Prediction of Simulator Sickness
in a Virtual Environment

Introduction

Reason and Brand (1975) stated that "Whenever we relinquish our intended status as self-propelled animals and step aboard some vehicle or device that transports us passively we incur the risk of motion sickness" (p. v). In 1953, Crampton and Young demonstrated that motion sickness can also occur when viewing a video display depicting a compelling representation of self-motion in the absence of actual physical motion. In 1957, Havron and Butler were the first to document that a phenomenon similar to motion sickness may be associated with flight simulation systems. Recent research (e.g., Knerr, Lampton, Bliss, Moshell, & Blau, 1993; Regan, 1993) has documented that this phenomenon - known as simulator sickness - can occur in Virtual Environments (VEs) as well and may pose a threat to the ultimate usability of this new technology.

This research focused on the prediction of simulator sickness in a VE. To provide a background for the investigation of simulator sickness in VEs, the phenomenon of simulator sickness in general will first be briefly reviewed. The technology referred to as Virtual Reality (VR) will then be introduced and an overview of this research provided.

The Phenomenon of Simulator Sickness

Simulator sickness is akin to motion sickness except it can occur without actual physical motion. Kennedy and Fowlkes (1992) noted that the phenomenon is properly called a syndrome because of its wide variety of signs and symptoms. Because of the variety of symptoms, Kennedy and Fowlkes describe simulator sickness as being "polysymptomatic". The cardinal signs resemble those of motion sickness: vomiting, nausea, pallor, and cold sweating. Other symptoms include drowsiness, confusion, difficulty concentrating, fullness of head, blurred vision, and eye strain. One measure of simulator sickness - the Simulator Sickness Questionnaire (SSQ) - divides the symptoms into three global categories: nausea, oculomotor discomfort, and disorientation (Kennedy, Lane, Berbaum, & Lilienthal, 1993).

Kennedy and Fowlkes (1992) further describe simulator sickness as being "polygenic" because no single factor has been identified as the cause. Several factors which have been identified by researchers to be related to simulator sickness include factors associated with the individual such as age and gender (Reason & Brand, 1975); factors associated with the simulator such as lag (Frank, Casali, & Wierwille, 1988) and field of view (Kennedy, Lilienthal, Berbaum, Baltzley, & McCauley, 1989); and factors associated with the task performed such as duration (Fowlkes, Kennedy, & Lilienthal, 1987) and degree of control (Casali & Wierwille, 1986).

Along with the potential discomfort to the individual, there are several operational consequences of simulator sickness: decreased simulator use, compromised training, and ground and flight safety (Crowley, 1987). There are also additional effects of simulator exposure: delayed flashbacks and after-effects (a sudden onset of symptoms) (e.g., Baltzley, Kennedy, Berbaum, Lilienthal, & Gower, 1989); shifts in dark focus (the physiological resting position of accommodation) (Fowlkes, Kennedy, Hettinger, & Harm, 1993); eye strain (e.g., Stone, 1993); and performance changes (Kennedy, Fowlkes, & Lilienthal, 1993).

One potentially critical effect of simulator exposure is postural disequilibrium, referred to as ataxia. Baltzley et al. (1989) reported that unsteadiness and ataxia are the greatest immediate concern for safety because there have been reports of such posteffects lasting longer than 6 hours and, in some cases, longer than 12 hours. Although ataxia does not always result, this could be due to the exposure time or sensitivity of the postural test. Clearly, longer-lasting effects, especially those such as flashbacks and ataxia, pose a safety risk to both users of simulators and to others.

A Brief Summary of the Major Theories of Simulator Sickness

First proposed by Claremont (1931), one theory which has evolved to explain simulator sickness is the sensory conflict theory, also known as the sensory rearrangement or neural mismatch theory. This theory proposes that sickness occurs when the pattern of inputs from different senses and within a single sense do not correspond to the stored patterns of such inputs based on past experience. The two primary conflicts thought to be at the root of simulator sickness occur between the visual and vestibular senses (i.e., intersensory conflict) and within the vestibular sense between the canals and otoliths (i.e., intrasensory conflict) (Guedry, 1968). Secondary conflict comes by way of proprioception. Not only do these conflicts lead to sickness (according to the sensory conflict theory), but visual and vestibular adaptation to conflicting cues may also lead to a disruption in balance and coordination, resulting in ataxia (Fregly, 1974).

A second theory of simulator sickness has been proposed by Riccio and Stoffregen (1991). This ecological theory suggests that sickness occurs in situations in which the individual does not possess or has not yet learned strategies for maintaining functionally effective postural control. This theory takes issue with the underlying assumption of the sensory conflict theory that redundancy among/within the visual, vestibular, and proprioceptive senses is expected and that sickness results from a lack of this redundancy. According to Stoffregen and Riccio (1991), such redundancy is not necessarily expected and, therefore, is not a reliable standard for the determination of sensory conflict.

Although the ecological theory is a viable competitor to the sensory conflict theory, the latter currently remains the most widely accepted theory of simulator sickness, most likely because it has enjoyed wide exposure in the literature and appears to be supported by much of the data. This research makes several references to sensory conflict as it relates to sickness. This is not intended to "take sides" on the cause issue but, rather, is simply due to the fact that the sensory conflict theory underlies most of the current literature.

Prediction of Simulator Sickness

Given the potential discomfort to the individual, as well as the operational consequences of sickness, there would be clear benefits to the ability to predict who will become symptomatic in a simulator. If the occurrence or severity of sickness could be successfully predicted based on characteristics of an individual, at-risk users could be identified, properly warned and, perhaps, even trained in some way to reduce their risk. Furthermore, if prediction was based solely on characteristics of the individual, such prediction might generalize to many situations such as different simulators and different tasks performed in the simulator.

Although several characteristics of an individual have been found to be significant predictors of simulator sickness, few characteristics are consistently found to be significant predictors. It is likely that the wide individual differences in susceptibility and great variation in responsivity to simulators make consistent relationships difficult to attain. Although the large individual differences in susceptibility create a large amount of non-error variation which makes statistical prediction of sickness relatively easy, the great variation in responsivity means that different people get sick for different reasons. Thus, a set of individual characteristics which consistently predict sickness for all individuals in all situations may be very hard to find.

Virtual Reality Technology

Although a Virtual Environment (VE) could be defined in many ways, the definition used for this research is that of a three-dimensional, interactive, realistic, real-time, computer-generated simulation providing direct input to the senses via a head-mounted display (HMD), DataGloves™, or similar devices. Note that this definition may be more representative of the goal of VE technology rather than its current state. Today's VEs are typically not fully three-dimensional, interactive, or realistic; they do not run in exact real-time; and they often only provide input to only the visual sense rather than all of the senses. Despite the current limitations, VE - commonly known as Virtual Reality (VR) - technology has many promising applications in areas such as training, molecular modeling, astronomy, architecture, medicine, and entertainment (Durlach & Mavor, 1995). As the technology progresses and the cost decreases, the many potential applications are likely to lead to widespread use of VR technology.

Prediction of Simulator Sickness in Virtual Environments

Unfortunately, as with other simulation environments, simulator sickness can occur in conjunction with VR exposure. Although effects such as nausea and vomiting would be greatly discomforting to users, after-effects such as ataxia pose severe safety risks and, ultimately, raise serious liability questions should accidents occur following VR exposure. As a result, it is imperative that users of VR systems be warned about the potential risks associated with the use of such systems. In order to properly warn users, however, it must be known of what they should be warned and if any users having specific personal characteristics should be given special warnings. This underscores the importance of the ability to predict sickness. If it can be predicted who will experience sickness in VR systems, at-risk individuals can be identified and properly warned.

The literature is silent on prediction of sickness in VR systems. A first step in the investigation of this matter is the identification of related factors. Kolasinski (1995) reviewed the literature and identified 40 factors which may be associated with simulator sickness in virtual environments. These factors fell into three global categories: simulator-related, task-related, and individual-related. Because of the possible contribution of all three categories of factors to sickness, complete prediction would likely involve factors from each. However, as the state of VR technology progresses, factors related to the simulator (i.e., VR system) will change. Furthermore, given the wide variety of applications for VR technology, factors related to the task performed will vary in each situation. This leaves factors related to the individual. Thus, although various factors associated with both the system and task are likely important in the prediction of sickness, for results which generalize over systems and tasks, prediction of sickness will likely have to be based primarily on characteristics of the individual.

An Overview of this Research

In the beginning of this Introduction, the phenomenon of simulator sickness was reviewed. Because of the discomfort potentially associated with it, as well as its possible operational consequences, the ability to predict sickness was noted to have value. Virtual Reality technology was then introduced. The plethora of potential applications - as well as the likely eventual widespread use - makes it especially unfortunate to discover that simulator sickness can also occur in VR systems. The potential consequences of such sickness - particularly with widespread use of VR technology- serves only to underscore the need for successful prediction of sickness. Furthermore, for greatest generalizability without regard to characteristics of the system used or task performed, prediction of sickness in VEs should focus largely on characteristics of the individual.

This research investigated the phenomenon of simulator sickness as it occurs in virtual environments. The primary focus was as a first step in investigating the prediction of sickness in VEs. Because of the changing nature of VR systems and the wide variety of possible tasks, this research centered on prediction of sickness based on characteristics of the individual. In the Review of the Literature, literature concerning individual characteristics which may be important for the prediction of sickness is reviewed. Of the dozen characteristics discussed, four were investigated in this research: age, gender, mental rotation ability, and pre-exposure postural stability. Linear regression techniques - selected because of their wide general applicability - were used to examine the capabilities of these four variables to predict sickness as measured by a standard self-report questionnaire.

As part of the investigation of simulator sickness as it occurs in virtual environments, this research also examined one potentially critical after-effect of exposure, ataxia. Occurrence of ataxia could have disastrous consequences for both users of VR systems and innocent others. Thus, thorough investigation and understanding of this particular after-effect is an absolute must if the technology is to gain widespread use. This research is a first step in that investigation and understanding.

These two areas of investigation formed two separate phases of this research. Thus, each phase had its own research question. Phase I of this research aimed to investigate the prediction of simulator sickness in VEs by addressing the following question: Can predictive techniques be used to model sickness on characteristics of an individual? Phase II of this research aimed to provide insight into the effects of VR exposure by addressing the question: Is VR exposure associated with ataxic decrements in postural stability?

Significance of this Research

Although a wide body of literature exists on simulator sickness, very little exists on sickness as it occurs in VEs. Furthermore, with few exceptions (c.f., Regan & Price, 1994), the majority of VR studies currently reported in the literature were not designed to and did not focus specifically on the investigation of sickness. Instead, most studies investigated the use of VR systems with sickness being examined only as an aside.

It is believed that this study is significant because its sole purpose was the investigation of simulator sickness in a virtual environment. As such, it appears to be the first experiment to experimentally investigate this issue. Thus, it may be an important first step in understanding sickness in VEs.

It is believed that a second significant aspect of this research is that instead of employing the extremely expensive high-end equipment typically used for VR research, it extended the type of simulator investigated to a low-end system, where the term "low-end" is relative to the current state and price of VR technology. As the technology improves and the low-end/high-end spectrum shifts, the Head-Mounted Display (HMD) employed in this research is likely to decrease in price. This HMD, then, should be fairly representative of what will one day be in widespread use.

Finally, it is believed that a third significant aspect of this research is that it investigated ataxic effects of VR exposure. It is important that ataxia, as well as sickness, be investigated because both pose threats to privately-owned as well as public-use VR systems because of the many possible liability issues surrounding widespread use of such systems.

Prediction of Simulator Sickness in a Virtual Environment