| Publications Page | HITL Home |

The Adaptive Effects of Virtual Interfaces: Vestibulo-Ocular Reflex and Simulator Sickness
by Mark Draper

Appendix G: Classification Scheme To Aid Generalization of Simulator Sickness Results

No one doubts that sickness symptoms can occur in synthetic environments. However, a major drawback of simulator sickness research is the inability of empirical results derived from one environment to reliably generalize to another environment (Hash & Stanney, 1995; Pausch, et al., 1992). While simulator sickness is a complex phenomenon with many complicated interactions, this lack of generalization may be partially due to poor identification and classification of provocative environments.

The majority of researchers who study motion sickness and simulator sickness believe that a main source of sickness are visual-vestibular sensory rearrangements (i.e., the sensory rearrangement theory). Given this assumption, it would be advantageous to classify synthetic environments in terms of their specific visual-vestibular relationships.

This idea of classifying environments according to visual-vestibular coupling is not new. It essentially extends Reason and Brand�s (1975) motion sickness environment classification scheme to synthetic environments. Reason and Brand argued that all motion sickness results from either visual-vestibular rearrangements or canal-otolith rearrangements, each with three possible permutations of the involved variables (see Section 2.4.4.1). Using this classification scheme, Benson (cited in Griffen, 1990) asserted that all simulator sickness is caused only by visual-vestibular rearrangements, especially those in which there are no vestibular motion signals to accompany existing visual motion inputs.

As a further example of classifications, Kennedy has long recognized the importance of classifying flight simulators into such dimensions as fixed-base/motion-base and demonstrating how sickness varies between these different simulators (Kennedy, Drexler, & Berbaum, 1994; Baltzley, et al., 1989). Recently, he has also attempted to identify and quantify provocative visual stimuli at a much more detailed level than ever before, using an visual frame grabber to capture and calculate optic flow velocities and accelerations along several dimensions (Kennedy, Berbaum, Dunlap, & Hettinger, 1996). The argument is that the visual stimulus must be fully defined if there is to be any hope of determining true causes of simulator sickness.

Although this line of inquiry is perfectly acceptable within the domain of fixed-base flight simulators, it represents only one componant of the motion stimuli relevant in head-coupled VEs. To accurately define the motion stimuli in head-coupled VEs, both the visual and vestibular components must be considered, and quite possibly the most important aspect is the interaction between visual and vestibular signals (per the sensory rearrangement theory).

There is often more than one type of visual-vestibular coupling manifest in a single synthetic environment. Current simulator sickness literature frequently fails to recognize this fact. Papers discuss only the most obvious visual-vestibular interaction involved in a synthetic environment. For instance, consider a fixed-base flight simulator which has perfect visual-vestibular coupling during active eye-head gaze shifts but no such coupling during simulated vehicular movement. Therefore, an appropriate classification scheme needs to account for several potential couplings coexisting within a synthetic environment.

A review of the literature makes it clear that classification of synthetic environments in terms of specific visual-vestibular interactions has not yet occurred, which may have contributed to the complexity of the simulator sickness problem. Therefore, a classification scheme is proposed to more completely classify the visual-vestibular interactions of synthetic environments.

Visual-vestibular relationships should be separated into those arising as a result of �head movements� and those arising as a result of �vehicular movements�, because it is through these two types of movements that all visual-vestibular interactions occur (Table 25). These two categories can be further separated into rotational and translational motion.

The �head movement� category entails visual and vestibular motion stimuli in response to active, self-generated movements of the participant�s head (either through neck rotations, torso movements, or natural locomotion). Head movements are assumed to be actively controlled by the subject and would, therefore, always have an appropriate vestibular signal. These motions also include efference copy signals in perceptual-motor control loops.

The �vehicle movement� category applies in those situations where the participant experienced simulated �vehicle� motion (such as in car and flight simulators), either actively controlled or passively experienced. In VR, this vehicular motion would include �flying� through an environment by means of a hand controller button or other such device. Vehicular movement would also include the changing of rotational views by means of an artificial control device.

Each cell in Table 25 has one of the following possible visual-vestibular couplings: synchronized (both visual and vestibular motion signals are assumed to be perfectly coupled as in the real world), congruent (both systems signal congruent motion information but they may not be perfectly coupled due to latencies, tracking inaccuracies, scale factor deviations from 1.0X, etc.), and vest/no vis (vestibular inputs not accompanied by corresponding visual motion inputs). The vehicular movement category also includes a fourth potential coupling: vis/no vest (visual motion signals are not accompanied by corresponding vestibular signals). Furthermore, as control of vehicular motion has been shown to modulate the severity of sickness experienced, there is identification of whether vehicular control task is active or passive for translation as well as for rotational motion.

Figure 103 provides examples of how the visual-vestibular relationship can vary with category of movement (head movement vs. vehicular movement). The top figure pair indicates how head movements (in translation and rotation) are coupled to visual optic flow in a head-coupled environment. Note that head movements result in a visual image motion response. Whether that response is synchronized, coupled, or vest/no vis depends on the specific environment involved.

The remaining figure pairs in Figure 103 show examples of vehicular movement. The middle pair demonstrates vehicular control in a motion-based simulator, where small, transient, externally-applied inertial motion stimuli (represented by the small dot to the right of the head) are provided to the entire subject�s body and to which the visual image is coupled. The visual-vestibular relationship is congruent in this example, as vestibular cues do not perfectly correspond with visual optic flow. The bottom pair represents vehicular movement in a fixed base simulator, where there is no vestibular coupling (internally or externally produced) with optic flow. This results in a vis/no vest relationship. These examples are intended simply to highlight the significant differences in visual-vestibular couplings that can occur in synthetic environments.

Head Movement

Figure 103: Various Visual-Vestibular Couplings

This proposed system specifies the visual-vestibular couplings of all synthetic environments according to the four quadrents of Table 25. Table 26 gives a few examples of how specific synthetic environments would be categorized. Note the differences in visual-vestibular couplings that exist between simulators, between VEs, and between simulators and VEs. This suggests that one should not speak generically of �visual-vestibular conflict� without detailing its specific nature.

Also note how DFOV modulates the potential for head movements to occur in non-head coupled simulators. If the DFOV is small (like a computer display monitor), there will be little if any resulting head motion but if the DFOV is large (like a dome simulator), there likely will be much more head movement (as users look around their environment). So in essence, DFOV may act as a gain modulator for the �head movement� visual-vestibular couplings in non-head-coupled displays, but this modulation may not carry over into head-coupled displays.

Table 1:  Classification of Synthetic Environments

		       Head Movement (A)                             Vehicular Movement


Synthetic environment	Rotation            Translation      	 Rotation                  Translation



1) Fixed-base domed flight sim	sync�d        	 sync�d     	A:vis/no vest    	A:vis/no vest

     (non-head coupled)

2) 6 DOF motion dome flight sim	sync�d	sync�d	A:congruent	A:congruent

3) fixed-base small FOV flt sim	N/A	N/A	A:vis/no vest	A:vis/no vest

4) 3 DOF head-cpld (rot. only) VR	congruent	vest/no vis	N/A	A:vis/no vest

    (gaze direction = translation dir.)

5) 3 DOF head-cpld (rot. only) VR	congruent	vest/no vis	A:vis/no vest	A:vis/no vest

    (hand direction = translation dir.)

6) 6 DOF head-cpld VR	congruent	congruent	N/A	A:vis/no vest

    (gaze direction = translation dir.)

8) 0 DOF small FOV passive	N/A	N/A	P:vis/no vest	P:vis/no vest 

    sim flythrough

9) 3 DOF (rot only) VR passive	congruent	vest/no vis	P:vis/no vest	P:vis/no vest 

   flythrough (active gaze shifts allowed)

In Table 26, fixed-base domed flight simulators involve active motion control and synchronized visual-vestibular couplings during head movements because the visual image always surrounds the pilot and it responds with 0 ms latency, 1.0 gain, and no phase shift to head rotations and translations (as in the real world). However, motion from vehicular control results in a significant conflict between optic flow and an uncorrelated vestibular signal (i.e., vis/no vest). Compare this to a commercially available head-coupled VR system that only responds to head rotations (3 DOF) and where the subject translates through virtual space in the direction he/she looks by virtue of a button press. In this case, although the subject actively controls the motion just like in the flight simulator condition, the underlying visual-vestibular interactions are different. Head movements result in �congruent� but not �synchronized� visual-vestibular interaction (due to latencies, scale factor deviations, etc.) and vehicular motion, only relevant in translation, involves visual motion cues only (i.e., vis/no vest). Also consider that translations only in the direction of current gaze result in more rotational head movements and less asynchronous translational optic flow in the visual stimulus. Thus these two synthetic "simulations" (fixed-base flight simulator and 3 DOF virtual interface) result in very different visual-vestibular couplings, and it should be apparent that generalization of simulator sickness factors/incidence from one to the other may be limited at best. Indeed, researchers have begun to investigate whether sickness symptoms arising from flight simulators are different than symptoms resulting from virtual interfaces (Stanney, Kennedy, & Drexler, 1997).

The purpose of this proposed classification scheme is to provide a rational grounds for generalization of simulator sickness results to other synthetic environments based upon the relative similarity of visual-vestibular couplings involved. A secondary purpose to offer a framework in which to discuss: 1) conflicting results in the literature, and 2) the potential saliency of specific visual-vestibular rearrangements. Given that any complete description of simulator sickness needs to fully include the task, the stimuli, and the individual, this framework is necessarily incomplete. However, for a particular task and individual profile, this classification scheme offers a way to specify and categorize the visual-vestibular couplings in synthetic environments while also specifying the salient task variable of vehicular motion control (active versus passive).

This classification also offers the potential for predictions to be made regarding the relative "potency" of synthetic environments, according to their aggregate of visual-vestibular rearrangements across vehicular and head movements. This presupposes that knowledge is first gathered on the relative saliency of each potential visual-vestibular coupling per movement category. If the task and individuals were equated across two environments to be compared, however, this scheme might at least provide a "suggestion" as to which environment may be more provocative based on some relevant combination of the visual-vestibular rearrangements involved. Nevertheless, the main purpose for developing this classification scheme was to guide the generalizing of simulator sickness results and to better frame discussions of existing research.