Subjects

The study protocol was approved in advance by TNO’s review board (reference: 2019-091). Each subject provided written informed consent before participating. Inclusion criteria for subjects included: being 18–45 yr old; not being pregnant; not being sensitive to motion sickness; and not having any health issues (cardiovascular, neurological or psychiatric, chronic headache or migraine, or vestibular). There were 14 nonpilot subjects (9 women, 5 men) between 19–41 yr old (mean = 24 yr, SD = 6 yr) recruited, and each received a monetary compensation for their time and travel costs. Two subjects did not complete the “post-roll test” phase of the experiment due to complaints about discomfort behind the ears caused by GVS. In the analysis of this phase of the experiment, data from only 12 subjects are used.

Equipment

The study was performed in the Advanced Spatial Disorientation (manufactured by AMST, Ranshofen, Austria) simulator at TNO. Although this is a moving base flight simulator, the simulator was operated in fixed-base mode for most of the experiment, using only the out-the-window visual system and the flight controls. The simulator motion was activated only during the familiarization phase of the experiment.

Electrical vestibular stimulation was delivered using carbon rubber electrodes coated with Spectra 360 electrode gel (Parker Laboratories, Fairfield, NJ, United States). Electrodes were placed over the mastoid processes on either side of the head with tape and a headcap. We used a binaural-bipolar configuration to modulate the firing rate of vestibular afferents in our participants.^20,22,24 In this arrangement, the stimulus evokes a perception of head-roll velocity³⁵ with a craniocentric sensation centered on an axis tilted 17–19° posteriorly and superiorly relative to Reid’s plane.^8,13 This net signal of virtual rotation is estimated based on the responses of afferent populations within the labyrinth, assuming that both canal and otolith afferents respond to the electrical stimulus.^15,24 The expectation of a minimal signal related to translation was based on the near symmetry of otolith afferents across the macular striola of the utricles.⁴¹ To minimize cutaneous cues of GVS, we applied an anesthetizing skin cream [Lidocaine (25 mg ⋅ g⁻¹) and Prilocaine (25 mg ⋅ g⁻¹), TEVA, Haarlem, Netherlands] on the skin 30 min prior to the experimental phase. The stimuli were delivered as analog signals via a data acquisition board (Labjack T4, LJTick-DAC, Lakewood, CO, United States) to an isolated constant current stimulator (STMISOLA, Biopac Systems Inc, Goletta, CA, United States). Using hardware and software measures, the maximum output current was limited to ±5 mA.

The electrical stimulus was coupled to the roll motion of the simulated aircraft. It was designed to mimic the dynamics of the semicircular canals for roll motion, represented by a first-order high-pass filter with a time constant ($\tau$) of 5 s, based on a study by Fernandez and Goldberg.¹² The relationship between current (I) and simulated aircraft roll motion $({\omega }_{ac}$) is given by:$$I=K_i\cdot G\cdot \frac{\tau s}{\tau s+1}\cdot {\omega }_{ac}$$where s is the complex frequency in the s-domain (Laplace transform). G is a gain factor to map roll rates to electrical currents and was set to 5/30 = 0.167 mA/(° ⋅ s⁻¹), based on a maximum current of 5 mA, and simulated aircraft turns with a maximum roll rate of 30° ⋅ s⁻¹. Note, however, that the manually flown maneuvers (see Procedure) could exceed this maximum roll rate, resulting in saturated GVS stimuli. We further note that although more advanced transfer functions replicating the dynamics of canal afferent responses to GVS were available,²⁴ we chose to implement the more straightforward transfer function as the difference between these two were minimal over the movement frequencies tested here. $K_i$ is the gain to explore the effect of stimulus intensity in the motion perception phase as described below (in the test phase, a fixed gain of $K_i$ = 1 was used). The direction of stimulation was such that the assumed sensation is in the same direction as the physical motion of the simulator.

Procedure and Design

After explanation of the experimental procedures, the subject signed an informed consent. Subsequently, the experimenter applied a small amount of anesthetizing skin cream to the left and right mastoid of the subject. In the 20 min needed to fully absorb the cream, the subject was seated in the flight simulator to get familiarized with the flight controls and task. As part of this simulator familiarization period, the subject was exposed to a predefined set of roll maneuvers using the outside visuals and the instruments. The subject also practiced stabilizing the aircraft and maintaining attitude after the operator had ended the roll motion and gave the voice command “you have control.” Throughout this familiarization phase, the simulator’s motion platform was active to allow subjects to experience natural physical motion feedback during simulated flight. The motion cues were generated by the stimulator’s standard built-in motion-cueing algorithm.

Following the simulator familiarization period, the subject stepped out of the simulator and the GVS electrodes were applied on the left and right mastoid. Subjects then returned to the simulator and underwent a second familiarization period to expose them to GVS without administering any flight procedure. This GVS stimulus consisted of a 0.2-Hz sine wave signal that varied in intensity from 2.5–5 mA. When the subject agreed to continue, and no unnecessary discomfort was experienced, the experimental trials were carried out in two successive phases. Note, during these subsequent phases, the simulator remained stationary at all times.

First, in the “motion perception” phase, we investigated the subjects’ motion perception during sinusoidal aircraft roll motions, with a maximum roll rate of 30° ⋅ s⁻¹, a frequency of 0.2 Hz, and a total duration of 10 s (i.e., two cycles). The purpose of this phase was to familiarize subjects with GVS, and to get more insight into the provoked sensations in a more controlled manner than in the next phase. The aircraft was flown manually by the simulator operator, and the participants were asked to judge how well the whole-body motion sensation delivered through GVS matched the visual motion observed in the outside view, a scenery of a typical daytime Dutch landscape. In five separate GVS conditions, the gain $K_i$ of the GVS signal (parameter in the equation above) was varied between 0, 0.5, 0.75, and 1 to investigate the effect of GVS stimulus magnitude on the motion sensation. One condition with a negative GVS gain of −0.5 was added to investigate whether a reversed current would diminish the subject’s perception of how well the GVS-evoked sensations of motion matched the visual cues of motion. All conditions were presented in random order and without simulator motion.

After each GVS condition, the subjects answered the following questions on an 11-point scale (0 = not at all; 10 = completely):

• a)

  How well could you feel the GVS-induced roll sensation? This will be referred to as “motion intensity”.

• b)

  How well did the magnitude of the GVS-induced whole-body roll sensation match the magnitude of the visual motion?

• c)

  How well did the direction of the GVS-induced whole-body roll sensation match the direction of the visual stimulus?

• d)

  How uncomfortable did you find the GVS stimulus?

Next, in the “post-roll test” phase, the subjects were exposed to prolonged roll angle changes with a sudden stop. Because the simulator was stationary during this period and the GVS-coupled stimulus delivers primarily a sensation of head roll, we considered these conditions as analogous to the roll stimulus applied to supine participants in a previous simulator study.³³ Each maneuver was flown manually by an experienced simulator operator, at one of two different magnitudes: a “small roll maneuver” with an intended roll angle change of 40° (starting at 20° roll tilt in one direction to 20° roll tilt in the other direction), or a “large roll maneuver” with an intended roll angle change of 90° (from 45° roll tilt in one direction to 45° roll tilt in the other direction). The simulator operator tried to perform each maneuver in 5 s, so that these two magnitudes theoretically corresponded to average roll rates of 8° ⋅ s⁻¹, and 18° ⋅ s⁻¹, respectively. Both roll magnitudes were presented during leftward and rightward maneuvers, resulting in four test conditions. All test conditions were presented without simulator motion, without instruments, and without outside visual references (as if flying in dense fog). Although the cockpit lights were dimmed, the reflection of the outside visual display made it possible to vaguely see the cockpit interior. Each condition was performed once, and the order of conditions was randomized between subjects. The subjects’ task was to take over the control at the end of each maneuver (upon a voice command after both the small and large roll maneuver), and to keep the aircraft’s orientation constant (that is, a constant roll angle) by moving the control stick left and right. Our hypothesis was that the GVS stimulus would produce an erroneous sensation of roll rate resulting in an overshoot of roll angle when subjects regained control.

Both after the motion perception phase and post-roll test phase, subjects were asked to rate possible discomforts. Furthermore, subjects were instructed to report any symptoms of motion sickness immediately to the experiment leader. Three individuals reported some minor motion sickness symptoms of dizziness. For one of these individuals, the experiment was paused for 15 min after reports of nausea. The complete study took about 105 min for each subject. Subjects performed the study at varying times of the day throughout the morning or afternoon.

We first simulated the predicted outcome by passing an artificial roll signal through the GVS transfer function (upper panel of Fig. 1). In the period prior to t = 0, a subject experiences an imposed motion by the operator, and, after t = 0, the subject takes over control. As shown in the middle panel of Fig. 1, the GVS stimulus peaks at the roll onset but quickly begins to decay. After 5 s of sustained roll the GVS signal is about half the peak response at roll onset, such that stopping the roll maneuver (at t = 0 s) results in a GVS signal of reversed sign, i.e., the post-roll effect. The simulation at t = 1 s represents a control input to counter this GVS-induced post-roll effect. Because this roll input is in the same direction as the preceding roll maneuver, it results in an overshot of bank angle, as is shown in the lower panel of Fig. 1.

View Figure

• Download as Powerpoint
• Download Figure

Statistical Analysis

To analyze the subjective responses in the motion perception phase, Kruskal-Wallis tests (a nonparametric variant of a one-way analysis of variance, because the variables were not normally distributed) and nonparametric post hoc tests were used. We tested whether there was a main effect of GVS gain on the subjects’ perception of whole-body motion and performed post hoc tests (Wilcoxon signed-rank tests, corrected for multiple comparisons using Bonferroni corrections) to compare results between the different gains, $K_i\ne 0$ to $K_i=0$, and to each other. We expected that all subjective responses would be highest at the largest GVS gain. A planned comparison of the negative GVS gain (i.e., $K_i=$ −0.5) was made to the equivalent positive GVS gain (i.e., $K_i$ = 0.5).

The dependent variable in the post-roll test phase, designated “roll response”, was computed as the change in roll angle 3 s (t = 3 s) after the moment at which the subject was given control (t = 0 s), i.e., the roll angle at t = 3 s minus the roll angle at t = 0 s. This analysis window was chosen because most subjects produced a response within this period when given control. Subsequently, this response was categorized as one of three types of behavior: 1) a response in the same direction as the preceding roll motion, corresponding to a post-roll effect; 2) a response in the opposite direction; and 3) a neutral response, with no noticeable response at all or a change in roll angle smaller than 6° (corresponding to a mean roll rate of 2° ⋅ s⁻¹). For both roll maneuvers, the percentage of roll responses within each category was computed.

Motion Perception Phase

We first examined the subjects’ ratings on the four aspects of the motion sensation induced by the GVS stimulus in the motion perception phase (see Fig. 2). The Kruskal-Wallis test showed a main effect of GVS gain on the perceived intensity of the motion sensation ($H\left(4\right)=20.46$, $P<0.001$; see Fig. 2A). The post hoc test revealed that relative to the control condition (i.e., $K_i=0$: median = 2.5), the perceived intensity was significantly larger for only the two high-gain GVS conditions ($K_i=0.75$: median = 6.0, $Z=2.77$, $P=0.014$; and $K_i=1.0$: median = 7.0, $Z=3.22$, $P<0.001$). Furthermore, the perceived intensities of motion sensation across all conditions with GVS (i.e., $K_i>0$) were not significantly different from one another (all $Z<2.20$, $P>0.05$). The effect of GVS gain on the match of perceived whole-body motion magnitude with visual motion was also significant ($H\left(4\right)=16.62$, $P=0.004$). According to the post hoc test, only the two high-gain GVS conditions ($K_i=0.75$: median = 6.5; and $K_i=1.0$: median = 7.5) were significantly different, and, thus, better matched in magnitude to the visual motion when compared to the control condition ($K_i=0$: median = 3.0) ($Z=3.09$, $P=0.031$; $Z=-3.09$, $P=0.003$, respectively). This means that for these two gains ($K_i=0.75$ and $K_i=1.0$), the GVS produces the closest sensation to that of real (visually perceived) motion. However, the comparison of perceived match in magnitude to visual motion across all GVS conditions ($K_i>0$) showed no difference between each other (all $Z<2.05$, $P>0.05$). We further found no significant main effect of GVS gain on the perceived match between the direction of the sensed motion and the direction of the visual motion ($H\left(4\right)=3.75$, $P=0.441$). Surprisingly, when we compared the condition with a negative gain ($K_i=-0.5$) with that of a positive gain ($K_i=0.5$), there were no significant differences in perceived intensity of the motion sensation ($Z=1.52$, $P=0.129$) or the matching of whole-body motion magnitude or direction with the visual motion as compared to equivalent positive GVS gain (magnitude: $Z=1.08$, $P=0.280$; direction: $Z=0.41$, $P=0.682$). Finally, there was a main effect of GVS gain on the discomfort ratings ($H\left(4\right)=34.65$, $P<0.001$), where stronger GVS stimuli gave more discomfort. Similar to all other measures, we found no difference in discomfort ratings between the positive and negative GVS gain conditions ($Z=0.07$, $P=0.942$).

View Figure

• Download as Powerpoint
• Download Figure

Post-Roll Test

Inspection of the simulated aircraft roll motions showed considerable variability in the test maneuvers flown by the simulator operator in the post-roll test phase. On average, the small roll maneuver with an intended roll change of 40° and roll rate of 8° ⋅ s⁻¹ was performed with a roll change of 47.6° (±4.6°), a roll rate of 13.2° ⋅ s⁻¹ (±2.1° ⋅ s⁻¹), and a duration of 3.71 s (±0.68). The large roll maneuver with an intended roll change of 90° and a roll rate of 18° ⋅ s⁻¹ was performed with a mean change of 127.8° (±11.2°), a roll rate of 26.1° ⋅ s⁻¹ (±4.7° ⋅ s⁻¹), and a duration of 5.33 s (±1.00).

Fig. 3 shows two examples of an individual response to a large roll maneuver (a rightward maneuver shown in red and a leftward maneuver shown in blue and inverted for easier comparison). Roll rate, GVS current, and roll angle are shown in separate panels. The moment that the simulator operator handed control over to the subject is represented by t = 0. After this hand-over, the subject had to keep the aircraft’s orientation constant, based solely on the motion sensation induced by the GVS stimulus. One can see that the GVS current slightly decays during the operator-controlled roll maneuver (t < 0 s). At the end of the maneuver, when the subject takes over control (at t = 0 s), the GVS current overshoots (to −2.2 mA and −2.6 mA in these two examples) in the opposite direction. Averaged over all subjects, the GVS level at the onset of the subject taking control amounted to −1.24 mA (±0.14 mA) for the small roll maneuvers and −2.87 mA (±0.53 mA) for the large roll maneuvers. In the blue example in Fig. 3, the stimulus resulted in a continued roll input to the right, which is in accordance with the predicted post-roll effect. In the red example, however, the subject’s response was in the opposite direction from the preceding roll maneuver, meaning opposite the expected direction for a post-roll effect. Furthermore, in 25% of the trials, we also observed that the subject’s steering control led to clipping of the GVS signal to the safety limit of 5 mA, though this was not the case for the two examples in Fig. 3. This clipping primarily took place after longer durations of control input by the subject, outside the analysis window of 3 s.

View Figure

• Download as Powerpoint
• Download Figure

Fig. 4 plots the individual roll responses relative to the GVS current at the moment of control handover for all subjects. Responses with a negative GVS current at onset were rightward maneuvers, while responses with a positive GVS current were leftward maneuvers. The GVS current values corresponding to small roll maneuvers (gray circles) and large roll maneuvers (black squares) cluster around ±1 mA and ±3 mA, respectively, with some scatter on account of the manual operation by the simulator operator. The change in roll angle shows large variability across subjects, especially for the large maneuvers, and the responses are uncorrelated with the GVS stimulus (Pearson correlation for all trials pooled is −0.0163; small roll maneuvers only −0.0403; large roll maneuvers only −0.0146). For the small roll maneuver, only 13% of the responses were in the post-roll direction (responses in upper left and lower right quadrant), 33% were in the opposite direction (responses in upper right or lower left quadrants), and 54% of the responses showed no noticeable response at all (responses in the horizontal band). For the large roll maneuver, these percentages were 33%, 46%, and 21%, respectively. These results indicate that a larger roll maneuver induced a larger number of trials containing a post-roll effect compared to the smaller roll maneuver, but the majority of responses in both maneuvers did not show a post-roll effect.

View Figure

• Download as Powerpoint
• Download Figure