Prediction of Simulator Sickness
in a Virtual Environment

[Previous Chapter][Table of Contents][Next Chapter]


Phase I - which investigated the prediction of sickness - and Phase II - which investigated ataxic decrements associated with VR exposure - served as different aspects of the same experiment. In this section, details concerning the research participants, apparatus, and procedure - which were all common to both Phase I and Phase II - are presented first. Then, the two experimental phases are described separately in terms of the independent variables, dependent variables, and design employed.

Research Participants

Research participants were college students recruited from undergraduate psychology and industrial engineering courses. Course credit was offered for participation. Prospective research participants were screened during a phone conversation to determine their eligibility to participate in the research. The screening protocol used - Protocol for Screening Potential Research Participants - appears in Appendix A. In order to be deemed eligible to participate in the research, prospective participants had to indicate that they were 18 years old or older, did not have a personal history of epilepsy, were not color blind, and had not been exposed to any form of virtual environment or virtual reality system within the previous 30 days. In addition, participants were required to have transportation away from the experimental site in which they were not the operator. Research participants were advised to arrive for the experiment in their usual state of fitness, in as much as this was under their control. They were told not to arrive under the influence of alcohol or any drugs, especially sedatives or tranquilizers, decongestants, or anti-histamines. They were warned that their state of fitness, as well as their eligibility to participate (based on the criteria stated above), would be verified again upon their arrival at the experimental site and that if any adverse conditions were noted, they would not be allowed to participate in the research. Furthermore, it was specifically requested that they reschedule in the case of illness.

Data from a total of 46 research participants were collected but only data from 40 participants were analyzed. Except for two participants, the first 40 participants were included in the analyzed data set. It was decided to exclude these two participants from the analyzed data set because of reported pre-exposure symptoms they indicated having and because of medication they indicated having taken. They were replaced by participants of the remaining six of the same gender and from the same type of course (psychology or engineering).

Of the 40 participants included in the analysis, there were equal numbers of males and females. Ten of the males and three of the females were from the industrial engineering course; the rest of the participants were from psychology courses. Because age was a variable of interest in this study, further details concerning the age of participants can be found in the Results section.


Two major pieces of equipment were necessary for this research: a device for measuring postural stability and a system for producing a VE.

Equipment for Measuring Postural Stability

Postural stability was measured using the video-based posture test equipment detailed in Kennedy and Stanney (1996). This equipment consisted of a video camera, a video recording device, a monitor to view the recording as it was taking place, and a reticle which was positioned on the back of the research participant's head. Participants were provided a hard surface to stand on and a piece of tape on the floor indicated where the heel of their back foot was to be placed. The camera was located five feet from the back of the participant's head. A black screen was positioned in front of the participant to provide a dark background for recording purposes.

Equipment for Producing the VE

The VE was produced by a PC-based low-end VR system. The PC used was a Compaq Presario CDS 972, 75 MHz Pentium with 725 MB hard drive, and 8 MB memory. The HMD used was the i*glasses!™ manufactured by Virtual i*O™ (1000 Lenora Street, Suite 600, Seattle, WA 98124).

The i*glasses!™ consist of a head piece with two .7 inch full color LCDs each having a field of view of 30 degrees. Each LCD panel has a resolution of 180,000 pixels. A tracker having a sample rate of 250 Hz is mounted on the back of the unit to monitor head position. The update rate is 60/70 Hz and the entire unit weighs 8 oz.

The stimulus for the experiment was a computer game called "Ascent", produced by Gravity, Inc. for Virtual i*O. This rock jumping game comes bundled with the i*glasses!™ . Ascent was selected because it is relatively easy to learn, uncomplicated, nonviolent, and moderately engaging. Furthermore, the game is such that each research participant received essentially the same stimulus and it could be cycled continuously for the required amount of time.

The control device was a standard mouse. Because only the left mouse button was necessary while playing the game in this study, the right mouse button was disabled.

The basic premise of the game is as follows. Players start the game on a ledge overlooking a canyon with rock walls on both sides and hot lava flowing on the ground below. A path of stones is suspended in the air in front of them. The player's task is to jump from stone to stone and "ascend" to the final stone. When the final stone is reached, players automatically proceed to the next level - not necessarily of increased difficulty - where they begin on a new ledge and repeat the task again. While jumping from stone to stone, players must be careful not to miss any stones or they will fall into the hot lava. They must also be careful to avoid "Tikis", an animated posed threat, which growl at players and push players off stones and ledges. If, at any time, players fall into the lava, they automatically start again at the beginning of the level they were in.

Direction of movement in the game is controlled by a tracker located at the back of the i*glasses!™. A cross-hairs is located in the center of the field-of-view. Thus, players must move their heads to look around the virtual world and aim the cross-hairs where they want to go, naturally creating eye and head movements. When a stone or ledge is within the player's jumping range, it becomes outlined in green. The player then presses the left mouse button to jump. Some stones and ledges are red, indicating that they will disappear in a few seconds, although they can still be jumped to until then. Ascent can be played in either stereo- or monoscopic mode. To take full advantage of the VR technology, stereoscopic mode was used for all research participants.

While engaged in the VR task, research participants were seated in a swivel chair. A table was located to the participant's right and the mouse was placed on this table. Participants were free to move the mouse to wherever on the table was most comfortable for their use. During immersion, participants faced a black screen so that they were forced to look into the HMD rather than at the monitor. Lights in the room were dimmed to reduce glare and reflections within the HMD during immersion.


When participants arrived for the experimental session, their eligibility to participate in the research was evaluated. They were first given the Research Participant Information Questionnaire, a copy of which appears in Appendix A. If participants indicated that they (a) had a history of epilepsy or seizures, (b) were color blind, (c) had been exposed to a VE within the past 30 days, (d) did not have the requisite means of transportation away from the experimental site, (e) were not in their usual state of good fitness, or (f) had taken sedatives or tranquilizers, decongestants, anti-histamines, or other drugs (evaluated on a case-by-case basis) they were not allowed to participate in the research at that time. The participants also indicated their age and gender on this questionnaire.

A pre-exposure SSQ was then administered. The results of this questionnaire were evaluated after its administration on a case-by-case basis in conjunction with the responses on the Research Participant Information Questionnaire to determine if the prospective participant was eligible to participate in the experiment.

As a final part of assessing eligibility, a color blindness test was administered to verify adequate color vision. This test consisted of tests 5 and 6 of the Physicians' Visual Rating Test manufactured by Keystone View. Participants had to be able to identify the numbers in all three circles on each of the plates in order to pass the test.

None of the participants who arrived for experimental sessions were eliminated by any of the criteria described above. This was most likely due to self-selection based on the briefing about the experiment given to the classes in which eligibility requirements were described. In addition, prospective participants had been pre-screened to verify eligibility before they were scheduled for an experimental session.

Once participants were deemed eligible to participate, they were read the Research Pre-Briefing, a copy of which appears in Appendix A, and were provided a hard copy of the briefing so that they could read along to themselves. Following that, they were asked to read and sign the Research Participant Consent Form - a copy of which appears in Appendix A - and were given a carbon copy of this form.

Mental rotation ability was then assessed with the Cube Comparison Test. Next, the participant's height and inter-pupillary distance were measured. Participants then performed the pre-exposure postural stability test. Following the procedure of Kennedy and Stanney (1996), participants assumed a Tandem Romberg stance with eyes closed and arms folded across their chests. They were asked to hold this position for 30 seconds. The trial ended at the end of the 30 seconds or when the participant fell out of position. Two trials of this postural stability test were administered.

Research participants were then prepared for using the VR system. They were instructed on how to use the HMD and the system was checked to be sure it was operating properly. Participants were then instructed on how to play Ascent. These instructions were standardized and a copy appears in Appendix A. Research participants were asked to play the game for 20 minutes but were reminded that they could stop at any time if they felt uncomfortable. The length of time each participant was immersed in the VE was noted. While the participant was immersed in the VE, informal notes were taken by the researcher on such things as room temperature and performance on the game. A copy of the notes form appears in Appendix A.

Following the procedure established by the U.S. Army Research Institute (ARI) for its VR research program, research participants were kept at the experimental site for 30 minutes after exposure. Immediately after exiting the VE, postural stability was re-assessed in the same manner as pre-exposure postural stability. The post-exposure SSQ was then administered. This questionnaire was the same as the pre-exposure SSQ but participants were asked to complete it based upon how they felt when they exited the VE. During administration of this questionnaire, participants were offered a can of ginger ale for refreshment. Next, the Low-End Virtual Reality Assessment Tool - a copy of which appears in Appendix A - was administered to occupy time. This survey was not a formal part of this research and the results are reported elsewhere (Vaden, Ehrlich, & Kolasinski, 1996). To further occupy any time remaining in the 30 minute waiting period, several puzzle-type games were made available for participants to play.

At the end of 30 minutes, participants were asked how they were feeling. If they indicated that they were feeling well enough to leave, the necessary paperwork for course credit was completed and they were asked to read and sign the Debrief Agreement, of which they were given a carbon copy. A copy of this document appears in Appendix A. A third and final postural stability test was then administered before participants were allowed to leave. With the procedure outlined above, the total time for each experimental session was approximately 1 hours.

Because it was expected that some research participants would experience discomfort as part of their participation in this research, arrangements were made with the on-campus Student Health Center in the event that medical attention was necessary. Action taken for dealing with research participants who felt unwell during the course of the experiment was determined on a case-by-case basis. No participants requested medical attention.

During the evening of the experimental session - or as soon as they could be contacted - follow-up data were collected by telephone. Participants were asked how they were feeling and if they experienced any aftereffects which they thought were due to their experience in the VE. They were also provided an opportunity to make comments about whatever they wished. They were requested to call the researcher if they experienced any effects in the near future which they thought were related to their experience. A copy of the protocol for collecting this data appears in Appendix A.

Phase I

Experimental Design

Phase I of this research addressed the question: Can predictive techniques be used to model sickness on characteristics of an individual? Modeling was attempted using linear regression analyses with the goals of the regressions being variable selection and assessment of model performance. Techniques, measures, and diagnostics were chosen to meet those goals. Analyses were conducted using the Minitab™ statistical software package (available from Minitab Inc., 3081 Enterprise Drive, State College, PA 16801-3008).

A pre-experiment power analysis revealed that a sample size of 40 would be adequate for the regression analyses. Details of this power analysis can be found in Appendix B.

Independent Variables

Age and Gender. Age and gender were easily documented by the individual. Descriptive statistics for these two variables specific to this research appear in the Results section.

Mental Rotation Ability. The Cube Comparison Test (test S-2 from the Kit of Factor-Referenced Cognitive Tests, available from Educational Testing Service, Princeton, NJ 08541) provided the measure of mental rotation ability for this research. Because this ability has not been previously investigated in a simulator sickness study using a paper-and-pencil test, there was no precedent for the selection of an appropriate test. The ETS test was selected because it is part of a well-known kit of cognitive tests. Furthermore, being a paper-and-pencil test, it could easily be administered as part of the study.

The Cube Comparison Test consists of two timed parts consisting of 21 items each. Three minutes are allotted for each part. The items are drawings of pairs of cubes with letters on three sides. The task is to determine whether or not the pair represents a drawing of different cubes or if the two cubes may be the same. The score on the test is the number marked correctly minus the number marked incorrectly. Thus, if an individual answers every item on this test correctly, the resulting score is 42. If every item is answered incorrectly, the resulting score is -42. The expected score for an individual guessing at each item is 0. Descriptive statistics for this test for the participants in this research appear in the Results section.

Postural Stability. The fourth independent variable was pre-exposure postural stability. The measure of postural stability proposed for this research was a measure used by Kennedy and Stanney (1996). The technique of Kennedy and Stanney employs Frame Grabber technology and results in 10 measures: velocity, position, and acceleration for each axis of motion (x, y, and z), plus an average of the three velocities. Any of the pre-exposure scores for each of these 10 measures can be used as a measure of pre-exposure postural stability. Because the stance employed is the Tandem Romberg (i.e., heel-to-toe) (Nerlove & Kennedy, 1993), Kennedy and Stanney suggest that the measures for the y direction are the best and that the velocity measures are the most potent. They used velocity in the y-direction as the measure of postural stability and it was originally planned that it would be the measure of pre-exposure postural stability employed in this research as well. However, since the time of their research, a problem was discovered in the way their measures were calculated. As they currently work to correct the situation, they proposed an alternative measure, a measure of postural instability referred to as the Prototype, also introduced in Kennedy and Stanney (1996).

To explain how the Prototype measure is calculated, it should be noted that, for each trial of a postural stability test, several quantities result. First, the time an individual is able to maintain the required position - either until falling out of position or until the end of the trial - is recorded. Second, a video record of the trial is made. This video is then used to calculate the 10 measures discussed above. However, independent observers can be used to obtain ratings of the amount of head sway of an individual during the trial. Such observer ratings were obtained for this research by two trained observers having no knowledge as to experimental details.

The Prototype value proposed by Essex Corporation is a combination of the time for the trial plus the average rating of head sway for the trial using the following formula:

Let P = 10 - trunc[ [TIME+4.99) / 5]

If TIME <= 25 then PROTOTYPE = P

Else if (RATING1 + RATING2) => 8 then PROTOTYPE = P

Else if 6 <= (RATING1 + RATING2) < 8 then PROTOTYPE = P - 1

Else if 5 <= (RATING1 + RATING2) < 6 then PROTOTYPE = P - 2

Else PROTOTYPE = P - 3

where TIME = time for the trial

RATING1 and RATING2 = the observer ratings of head movement by the first and second observers, respectively

PROTOTYPE = value of Prototype measure

Thus, for an individual who is able to maintain the position for 25 seconds or less, the Prototype value is a function of only the time for the trial. For an individual who is able to maintain the position for longer than 25 seconds, the ratings are taken into account. Thus, the Prototype value represents a refinement of the time measure for those individuals who are able to perform the test well. A Prototype value ranges from a theoretical maximum of 10 (for a trial lasting 0 seconds) to a minimum of 1 (for a trial lasting 30 seconds and each rater giving a rating of 1). As a measure of postural instability, lower values of the Prototype reflect better postural stability.

Because each trial resulted in one Prototype value, there were two Prototype values, one for each of the two pre-exposure postural stability trials. The mean of these two values was then computed to give the measure of pre-exposure postural stability used for analysis. Descriptive statistics for the measure of pre-exposure postural stability obtained for the participants in this research appear in the Results section.

Interrater reliabilities for each of the four trials used in this study (two pre-exposure trials and two post-exposure trials) were determined by means of a Pearson correlation. They were also determined for the two pre-exposure and two post-exposure trials overall, for the first and second trial overall, and for all four trials overall. These nine correlations ranged from .837 to .909. The overall interrater reliability was .885.

Other Variables Controlled. It should be noted that, although they were not explicitly investigated in this study, attempts were made to control for several other variables: experience with the task, level of adaptation, and illness. First, due to the nature of the stimulus employed in this research - the computer game, Ascent - experience with the real-world task was assumed to be zero because there is no real analogous real-world task.

Second, although the simulation equipment employed in this research falls at the low-end region of today's VR equipment, the cost is still prohibitive enough to support the assumption that most prospective research participants in this study would not have experience with the specific system used. However, it was recognized that prospective participants could have experience with other VR systems either through using one of the increasingly available public-use entertainment VR systems or through past experience in a VR-related experiment at the University of Central Florida. Instead of specifically investigating simulator experience in this study, it was decided to simply reduce any possible adaptation effects from other VR systems. Prospective participants were not allowed to participate in the research if they indicated having been in a VE within 30 days prior to participating in the experiment, regardless of the VR system used. Thus, as with experience with the real-world task, level of adaptation (i.e., VR experience) was not explicitly investigated in this study but was assumed to be a constant (low to zero) for all research participants.

Finally, during the screening process, prospective research participants were asked to arrive for the session in their usual state of fitness, in as much as it was within their control. Data from research participants suffering from colds or flu or taking certain medications were not included in the analysis. Pre-exposure SSQ data were evaluated on a case-by-case basis. Data from research participants who were highly symptomatic prior to exposure were not included in the analysis. Thus, illness was not explicitly investigated in this study but attempts were made to control for it.

Dependent Variables

The dependent variable of interest in this phase was simulator sickness. Scores obtained from the Simulator Sickness Questionnaire (SSQ) provided the measures of simulator sickness for this research. The SSQ (Kennedy, Lane, et al., 1993) is a 16-item symptom checklist. Each symptom is rated by the individual as either "none", "slight", "moderate", or "severe". There is also a place on the questionnaire for any additional symptoms the individual is experiencing. The ratings from the 16 symptoms are used to compute four scores: subscale scores for the Nausea, Oculomotor Discomfort, and Disorientation subscales as well as an overall Total Severity score. Details concerning the computations of these scores can be found in Kennedy, Lane, et al. (1993). Four separate analyses were performed for each of these measures. Descriptive statistics for each of these four measures for the participants used in this research appear in the Results section.

Phase II

Experimental Design

Phase II of this research addressed the issue of postural stability. Specifically, is VR exposure associated with ataxic decrements in postural stability? A paired t-test was used to answer this question. A pre-experiment power analysis revealed that a sample size of 40 would be more than adequate to detect moderately-sized ataxic decrements due to low-end VR exposure. Details of this power analysis can be found in Appendix B.

The paired values consisted of the pre- and post-exposure postural stability values. Post-exposure postural stability data were collected in the same manner as described in the previous subsection for the pre-exposure data. As with the pre-exposure measure, the post-exposure measure of postural stability consisted of the mean of two Prototype values. Descriptive statistics for the measure of post-exposure postural stability obtained for the participants in this research appear in the Results section.