Sensitivity to Visual Speed Modulation in Head-Mounted Displays Depends on Fixation

Abstract

A primary cause of simulator sickness in head-mounted displays (HMDs) is conflict between the visual scene displayed to the user and the visual scene expected by the brain when the user’s head is in motion. It is useful to measure perceptual sensitivity to visual speed modulation in HMDs because conditions that minimize this sensitivity may prove less likely to elicit simulator sickness. In prior research, we measured sensitivity to visual gain modulation during slow, passive, full-body yaw rotations and observed that sensitivity was reduced when subjects fixated a head-fixed target compared with when they fixated a scene-fixed target. In the current study, we investigated whether this pattern of results persists when (1) movements are faster, active head turns, and (2) visual stimuli are presented on an HMD rather than on a monitor. Subjects wore an Oculus Rift CV1 HMD and viewed a 3D scene of white points on a black background. On each trial, subjects moved their head from a central position to face a 15° eccentric target. During the head movement they fixated a point that was either head-fixed or scene-fixed, depending on condition. They then reported if the visual scene motion was too fast or too slow. Visual speed on subsequent trials was modulated according to a staircase procedure to find the speed increment that was just noticeable. Sensitivity to speed modulation during active head movement was reduced during head-fixed fixation, similar to what we observed during passive whole-body rotation. We conclude that fixation of a head-fixed target is an effective way to reduce sensitivity to visual speed modulation in HMDs, and may also be an effective strategy to reduce susceptibility to simulator sickness.

1. Introduction

Head-mounted displays work by presenting a rendered view of a virtual environment that is updated based on the users head movement. Consequently, when the user turns the head, optic flow is presented on the HMD that is consistent with the users head movement. Disagreement between the head movement and the visual motion that is rendered is the most widely accepted explanation for the initiation of simulator sickness symptoms [1, 2, 3, 4, 5, 6, 7, 8]. However, user tolerance for this disagreement has not been extensively studied [9, 10]

In prior work, we evaluated this tolerance by introducing conflicts between the physical head motion and the visual scene motion and measuring participants’ ability to detect these conflicts [11]. We found that sensitivity to conflict depended on how participants moved their eyes, with the best sensitivity observed when participants moved their eyes to track scene-fixed targets. Head motion in these experiments was generated through passive full-body rotation with participants seated on a moving platform and with visual stimuli presented on a display mounted to the platform. Here we examine whether our previous findings generalize to the most common VR use-case, that is active turning of the head relative to the body with visual stimuli presented on an HMD.

Active and passive head movements differ in several important respects. Perhaps most importantly, during active head movements, additional non-visual cues to head motion are available, including proprioception and efference copies of motor commands. Head movements are also more variable from trial to trial; self-generated head movements lack the absolute consistency of velocity and duration afforded by the motion platform.

Using active head movements, we therefore address a slightly different set of questions in the current study. Similar to the previous study, we aimed to (1) measure sensitivity to conflict, (2) measure sensitivity to the visual optic flow stimulus, and (3) measure how these depend on fixation. However, in contrast with the previous study, we were not able to measure sensitivity to non-visual head motion cues during active movement, since we have not yet devised a method that would allow us to make this measurement. Another difference from the previous study is that speed modulation is achieved by adding or subtracting a constant amount relative to the head-specified visual speed. In other words, the speed manipulation is additive rather than multiplicative. Also, finally by varying the medium of display relative to the previous study (HMD vs. 3D monitor) we are able to validate the usefulness of the HMD as a tool in psychophysical experimentation. We also sought to investigate whether an individual’s sensitivity to conflict is predictive of their susceptibility to motion sickness, paying particular attention to possible effects of gender. While some studies claim that females are three times more susceptible [12,9,13], many others fail to replicate this finding [14,15]. In this study, participant numbers were balanced with respect to gender allowing us to investigate whether gender is predictive of sensitivity to conflict, susceptibility to simulator sickness, or both.

Overall, the experimental design of the current study allows us to test the below hypotheses related to active head movement (Table 1).

Table 1:

Hypotheses

H1 : Conflict detection is improved during scene-fixed compared with head-fixed fixation.

H2 : Fixation does not influence performance in a visual speed-discrimination task.

H3 : Females exhibit more VR sickness than males.

H4 : VR sickness ratings and conflict sensitivity are correlated.

2. Methods

2.1. Participants

Nineteen healthy participants (nine F, ten M), ranging in age from 20 to 41 (mean age = 26.5 years) completed the study. All possessed normal, or corrected-to-normal vision, unrestricted head/neck movement, and had no history of visual, or vestibular sensory disorders. All but two were naive to the hypothesis being tested; results of these subjects did not differ significantly from those of the other subjects. Three further participants began the study. Two were excluded due to an inability to consistently execute the trials, one dropped out due to time demands. The subjects in the current study were not the same as those that participated in our previously published study [11]. All procedures were approved by the Institutional Review Board of the University of Nevada, Reno and all participants provided informed consent.

2.2. Equipment

The experiment was conducted using an Oculus Rift (CV1) head mounted display (HMD) and Oculus-ready Alienware PC with NVIDIA GeForce GTX 1060 video card. Latency of this VR system was measured using the built in Oculus latency tester. Average latency measurements showed Flip to Mid-Photon as 7.23ms, Timewarp to Mid-Photon as 10.19ms, and App Tracking to Mid-Photon as 13.84ms. Participants were seated in a fixed, high-backed chair, to ensure head rotations originated from the neck while avoiding postural rotation. Participant responses were input using a standard keyboard. Textured Velcro tape was attached to the response keys to allow them to be identified haptically. All conditions were conducted in a quiet darkened room. Headphones in the HMD enabled delivery of auditory beeps, tones and instructions that helped orient the subjects in the experiment. The virtual environment was programmed using C# within the Unity programming environment. Visual stimuli were rendered with a refresh rate of 90 Hz. Participants altered the HMDs interpupillary distance (IPD) themselves to a comfortable setting to account for individual differences in interpupillary distance. However, the IPD setting was not checked or recorded. We did not have the capability to record eye movements inside the HMD in order to verify that subjects were accurately following the fixation instructions. However, in a previous, similar study [11], we did record eye movements in a subset of subjects, and the data show that subjects successfully followed instructions (see Fig. S4 in [11]).

2.3. Conflict Detection Task

During normal use of an HMD, the view on the visual environment is updated based on head movement, resulting in movement of the visual scene that is equal and opposite the head motion, consistent with a stationary, earth-fixed environment. However, in this task, the visual speed was modified to be either faster or slower than the head movement. Visual speed was increased or decreased by a constant amount so that it appeared to be drifting with or against the head movement. In other words, the speed manipulation was additive, not multiplicative. Participants were asked to judge if the visual motion was too fast or too slow compared to their head movement. In other words, they were asked to detect the conflict between visual motion and head motion. The goal of the experiment was to measure the threshold for detection of this conflict.

The exact timecourse of an individual trial is illustrated in Fig. 1. Participants initiated each trial by adjusting their head position to the designated central position. This was achieved by making small head movements to align a head-fixed target (rendered 300cm in front of the cyclopean eye) with a central, scene-fixed target in an otherwise black, featureless environment. Once aligned, one of the red targets vanished, and a yellow target flashed (0.1s duration) within the participants left or right near-peripheral vision at an eccentricity of 15°. Then a randomly generated 3D starfield appeared and the participant performed a head rotation to point the head to where the yellow rotation target had flashed. The starfield consisted of 8000 randomly distributed white spheres (radius = 33cm), at a minimum distance of 1000cm, and average distance of 5000cm from the participants position. After the head rotation, the starfield disappeared, and an audible beep and text on the screen prompted the participant to indicate via keypress if they perceived the visual scene as moving with or against their head motion. If the visual scene moves too slowly, it is perceived to move with the head movement, and if it moves to quickly, it is perceived to move against the head movement. After the response, the next trial was initiated.

Figure 1.

Experimental set up. Tasks focused on either visual speed discrimination (A,B,C) or conflict detection (D,E,F). A) illustrates the timecourse for the visual speed discrimination task. Participants focused on a red fixation target in the center of the scene and judged which of two yaw rotations of the background scene was faster (2AFC). Rotations were separated by a 0.5s inter-stimulus interval. B) and C) highlight the different conditions for this task. In B) participants eyes were fixed in their head while the visual stimulus moved across the retina. In C), the fixation target moved with the scene movement and thus by focusing on the target the image was stable on the retina. D) illustrates the timecourse for the conflict detection task. To ensure a consistent starting point, participants initiated each trial by aligning a red fixation target fixed in the scene with one attached to their virtual head. Once the targets were aligned, a yellow rotation target flashed (0.1s) at 15° eccentricity and participants rotated their heads to point towards this target. The visual speed was modulated to create conflict between the physical motion and displayed motion of the visual scene. Then participants answered whether the visual scene had moved too slowly or too quickly, i.e. with or against the direction of head rotation in world coordinates. E) and F) highlight the different conditions. In E), the fixation point moved with the head, so participants kept their eyes fixed in their heads. In F), the fixation point stayed fixed with respect to the scene, so participants had to counterrotate their eyes to maintain fixation.

Because head movements were actively generated, care was taken to ensure consistency of head movements across trials. Specifically, the duration and speed of head movement was monitored and if duration was too short or long (0.5s ≤ dur ≤ 2.78s), or speed was too slow or fast (5.4°/s ≤ speed ≤ 20°/s), the trial was rejected and the participant received feedback in the form of a metal clang sound and a warning corresponding to the nature of the fault (e.g. too fast, or too slow). Parameters for acceptable head movements were identified in a pre-experiment pilot study. The movement profile was chosen as one that felt natural and was easy enough to repeat trial after trial. Example head movements from this pilot study along with exclusion criteria are illustrated in Figure 2

Figure 2.

Example head movement traces. Thin colored lines indicate individual head movement traces. The thick blue line indicates the average over all of the individual traces. The dashed horizontal lines indicate the exclusion criteria for minimum and maximum average head velocities. The vertical solid red lines indicate the exclusion criteria for minimum and maximum duration of head movement.

When a trial was rejected, to maintain consistency and encourage a smooth flow through the experiment, the starfield disappeared as before, and the participants were still presented with the same forced choice task. While responses on these trials were recorded, they did not affect the adaptive psychophysical procedure and were not used to calculate thresholds. There was an average of 7.8 unacceptable trials per block.

Each experimental block consisted of 150 trials with acceptable head movement profiles. The stimulus for a given trial was generated using two interleaved adaptive staircases (2up1down, 1up2down) of 75 trials each (Fig.3). New staircases were initiated in subsequent blocks.

Figure 3.

Psychophysics. Stimuli presented to a single participant in the conflict condition were modulated adaptively based on previous responses and a cumulative Gaussian psychometric function was fit to responses. A) illustrates two interleaved adaptive staircases (2 up 1 down and 2 down 1 up) which made up each block of 150 trials. B) illustrates how participants answers of with or against were used to fit a psychometric function. The parameters of interest are the mean or position of the curve (i.e. the point of subjective equality), and the standard deviation or steepness of the curve (i.e. the just noticeable difference).

There were two conditions, and participants completed three blocks for a total of 450 trials in each. The conditions were distinguished based on fixation behavior during the active head turn. In the head-fixed fixation condition, after head- and scene-fixed targets were aligned (Fig. 1), the scene-fixed target disappeared, and participants were left to fixate the target that moved with the head (Fig. 1). This behavior maximized retinal image motion (optic flow), while minimizing eye movement. In the scene-fixed fixation condition, the head-fixed target disappeared, and participants were left to fixate the scene-fixed target during the head movement (Fig. 1). This behavior maximized eye movement, while minimizing optic flow.

2.4. Visual Speed-discrimination Task

In this task, participants kept their heads still and were presented with full-field visual motion similar to the optic flow presented during the conflict detection task. Two consecutive motion intervals of 1.0 s were presented on each trial. One motion interval was the standard stimulus with bell-shaped velocity profile and a peak velocity of 29.5°/ s. The other interval was the comparison stimulus which also followed a bell-shaped velocity profile but with a peak-velocity that was incrementally faster or slower than the standard. This speed increment was varied from trial to trial. After both intervals were presented. Participants responded which interval contained the faster movement (Fig. 1). Motions were in the same direction in both intervals, and participants completed 150 trails in each block. There were two conditions, head-fixed and scene-fixed, similar to those described above.

2.5. Training, order of experiments, and simulator sickness ratings

Experimentation was split over five sessions, each separated by 24 hours or more. These sessions always began with a thorough training session to ensure that participants understood the task. The training protocol explained and demonstrated each element of the required movement step-by-step. Participants performed each component part numerous times, before they finally executed 20 practice trials. The training was implemented to elicit consistent head-turning behavior within and across participants; training is often required when movements are generated actively rather than presented passively [16, 17]. After the training, participants completed a maximum of two experimental blocks of 150 trials each per session. All blocks for a given condition were completed in sequence, but the order of conditions was counterbalanced across subjects. To encourage focus/attentiveness within each session, mandatory breaks of 30 seconds were implemented after 50 and 100 trials. Longer breaks of 4 - 5 minutes were enforced between blocks. During this time participants removed the HMD and the light was turned on. Participants sat comfortably, stretched, or walked around before beginning the next block of trials. Each session was performed with an average running time of 25 minutes.

At the end of each condition, subjects provided a single rating of their feeling of simulator sickness on a 4-point scale, including ratings of none (0), slight (1), moderate (2), and severe (3). This scale is identical to that used in the more extensive Simulator Sickness Questionnaire (SSQ) [6] which asks participants to rate the severity of several specific symptoms.

2.6. Statistical Analysis

Analyses were conducted using Matlab R2016a together with the Palamedes toolbox package developed by Kingdom & Prins [18]. PAL PFML (Palamedes psychometric function: maximum likelihood) functions enabled us to fit a cumulative Gaussian to a participants response data (Fig.3). Parameters for lapse rate λ and guess rate γ were both set to zero. We refer to the mean parameter of the fit as the point of subjective equality (PSE), and standard deviation parameter as the just noticeable difference (JND).

Paired t-tests were carried out using participants JNDs across conditions. We additionally conducted one-sample t-tests to examine whether PSE values differed significantly from a speed increment of zero, i.e. the point where the visual stimuli has not been manipulated and matches the physical head motion. We analyzed VR sickness ratings using rank-based Wilcoxon tests to examine possible effects of Task (Conflict, Visual), Fixation (Head-fixed, Scene-fixed) and Gender (Male, Female). Additionally, the correlation between threshold and sickness ratings was analyzed using Spearmans rank correlation.

3. Results

3.1. Conflict Detection Task

Psychometric fits to the data from each individual subject and condition (e.g. Fig. 3B) provide a measure of the visual speed increment that is perceived to match the physical head motion (the PSE) as well as the threshold for the speed increment that leads to detection of conflict (the JND). Thresholds in the head-fixed condition (mean=2.88; SD=1.89) were significantly larger (t=2.828, p=0.011) than those in the scene-fixed condition (mean=1.98; SD=0.81), in support of hypothesis H1. These thresholds are plotted in Fig. 4A. The slope of the blue line is the log-average (i.e. geometric mean) of the ratio of the scene-fixed versus head-fixed JND (1.33) and the shaded area represents the standard error of this ratio (0.11). For comparison we have also plotted the thresholds measured in our previous study with passive head movements (Fig. 4C). As in the current study, thresholds in the head-fixed condition were larger than those in the scene-fixed condition. Again, the slope of the blue line is the log-average of the ratio of the scene-fixed versus head-fixed JND (1.27) and the shaded area represents the standard error of this ratio (0.22). The relationship between head-fixed and scene-fixed thresholds appears to be independent of whether head movement was active or passive.

Figure 4.

Results. Plots compare performance between Scene-fixed and Head-fixed fixation in the current study (top row) and the previous study (bottom row) in the Conflict condition (left column) and the Visual condition (right column). (A) Demonstrates that conflict detection was better (JNDs were lower) during scene-fixed compared to head-fixed fixation, a pattern that agrees with results of our previous study, shown in panel (C). (B) shows that the just-noticeable gain increment in the head-fixed and scene-fixed visual conditions in the current study were comparable and significantly correlated. Result of our previous study, shown in (D), show that performance was better during head-fixed compared to scene-fixed fixation. Thus fixation dependent changes in visual variability (B,D) cannot explain fixation dependent changes in conflict detection (A,D).

It is tempting to compare JND values observed in the present (Fig. 4A) and previous (fFig. 4C) studies. However, there are many methodological differences between the two studies, which makes comparison difficult. We expand on these differences in the discussion section.

The speed increment that resulted in visual speed perceived as matching head motion was close to zero in both the head-fixed condition (mean=−0.29; SD=0.53) and the scene-fixed condition (mean=−0.15; SD=0.35).

3.2. Visual speed-discrimination task

In addition to measuring visual-vestibular conflict detection, we also measured visual speed discrimination thresholds. These thresholds were measured to examine whether the effect of fixation behavior on conflict detection could simply reflect an effect of fixation on visual speed estimation. For example, more variable visual speed estimation during head-fixed fixation could explain higher conflict detection thresholds. However, results do not support this hypothesis. Discrimination thresholds were approximately equal during head-fixed (mean=7.20, SD=3.51) and scene-fixed (mean=7.05, SD=3.22) fixation (t=0.297, p=0.770), in support of hypothesis H2, so differences in visual variability alone cannot explain the observed dependence of conflict detection on fixation. Thresholds in the two fixation conditions were not only approximately equal, they were also significantly correlated (r=0.846, p< 0.001; Fig. 4B), likely reflecting their dependence on similar underlying visual motion processing mechanisms.

For comparison, we also examined visual speed discrimination thresholds from our prior study which used much slower movements (Fig. 4D). Again, thresholds were not significantly different during scene-fixed (mean=2.49, SD=0.4) compared with head-fixed (mean=2.0, SD=0.71) fixation (t=2.0541, p=0.070). Thresholds were not significantly correlated (r=0.177, p=0.626). The magnitude of the thresholds were comparable between the current (Fig. 4B) and previous study (Fig. 4D), unlike during the Conflict detection condition (Fig. 4A vs. Fig. 4C) where thresholds were found to be much lower in the current study.

3.3. Simulator Sickness Questionnaire

At the end of each condition, participants rated their feeling of simulator sickness on a 4-point scale, similar to that used in the more comprehensive SSQ [6]. A summary of all responses is shown in Figure 5. Responses were not obtained from 2 subjects (1M, 1F). We hypothesized that greater sensitivity to conflict in the Conflict Detection task would be associated with more severe sickness ratings, but this association was not present in the data (Spearman rank correlation r=0.15, p=0.35; see also Fig. 5A). This does not support hypothesis H4. Surprisingly, we observed a significant correlation between performance in the visual speed-discrimination task and sickness with less sensitivity associated with more severe sickness (Spearman rank correlation r=0.49, p=0.002; see also Fig. 5B). Possible reasons for this association are explored in the discussion section.

Figure 5.

Sickness ratings by condition and gender. Head-fixed indicated by blue, and scene-fixed by red. Males indicated by triangles, females by circles. A) Ratings of sickness following the Conflict Detection conditions. B) Ratings of sickness following the Visual Speed Discrimination conditions.

Sickness ratings were additionally analyzed to examine the effects of Task (Conflict, Visual), Fixation (Head-fixed, Scene-fixed) and Gender (Male, Female). There was a significant effect of Task (Wilcoxon signrank test, p<0.001), with more adverse responses reported during the visual discrimination task. This makes sense because in this task, visual optic flow consistent with head motion was presented to stationary observers. We also observed a significant effect of Fixation (Wilcoxon signrank test, p=0.02), with more adverse responses reported during Scene-fixed fixation. Conflict detection thresholds were also lower during scene-fixed fixation; this association is consistent with the hypothesis that conflict detection mechanisms that underlie perceptual reports may be the same as those that can ultimately elicit simulator sickness. Finally, there was a significant effect of Gender (Wilcoxon ranksum test, p<0.001) with more sickness reported by female participants. This is consistent with gender-dependent effects reported previously, but the cause for these effects remains unknown.

4. Discussion

In the current study, we examined sensitivity to visual speed modulation during active head-on-body rotation using an HMD. This work builds on our prior study [11] in which we measured this sensitivity during slow, passive full-body rotation. We found that conflict detection depends similarly on fixation behavior, regardless of whether head movements were actively or passively generated. This effect of fixation cannot be explained based on the variability of the visual self-motion estimates because visual discrimination thresholds do not depend similarly on fixation behavior. Instead, we hypothesize that cue comparison mechanisms operate differently depending on oculomotor behavior, regardless of whether movements are active or passive. We also found that conflict detection thresholds were an order of magnitude lower during the current study compared to the previous study. Possible explanations for this large difference in thresholds are discussed below along with a discussion of the relation between conflict sensitivity and simulator sickness.

4.1. Visual-vestibular conflict detection during active head movements

While the experimental design and procedure used to measure conflict detection in the current study was very similar to our prior study, there were several important methodological differences. First, head movements were generated actively in the current study via yaw head turns, whereas the whole body was passively rotated in the previous study. Second, the speed of rotation was much slower in the previous study ( 10°/s peak vel.) compared to the current study ( 30°/ s peak vel.). Third, manipulation of visual speed was additive in the current study (i.e. speed increased/decreased by a constant amount throughout the trial) versus multiplicative in the previous study (i.e. speed increase/decrease proportional to the instantaneous head velocity). Fourth, the visual stimuli were presented on an HMD in the current study, compared with a 3D TV in the previous study; the HMD moved with the head while the screen moved with the platform. Finally, the characteristics of the visual scene were different. Both scenes were composed of a volume of random dots, but the volume, size, density and color of the dots differed.

Despite these differences, the effect of fixation was similar across studies (Fig. 4A, C). Conflict detection is facilitated when fixating a scene-fixed point compared with fixating a head-fixed point, but the reason for this scene-fixed advantage is unclear. We speculate that during scene-fixed fixation, natural gaze stabilization behaviors, including the vestibulo-ocular reflex (VOR), are allowed to operate more or less normally. Increased sensitivity to conflict under these conditions may reflect perceptual access to error signals, in the form of low-velocity retinal slip, that are normally used to drive calibration of the VOR. During head-fixed fixation, on the other hand, the VOR is suppressed resulting in high-velocity retinal slip as the eye moves relative to the scene. Conflict detection in this case depends on comparison of retinal slip and vestibular velocity, and this comparison process appears to operate less efficiently, perhaps because it is not integral to everyday reflexive stabilization behaviors [22].

While it is reasonable to compare ratios of scene-fixed to head-fixed JNDs across studies, direct comparison of JND magnitude across studies is less informative because of the many methodological differences. Generally, both head-fixed and scene-fixed JNDs were slightly lower in the current study (Fig. 4A) compared to the previous study (Fig. 4C). This advantage could be related to the speed manipulation, which was additive rather than multiplicative in the current study. An additive factor could be more perceivable at the beginning and end of a movement, when head velocity is slow.

Alternatively, lower JNDs in the current study may be a consequence of active head movement, which would allow generating a prediction of the visual scene motion based on efference copy signals [23]. Intuitively, extra information flows should reduce uncertainty and allow for a more accurate assessment of whether any conflict was present. Indeed, several previous studies [24-28] have shown greater precision of responses when all modalities (vestibular, proprioception, efference copy) signaled a rotation [17].

In addition to the role of active head movement, the difference in performance between current and prior studies could also be due to other methodological differences including speed of head movement ( 10°/s vs. 30°/s peak vel.), display type (HMD vs. screen), and differences in the visual scenes. Previous studies on the detection of latency in HMDs have revealed significant improvements in sensitivity when peak velocity of rotation is increased [29,30]. Additional research is needed in future using identical methods (i.e. same procedure, display, and scene) to investigate conflict detection during identical active and passive head movements (i.e. same speed and trajectory of head motion with multiplicative manipulation of visual speed).

4.2. Visual speed perceived as matching

Previous work, including our own, has generally demonstrated that the visual gain necessary to match a physical head rotation is significantly greater than 1 [19-21]. In this study, we instead observed that speed increments very close to zero were perceived as matching. This disparity can likely be explained by methodological differences. In particular, the speed manipulation in the present study was additive rather than multiplicative. We suspect that using an additive factor allowed for greater sensitivity to conflict at the beginning and end of the movement, when the head was moving slowly. This could have led to better detection of a mismatch in the magnitude of visual and nonvisual head-motion signals. Alternatively, the visual-vestibular comparison process may operate differently during passive compared to active head movement. Direction and speed of head movement may also play a role. Examining the previous literature in greater detail we observe that measurement of visual gain perceived as matching during yaw, rather than roll, pitch or linear head movement, tends to result in matching gains that deviate less from gain of unity [19,21]. Size and depth cues also influence self-motion perception [31]. Previous studies have shown that manipulating size and distance attributes of the visual scene can effect the matching visual gain [32-34]. Due to these many methodological variations, prior studies often report that a wide spread of gains (0.8–1.4) are accepted as stable [19].

4.3. Sickness and relation to conflict detection

Perceptual measures of conflict detection, like those presented here, may prove to be valuable predictors of susceptibility to motion sickness. Stimulus conditions that lead to improved conflict detection may also be those that lead to increased incidence of simulator sickness. Likewise, individuals who are more sensitive to conflict may be those that are more likely to experience simulator sickness.

In addition to measurements of perceptual sensitivity, we obtained from each participant a 4-point sickness rating after each condition. Participants reported greater discomfort during scene-fixed compared to head-fixed fixation (Fig. 5, red vs. blue). Perceptual sensitivity to conflict was also greater (i.e. JNDs were lower) during scene-fixed compared to head-fixed fixation (Fig. 4A). Thus, in support of hypothesis H₄, greater perceptual sensitivity to conflict was associated with greater likelihood of discomfort in response to conflict. Despite this association, the correlation between conflict JND and sickness ratings (Fig. 5A) was not significant. This non-significant result may be the consequence of insufficient statistical power. There was little sickness reported (only ratings of none or slight reported). Future studies may seek to drive increased levels of sickness, or use a finer scale of reporting, in order to observe a significant correlation.

We observed significantly more severe sickness reports in the visual speed-discrimination task than in conflict detection (Fig 5A vs 5B). This is perhaps not surprising because visually-simulated self-motion in the absence of physical motion is known to be conducive to simulator sickness [35, 36, 37]. The reduced sickness in the conflict detection task is likely because non-visual cues to self-motion were approximately in agreement with the visual cues. The only difference was due to the manipulation of visual gain. The conflict detection task included active head rotations. Motor activation may mitigate the effects of conflict due to agency or control [38]. Previous studies have shown how postural instability [39, 40] and control and the resulting expectations [41, 38] mediate motion sickness. An active participant necessarily has greater control which may account for the reduction in symptoms.

Interestingly, in the visual speed-discrimination task, sickness rating and JND were positively correlated (Fig. 5B); greater sickness was reported by those subjects who were least sensitive. This result appears to contradict hypothesis H₄, but only if the following assumptions are valid: (1) performance on the visual discrimination task is an accurate measure of noise on the visual estimate, and (2) conflict detection is limited by this noise (i.e. signal detection model of conflict detection [42]). Alternatively, it may be that participants who reported greater levels of sickness were simply less able to concentrate on the task at hand. This association deserves to be investigated more deeply in future experiments.

Finally, we observed an effect of gender of similar to previous studies, with more severe sickness reported by females than males (Fig. 5, triangles vs circles). With so many contradictory findings in existence, this additional result does little to swing the balance of evidence either way. Far more work is necessary in this arena before any definite conclusions can be made.