16 g the participants in that previous study were no longer able to estimate their self-tilt. Instead, they aligned the SVV with their longitudinal body axis. This result suggests that the vector model should only include an internal gravity vector component when the gravity level is above a perception threshold. Below that threshold, the SVV would be solely dependent on the idiotropic vector. According to the results from [9] this threshold is comprised between 0.16 g and 0.38 g. In another, ground-based centrifuge study found that the centrifuge-induced acceleration had to exceed 0.15 g to affect the perception of upright, which suggests that the threshold value is closer to Lunar than Martian gravity [17].

Footnotes

Acknowledgments

The authors would like to thank Maj. Erik Frijters for assisting as operator during the flight experiments; Ing. Ingmar Stel for programming the SVV task; Dr. Pierre Denise for his help with the ethical committee documentation; and the eight volunteers for their participation in this study. We also thank the European Space Agency and Novespace for their support and their help in accommodating this experiment on board the A-310 Zero-G aircraft. The TNO contribution to this work was funded by the Netherlands Ministry of Defence (Research program V1530 “Flying 2020”).

References

Aubert

, Eine scheinbare bedeutende Drehung von Objecten bei Neigung des Kopfes nach rechts oder links, Archiv für Pathologische Anatomie und Physiologie und für Klinische Medicin 20 (1861), 381–393.

Bischof

, Optic-vestibular orientation to the vertical, In: Handbook of sensory physiology Part 2 Vol. 6, H.H. Kornhuber, ed., Berlin: Springer-Verlag, 1974, 155-190.

Carriot

, Jamali

and Cullen

K.E.

, Rapid adaptation of multisensory integration in vestibular pathways, Frontiers in Systems Neuroscience 9 (2015), 59.

Clark

T.K.

and Young

L.R.

, A case study of human roll tilt perception in hypogravity, Aerospace Medicine and Human Performance 88 (2017), 682–687.

Clemens

I.A.H.

, De Vrijer

, Selen

L.P.J.

, Van Gisbergen

J.A.M.

and Medendorp

W.P.

, Multisensory Processing in Spatial Orientation: An Inverse Probabilistic Approach, Journal of Neuroscience 31(14) (2011), 5365–5377.

Clément

, Arnesen

T.N.

, Olsen

M.H.

and Sylvestre

, Perception of longitudinal body axis in microgravity during parabolic flight, Neurosci Lett 413 (2016), 150–153.

Clément

, Delière

and Migeotte

P.F.

, Perception of verticality and cardiovascular responses during short-radius centrifugation, Journal of Vestibular Research 24 (2014), 1–8.

Clément

, Moore

, Raphan

and Cohen

, Perception of tilt (somatogravic illusion) in response to sustained linear acceleration during space flight, Experimental Brain Research 138 (2001), 410–418.

De Winkel

, Clément

, Groen,

and Werkhoven

, The perception of verticality in lunar and Martian gravity conditions, Neuroscience Letters 529 (2012), 7–11.

10.

Dyde

R.T.

, Jenkin

M.R.

and Harris

L.R.

, The subjective visual vertical and the perceptual upright, Experimental Brain Research 173 (2006), 612–622.

11.

Field

and Discovering

(2018).

, Discovering statistics using IBM SPSS Statistics (5th edition). London: SAGE Publication Lda.

12.

Glasauer

and Mittelstaedt

, Determinants of orientation in microgravity, Acta Astronautica 27 (1992), 1–9.

13.

Godwin

, Apollo 16. The NASA Mission Reports Volume One, Apogee Books, Bulington, Ontario (2002).

14.

Groen

E.L.

, Jenkin

H.L.

and Howard

I.P.

, Perception of self-tilt in a true and illusory vertical plane, Perception 31(12) (2002), 1477–1490.

15.

Guerraz

, Poquin

, Luyat

and Ohlmann

, Head orientation involvement in assessment of the subjective vertical during whole body tilt, Perceptual and Motor Skills 87 (1998), 643–648.

16.

Harris

L.R.

, Jenkin

and Dyde

R.T.

, The perception of upright under lunar gravity, Journal of Gravitational Physiology 19 (2012), 9–16.

17.

Harris

L.R.

, Herpers

, Hofhammer

and Jenkin.

M.R.

, How much gravity is needed to establish the perceptual upright? PLoS One 9 (2014), doi: 10.1371/journal.pone.0106207

18.

Howard.

I.P.

, Human visual orientation. Wiley-Blackwell: New York, 1982.

19.

Jian

B.J.

, Shintani

, Emanuel

B.A.

and Yates

B.J.

, BJ, Convergence of limb, visceral, and vertical semicircular canal or otolith inputs onto vestibular nucleus neurons, Experimental Brain Research 144 (2002), 247–257.

20.

Lackner

J.R.

and DiZio

, Vestibular, proprioceptive, and haptic contributions to spatial orientation, Annual Review of Psychology 56 (2005), 115–147.

21.

Lee

S.M.C.

, Martin

D.S.

, Miller

C.A.

, Scott

J.M.

, Laurie

S.S.

, Macias

B.R.

, Mercaldo

N.D.

, Ploutz-Snyder

and Stenger

M.B.

, Venous and Arterial Responses to Partial Gravity, Frontiers in Physiology 11 (2020), 863.

22.

MacNeilage

P.R.

, Banks

M.S.

, Berger

D.R.

and Bülthoff

H.H.

, A Bayesian model of the disambiguation of gravitoinertial force by visual cues, Experimental Brain Research 179 (2007), 263–290.

23.

Mittelstaedt

, A new solution to the problem of the subjective vertical, Naturwissenschaften 70 (1983), 272–281.

24.

Schöne

, On the role of gravity in human spatial orientation, Aerospace Medicine 35 (1964), 764–772.

25.

Udo de Haes

H.A.

and Schöne

, Interaction between statolith organs and semicircular canals on apparent vertical and nystagmus, Acta Otolaryngologica 69 (1970), 25–31.

26.

Van Beuzekom

A.D.

and Van Gisbergen

J.A.M.

, Properties of the internal representation of gravity inferred from spatial-direction and body-tilt estimates, Journal of Neurophysiology 84 (2000), 11–27.

27.

Vingerhoets

R.A.A.

, Medendorp

W.P.

and Van Gisbergen

J.A.

, Body-tilt and visual verticality perception during multiple cycles of roll rotation, Journal of Neurophysiology 99(5) (2008), 2264–2280.

28.

De Vrijer,

, Mdorp,

W. P.

, & Van Gisbergen,

J. A. M.

, (2009). Accuracy-precision trade-off in visual orientation constancy, Journal of Vision 9(2), 9, 1–15, http://journalofvision.org/9/2/9/, doi:10.1167/9.2.9.

29.

Yates

B.J.

, Jian

B.J.

, Cotter

L.A.

and Cass

S.P.

, Responses of vestibular nucleus neurons to tilt following chronic bilateral removal of vestibular inputs, Experimental Brain Research 130 (2000), 151–158.

30.

Zupan

L.H.

, Merfeld

D.M.

and Darlot

, Using sensory weighting to model the influence of canal, otolith and visual cues on spatial orientation and eye movements, Biological Cybernetics 86(3) (2002), 209–30.