CYBERSICKNESS PRIORITIZATION AND MODELING
By
Lisa Renee Rebenitsch

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Computer Science - Doctor of Philosophy
2015

ABSTRACT
CYBERSICKNESS PRIORITIZATION AND MODELING
By
Lisa Renee Rebenitsch
Motion sickness-like symptoms due to visual stimuli, or cybersickness, are a topic of significant
public concern. From three-dimensional movies to new head mounted displays such as the
Oculus Rift, the general public is being offered ever-increasing opportunities for immersive
experiences. However, virtual environment opportunities are often accompanied with
increased reports of “movie theatre sickness” and other cybersickness symptoms. Research in
the field has posed over forty potential factors that affect cybersickness. The objective of this
thesis is to prioritize cybersickness human factors and develop models to predict their effects. A
thorough search of the literature is given, which summarizes the support, or lack thereof, for
each factor. New experimentation is conducted for the factors with insufficient existing
evidence. The factors with sufficient support were fitted to single-term models to use for
guidelines and normalization. Cumulative models are then developed to predict individual and
configuration level cybersickness effects.
The literature on certain factors is unclear due to selective reporting of configuration features,
various configurations, different measurement methods, and limited studies. The literature
regarding individual susceptibility to cybersickness is particularly limited. Therefore,
experiments were performed that specifically addressed individual susceptibility with different
display modalities. Common display methods such as head-mounted displays and screens were

examined. Stereoscopic effects were also examined due to recent renewed interest and unclear
effects. All the experiments included surveys to collect data on individual susceptibility factors.
Cybersickness effects are not normally distributed, which decreases the reliability of common
statistical measures on cybersickness data. A new statistical method for cybersickness research,
called the zero-inflated model, is examined as an improved method for comparing the impact
of the features involved in cybersickness. The zero-inflated model has the benefit of having a
"non-susceptible" or "well" group built into the statistical test, which can be upwards of 50% of
the participants in a study.
An analysis of the literature permitted several variables to be removed from further
consideration and allowed single factor models to be created. Three independent variability
models for individual susceptibility were found. The individual model explains 37% of the
adjusted variance of a model and the linear models created for the hardware and software
explain 55-70% of the adjusted variance. Since the variance as predicted by the models is not
purely additive, the total explained variance with these models is likely 70-80%, which is a
significant improvement over the models proposed by Kolasinski's that explain 34% of the
variability of the model [1]. The models created using experimental results incorporating the
individual, hardware and software had an average absolute residual accuracy of less than 1%.

TABLE OF CONTENTS

LIST OF TABLES .............................................................................................................................. VII
LIST OF FIGURES ............................................................................................................................. IX
KEY TO ABBREVIATIONS AND SYMBOLS ........................................................................................ XI
1

Introduction ............................................................................................................................ 1
1.1 Cybersickness Factors ...................................................................................................... 3
1.2 Contributions of this Thesis.............................................................................................. 6

2

Background ............................................................................................................................. 9
2.1 Definition of cybersickness ............................................................................................ 11
2.2 Theories .......................................................................................................................... 12
2.3 Cybersickness Detection ................................................................................................ 14
2.3.1 Questionnaires ........................................................................................................ 15
2.3.2 Postural Instability .................................................................................................. 17
2.3.3 Biometrics and Physiological State ......................................................................... 19
2.4 Factors ............................................................................................................................ 22
2.5 Cybersickness Models .................................................................................................... 23

3

Analysis of Factors ................................................................................................................ 26
3.1 Individual Factors ........................................................................................................... 29
3.1.1 Demographics ......................................................................................................... 29
3.1.1.1
Age — Possible ................................................................................................ 29
3.1.1.2
Gender — Conflicted ....................................................................................... 30
3.1.1.3
Ethnicity — Untested....................................................................................... 31
3.1.1.4
Vision Correction — Untested ......................................................................... 31
3.1.2 Experience............................................................................................................... 32
3.1.2.1
Experience with Real-World Task — Conflicted .............................................. 32
3.1.2.2
Experience with Simulator or Habituation — Confirmed, Partial ................... 32
3.1.2.3
Video Game Play — Untested ......................................................................... 36
3.1.2.4
Hats — Untested ............................................................................................. 36
3.1.2.5
Duration — Confirmed .................................................................................... 37
3.1.3 Mental Attributes.................................................................................................... 39
3.1.3.1
Concentration Level — Probable..................................................................... 40
3.1.3.2
Mental Rotation Ability — Conflicted ............................................................. 41
3.1.3.3
Perceptual Style — Possible ............................................................................ 42
3.1.4 Physical Attributes .................................................................................................. 42
3.1.4.1
Eye Dominance — Unlikely.............................................................................. 43
3.1.4.2
Stereoscopic Visual Ability — Unlikely ............................................................ 43
3.1.4.3
Postural Instability — Confirmed .................................................................... 44
iv

3.1.4.4
History of Headaches/Migraines — Possible .................................................. 44
3.1.4.5
History of Motion Sickness — Probable .......................................................... 45
3.1.4.6
Weight and Body Mass Index — Possible ....................................................... 45
3.2 Hardware Factors ........................................................................................................... 45
3.2.1 Rendering ................................................................................................................ 46
3.2.1.1
Stereoscopic Rendering — Conflicted ............................................................. 46
3.2.1.2
Inter-Pupillary Distance — Possible ................................................................ 49
3.2.1.3
Screen Distance to the Eye — Untested.......................................................... 50
3.2.1.4
Update Rate — Conflicted ............................................................................... 50
3.2.2 Tracking ................................................................................................................... 51
3.2.2.1
Method of Movement — Probable ................................................................. 51
3.2.2.2
Calibration — Untested ................................................................................... 52
3.2.2.3
Position Tracking Error —Possible................................................................... 52
3.2.2.4
Tracking Method — Conflicted........................................................................ 53
3.2.2.5
Head Movements — Untested ........................................................................ 53
3.2.3 Screen...................................................................................................................... 54
3.2.3.1
Resolution/Blur — Possible ............................................................................. 54
3.2.3.2
Horizontal and Vertical Field of View — Confirmed, Partial ........................... 55
3.2.3.3
Weight of the Display — Possible.................................................................... 60
3.2.3.4
Display Type — Probable ................................................................................. 60
3.2.4 Non Visual Feedback ............................................................................................... 62
3.2.4.1
Type of Haptic Feedback — Untested ............................................................. 62
3.2.4.2
Ambient Temperature —Untested ................................................................. 63
3.2.4.3
Olfactory Feedback —Untested ...................................................................... 63
3.2.4.4
Audio Feedback —Untested ............................................................................ 63
3.3 Software ......................................................................................................................... 63
3.3.1 Movement............................................................................................................... 64
3.3.1.1
Rate of Linear or Rotational Acceleration —Untested .................................... 64
3.3.1.2
Self-Movement Speed and Rotation —Confirmed, Partial ............................. 65
3.3.1.3
Vection —Conflicted ........................................................................................ 71
3.3.1.4
Altitude above Terrain —Probable.................................................................. 72
3.3.1.5
Degree of Control —Confirmed....................................................................... 72
3.3.2 Appearance ............................................................................................................. 73
3.3.2.1
Screen Luminance —Untested ........................................................................ 74
3.3.2.2
Color —Possible ............................................................................................... 74
3.3.2.3
Contrast —Untested ........................................................................................ 74
3.3.2.4
Scene Content or Scene Complexity —Probable ............................................ 75
3.3.2.5
Global Visual Flow —Probable ........................................................................ 76
3.3.2.6
Orientation Cues —Possible ............................................................................ 76
3.3.3 Stabilizing Information ............................................................................................ 77
3.3.3.1
Focus Areas —Probable ................................................................................... 77
3.3.3.2
Ratio of Virtual to Real World —Untested ...................................................... 78
3.3.3.3
Independent Visual Backgrounds —Confirmed .............................................. 78
3.3.3.4
Sitting versus Standing - Confirmed ................................................................ 81
v

3.4

Summary of Factors ....................................................................................................... 82

4

General Procedure and Hypotheses ..................................................................................... 86
4.1 Individual Factors ........................................................................................................... 88
4.2 Hardware and Software Factors .................................................................................... 90
4.3 Environment ................................................................................................................... 92
4.4 Factor Limitations........................................................................................................... 94
4.5 Procedures and Measures.............................................................................................. 95

5

Experiments .......................................................................................................................... 98
5.1 HMD Weight Procedures ............................................................................................... 99
5.1.1 Specifications ........................................................................................................ 100
5.1.2 Results ................................................................................................................... 101
5.2 Screen Size Procedures ................................................................................................ 102
5.2.1 Specifications ........................................................................................................ 102
5.2.2 Results ................................................................................................................... 103
5.3 Display Type Procedures .............................................................................................. 104
5.3.1 Results ................................................................................................................... 104
5.4 Stereoscopic Rendering Procedures ............................................................................ 106
5.4.1 Specifications ........................................................................................................ 106
5.4.2 Results ................................................................................................................... 107
5.5 Individual Susceptibility Procedures ............................................................................ 107
5.5.1 Initial Results ......................................................................................................... 108
5.5.2 Additional Results ................................................................................................. 110
5.6 Participant Preferences ................................................................................................ 115
5.6.1 Informal Feedback ................................................................................................ 117
5.7 Cumulative Analysis and Results .................................................................................. 119
5.8 Linear Model using Reported Configurations .............................................................. 121
5.9 Zero Inflated Model using Experimental Results ......................................................... 124

6

Conclusion and Future Directions ....................................................................................... 128
6.1 Future Directions .......................................................................................................... 132

APPENDICES ................................................................................................................................ 135
Appendix A .............................................................................................................................. 136
Appendix B .............................................................................................................................. 142
Appendix C .............................................................................................................................. 151
BIBLIOGRAPHY ............................................................................................................................ 153

vi

LIST OF TABLES

Table I: Related Conditions Symptom Profiles ............................................................................. 10
Table II: Physiological Sensors Used in Cybersickness Research .................................................. 20
Table III: Potential Factors in Cybersickness ................................................................................. 24
Table IV: Re-categorized Potential Factors in Cybersickness ....................................................... 27
Table V: Removed Factors ............................................................................................................ 28
Table VI: Habituation Values......................................................................................................... 35
Table VII: The Results of Several Duration Studies ....................................................................... 38
Table VIII: FOV Values ................................................................................................................... 57
Table IX: Speed Values .................................................................................................................. 70
Table X: IVB Values........................................................................................................................ 81
Table XI: Confirmed Factors .......................................................................................................... 83
Table XII: Probable factors ............................................................................................................ 84
Table XIII: Possible Factors ............................................................................................................ 84
Table XIV: Conflicted Factors ........................................................................................................ 84
Table XV: Untested Factors ........................................................................................................... 85
Table XVI: Eliminated Factors ....................................................................................................... 85
Table XVII. Individual Factors Question Location, Reason, and Hypotheses................................ 89
Table XVIII: Experiments Rational and Hypotheses ...................................................................... 91
Table XIX: Coefficient of Determination due to Characteristics ................................................. 109
Table XX: MSSQ Correlations ...................................................................................................... 112

vii

Table XXI: Binned Game Play and SSQ Correlations ................................................................... 113
Table XXII: Binned Game Play and MSSQ Correlations............................................................... 113
Table XXIII: SSQ-T Values for Vision ............................................................................................ 113
Table XXIV: Headache Correlations ............................................................................................ 113
Table XXV: HMD Preferences ...................................................................................................... 116
Table XXVI: Large Preferences .................................................................................................... 117
Table XXVII: Cumulative Preferences.......................................................................................... 117
Table XXVIII: ZINB Model Terms ................................................................................................. 126
Table XXIX: Simulator Sickness Questionnaire Symptom Categories ......................................... 136
Table XXX: Literature Experiment Configurations ...................................................................... 142
Table XXXI: Cross Configuration Data Set ................................................................................... 146
Table XXXII: Some Factors that Require Further Study .............................................................. 150

viii

LIST OF FIGURES

Figure 1: Common Implementations of Virtual Environments ...................................................... 4
Figure 2: Orientation Conflicts with Incorrect Visual Stimuli ....................................................... 12
Figure 3: Rotational and Translational Axes. ................................................................................ 28
Figure 4: Habituation Data ............................................................................................................ 36
Figure 5: Fitted Duration Values ................................................................................................... 40
Figure 6: Heterophoria.................................................................................................................. 47
Figure 7: Virtual Field of View and Real Field of View Ratios ....................................................... 55
Figure 8: Real Field of View Data .................................................................................................. 58
Figure 9: Percent Change in SSQ-T with Change of FOV .............................................................. 59
Figure 10: Seasickness Model from McCauley et al. [116] ........................................................... 65
Figure 11: Nausea versus Navigation Speed from So, Lo, and Ho [77] ........................................ 67
Figure 12: Fitted Speed Values ..................................................................................................... 71
Figure 13: Scene Complexity (a-c) from So et al. [15] and (d) van Emmerik, Vries, and Bos [49] 76
Figure 14: Independent Visual Background .................................................................................. 79
Figure 15: Screenshots from the Application ............................................................................... 93
Figure 16: X-Box Controls.............................................................................................................. 94
Figure 17: Sample SSQ-T scores .................................................................................................... 99
Figure 18: The Head Mounted Display ....................................................................................... 100
Figure 19: Sample Textures......................................................................................................... 119
Figure 20: Original Usability Survey ............................................................................................ 137

ix

Figure 21: Revised Usability Survey ............................................................................................ 138
Figure 22: Full Golding's Motion Sickness Susceptibility Questionnaire (MSSQ) [141] ............. 139
Figure 23: Golding's [144] Short Motion Sickness Susceptibility Questionnaire ........................ 141

x

KEY TO ABBREVIATIONS AND SYMBOLS

AR - Augmented reality
R2 - Coefficient of determination or variance
CAVE - Computer Aided Virtual Environments
-

.

DOF - Degrees of Freedom
FOV - Field of view
IVB - Independent Visual Backgrounds
MHQ - Motion History Questionnaire
MSSQ - Motion Sickness Susceptibility Questionnaire
Our HMD - our experiment that compared a lightweight HMD to a heavy HMD
Our Screen - our experiment that compared a small screen to a large screen, but held the field
of view and standing position constant
Our Render - our experiment that compared a monoscopic application to a stereo application
SSQ-T/N/D/O - Simulator Sickness Questionnaire, total, nausea score, disorientation, and
oculomotor, respectively
VE - Virtual environment
VR - Virtual reality
ZINB - Zero Inflated Negative Binomial
** Correlations = p < 0.01
* Correlations = p < 0.05
∙ Correlations = p < 0.1
- lambda

xi

(x) - the Gamma function

xii

1

Introduction

Virtual and augmented reality technologies have been in development for decades. Throughout
that period of time, a significant subset of users has suffered from motion sickness-like
symptoms, which are commonly referred to as cybersickness or simulator sickness, depending
on the setting. Virtual environments provide a generally safe setting in which to learn control
systems, explore alternate realities, or consume media. Initially, these virtual systems were
expensive to build and maintain and were predominately limited to military and research
applications. Relatively recently, new systems have been created that are, or soon will be,
available for purchase by general consumers. These systems include the Oculus Rift and Sony
Morpheus projects as well as tracking equipment such as the Xbox Kinect and the Nintendo Wii.
This technological realm, once predominately developed for the military, is opening up to the
general consumer, which has brought attention to new applications as well as the human
factors involved.
There are few empirically supported guidelines for cybersickness in virtual environments. The
research for these guidelines was completed in three phases. The first phase involved an
analysis of the literature to identify the possible factors of cybersickness and then model those
factors. For example, a suggestion to decrease the size of the screen commonly appears in
warning labels. The literature suggests that reducing the size of the screen to one half its size
leads to an approximately 50% decrease in the severity and incidence of cybersickness
symptoms. To aid future studies all source factors were labeled as untested, unlikely, possible,
probable, confirmed, or conflicted. The second phase involved experimentation to clarify factors
that were ambiguous or lacking in the literature. For example, with the exception of gender,
1

demographic studies were largely absent from the literature. Therefore, background surveys
were utilized during this phase to collect demographic information. Experiments were
approved by Michigan State University's institutional review board with the approval number
13-460.
The final phase involved the creation of cumulative models from multiple factors. There are
many independent elements involved in virtual environments. A virtual environment can utilize
a variety of displays, interaction techniques, rendering methods, and applications. To predict
what will happen in a new system, several of these elements must be combined into one
model.
Cybersickness has been a known concern in virtual systems for decades, but the shift in its
availability to the general public has renewed attention to the frequency of these issues. A
quick Internet search for "movie sickness" returns thousands of results from people
complaining of motion sickness after watching movies such as “The Hunger Games,” “The Blair
Witch Project,” and others, particularly movies that were filmed in the ‘shaky camera’ style. In
fact, "The Blair Witch Project” has been used as a cybersickness stimulus for research purposes.
One particularly severe example of “movie sickness” is a movie that was shown to 294 seventh
grade students wherein 36 students became ill enough to warrant hospital treatment [2].
Oculus Rift's best practices specifically mention cybersickness and offer suggestions on camera
control to avoid inducing symptoms [3]. Nonetheless, cybersickness researchers report that
incidences of cybersickness affect 30% [4] to 80% [5] of the general population. Given that
virtual reality may cause disorientation, this is a safety issue as well as a usability issue. Despite

2

this high incidence of symptoms, the potential for cybersickness does not necessarily prevent
individuals from using virtual systems [6]. However, it does signify that improved methods to
minimize the impact of cybersickness are of critical importance.

1.1 Cybersickness Factors
Despite that fact that cybersickness has been a known issue for decades, its general theoretical
foundations are fundamentally incomplete. There are very few guidelines on the use of virtual
systems and the durations of safe exposure to them. Television and game manufacturers place
warnings on their devices suggesting limited usage, but the suggested durations are vague and
are typically not based on empirical or theoretical evidence [7,8]. One reason for the difficulty
in identifying the factors of cybersickness is the large number of potential factors of
cybersickness.
Kolasinski presented over 40 possible cybersickness factors in 1995 [9]. Examples of such
factors include screen size, stereoscopic rendering, the tasks performed, gender, mental
rotation ability, and interaction paradigms. These factors were grouped into three sources:
simulator, task, and individual factors. Moreover, there are many possible system
configurations and many of the factors can be selected independently of each other. For
example, a new virtual environment may allow a user to choose from a head-mounted display
or a large screen, monoscopic or stereoscopic rendering, a joystick or mouse for movement,
and the ability to track participant rotation only or to track position as well. This results in an
exponential number of configurations. For example, Figure 1 shows three different ways that
an identical scene may be presented, with or without tracking to update the virtual view.

3

Figure 1: Common Implementations of Virtual Environments
Given the numerous factors and various installations, it is sometimes difficult to determine if
any given cybersickness effect is due to the system configuration or the factor under
consideration. Another impediment is that cybersickness is highly individualized in nature. One
participant may be very sensitive to visual stimuli, whereas the next participant may show little
or no sensitivity to the same stimuli. Cybersickness in one participant may induce eye strain and
headaches, while for another participant it may induce balance and nausea symptoms. Since
the effects are individualized, the experimental design of cybersickness studies could benefit
from studying the same participants for each condition. This is difficult to accomplish because a
participant’s symptom incidence and severity tends to decrease with each session. Participants
are also frequently hesitant about additional time requirements. Currently, the decrease in
symptoms and the difference between participants are ignored as there are no models
available to normalize these effects. Phase one and three of our cybersickness research directly
addressed these problems by creating models to normalize some of these effects.

4

Prior general predictive measures of cybersickness have been suggested by Kolasinski [1] and
So [10]. Simon [11,12,13], and Jones, Kennedy, and Stanney [14] have suggested research
programs that would ultimately lead to the creation of predictive models, but these research
programs have not been realized. Kolasinski's model is the only model that accounts for the
individual. Kolasinski proposed a testing/qualification procedure and a predictive linear model
based on testing results, but the procedures could not be completed outside an academic
setting since it required special equipment which requires several square feet of room. This is a
problematic issue as it prevents the results from providing guidance unless users can visit a lab
to be tested. So's cybersickness dose method, which only considers the visuals of a scene
explained 85-97% of the proportion of variability the model, or coefficient of determination.
While this seems to have good accuracy, their experimentation only included six points and
small data sets are highly prone to over fitting which results in higher coefficients of
determination than what would occur in practice. This model also assumes no interaction with
the system so the result can only be partially applied to the dynamic systems of general virtual
environments. In addition, the measure requires an application to be built before a prediction
can be made, which makes the measure difficult to apply during the application design stage. A
later study by So, Ho, and Lo expressed their reservation regarding the effectiveness of a
cybersickness dose value when they found that scene complexity did not significantly affect the
level of cybersickness [15]. Simon and Jones, Kennedy, and Stanney both suggest using the
results of every tested configuration and then adjusting the predicted results according to how
the configurations differ. Simon suggests doing this directly with large linear models that
include all the factors, while Jones, Kennedy, and Stanney suggest using the average effect of

5

each factor and adjusting the results accordingly. Simon and Jones, Kennedy, and Stanney
require a large number of configurations be tested as well as a uniform distribution of
configurations. It is difficult for cybersickness research projects to meet these requirements and
the literature does not provide the data to create the models used in their approach.

1.2 Contributions of this Thesis
This thesis presents an analysis of the factors that influence cybersickness symptomology, new
predictive models for individual and novel configurations, and new statistical methods for
analyzing the data acquired during cybersickness research. The primary results of the research
were:


Several factors, such as display type, are shown to have negligible effects and those
effects can be explained by other factors.



Some factors were found to be too general and were consequently broken into subfactors. For example, ‘field of view’ can refer to the size of the real screen or the ratio of
the real screen to the virtual rendering plane. Also, researchers have only studied the
‘field of view’ measure along the diagonal or horizontally across the screen. The effect
of increasing the size of the screen vertically is unknown.



An analysis of individual factors resulted in several guidelines that can help to reduce
the incidence of symptoms, such as decreasing the size of the screen and seating
participants.



Three susceptibility factors that correlated with cybersickness were found, and these
factors are shown to be independent of each other. These factors are a history of

6

motion sickness, a history of headaches, and history of video game play of genres
known for large amounts of forward motion.


Models created to predict individual differences can be used to normalize the results of
different participants, allowing for the inclusion of different participants in each
condition.



The model that was created for individual susceptibility explains 37% of the adjusted
variance.



The linear models that were created from the literature results incorporating hardware
and software factors explain 55-70% of the adjusted variance.



The model that was created from experimental results incorporating individual,
hardware, and software factors has an average absolute residuals of less than 1%.

Because coefficients of determination are not purely additive, the total adjusted variance
explained is likely 70-80%, which is a significant improvement over Kolasinski's 34% of total
variance. Since different statistical models were used for the experimental and literature
results, these models could not be directly compared.
Using the results of this thesis, virtual environments can be dynamically adapted to
participants. The single factor models allow developers to create low, medium, and high
cybersickness probability settings, trading features for decreased susceptibility. For example, a
high field of view provides greater immersion, but also produces a higher probability of
cybersickness. Using the field of view model, participants can estimate how much they need to

7

restrict the display to reach a certain symptom level for the "low" setting. The individual
susceptibility model can be applied prior to entering a virtual environment, so developers can
suggest initial settings for the user.

8

2

Background

Research on cybersickness and simulator sickness has been ongoing for several decades. The
original objective of the studies was to determine the factors that were most likely to make
trainees ill in military simulators. As a result, the research prior to 1995 focused on hardware
with the belief that improving the hardware would eliminate the problem. This theory began to
be challenged in 1995 by Mon-Williams, Wann, and Rushton who suggested that improving the
screen hardware may actually increase symptoms [16]. The change in systems from simulators
with mock ups of controls to visor and large screen displays made the systems more
conventional. Unfortunately, reports of cybersickness began to increase. The change in systems
also resulted in new terms for the symptoms including cybersickness, visually induced motion
sickness (VIMS), virtual reality induced symptoms and effects (VRISE), and others. The
distinction between cybersickness and simulator sickness is that simulator sickness implies a
physical mockup of the control systems while cybersickness lacks such a mockup. In
“Cybersickness is not Simulator Sickness”, Stanney, Kennedy, and Drexler proposed that
cybersickness and simulator sickness are different maladaptations [17]. Stanney, Kennedy, and
Drexler reported that the symptom profile, or the relative severity of oculomotor,
disorientation, and nausea symptoms, differed between simulator and non-simulator systems,
as shown in Table I.
More recently, there has been a call to formally unify the two fields by Blea et al. [18], Bos, Bles,
and Groen [19], and Smart, Otten, and Stoffregen [20]. They claim that all the symptoms are
due to the same source, namely postural instability. Postural instability is the extent to which a

9

person sways when standing still. Whether these two fields will reunite is uncertain, but
simulator sickness is currently considered to be the ancestor of cybersickness research.
Table I: Related Conditions Symptom Profiles
Military
Simulators

Sea Sickness

Space Sickness

Cybersickness

Highest Rating

Oculomotor

Nauseagenic

Nauseagenic

Disorientation

Middle Rating

Nauseagenic

Oculomotor

Disorientation

Nauseagenic

Lowest Rating

Disorientation

Disorientation

Oculomotor

Oculomotor

Although research in cybersickness and simulator sickness has been ongoing for decades, there
are few fundamental theories about them. This is partially due to the large number of potential
factors involved in cybersickness and simulator sickness. Moreover, many of the hardware and
software factors such as display type, stereoscopic rendering, and navigation can be selected
independently of each other. This results in an exponential number of potential configurations
to test.
Exacerbating these issues is the fact that there is currently no standard testing system. The
virtual environment performance assessment battery (VEPAB), which was designed for
simulator performance and usability, was developed in 1994, but has not been widely used
[21]. The tasks in the VEPAB focus on the quality of visuals, locomotion, tracking, object
manipulation, and reaction time in virtual environments. There are 22 tasks in total, including
distance and height estimation, navigation tasks, and simple activation of controls. The virtual
environment used by the VEPAB was eventually deemed to be overly simplistic due to later
hardware and graphics improvements, but no replacement system has been developed. While

10

the original VEPAB environment is no longer in use, the variety of required activities/tasks it
used still applies to today’s virtual environment interfaces.

2.1 Definition of cybersickness
Cybersickness is defined as the feeling of symptoms similar to or related to motion sickness in a
virtual environment. Since there is typically no actual physical motion, cybersickness is
generally considered to be "visually induced." Medically, cybersickness symptoms may include
nausea, pale skin, cold sweats, vomiting, dizziness, headache, increased salivation, and fatigue
[22]. Because virtual environments place additional strain on the eye, the symptoms can also
include difficulty focusing and eyestrain [17]. Because cybersickness is polysymptomatic (many
symptoms) and polygenic (manifested symptoms differ from individual to individual), the
degree and sensitivity to the environment varies widely between individuals. When
cybersickness is said to increase, it typically means that its severity and incidence frequency has
increased.
The greatest risks in virtual systems are disorientation and nausea. Virtual systems can affect
eye function and diminish distance perception, and can therefore create a travel risk after using
them. Emetic responses are uncommon, with less than 2% of participants experiencing nausea
or vomiting, but it is known to be the least tolerated symptom [23,24,25]. More worrisome is
the fact that it may take time for the effects induced by virtual systems to subside. Studies by
Stanney and Kennedy [23] and Stanney et al. [24] demonstrate that the nausea caused by
virtual environments may persist for an hour and that disorientation may last for two to four
hours.

11

2.2 Theories
Cybersickness researchers rely on a variety of theories on the cause of illness to design and
support their experiments. The biological causes of cybersickness have not been firmly
established, and different theories hypothesize the existence of different responsible factors in
virtual environments. The most prevalent theories in cybersickness research are sensory
mismatch, postural instability, and rest frame theory [18,19,26,27,28].

Figure 2: Orientation Conflicts with Incorrect Visual Stimuli
Sensory mismatch is the most common theory. It posits that if the stimulus from the outside
environment is perceived differently by different senses, it will induce motion sickness. For
example, if a picture is tilted as shown in Figure 2, the vestibular system of the participant
experiences gravity as acting straight down, but the visual system perceives that gravity should
be tilted with the picture. Sensory mismatch is often the explanation as to why vection, the
perception of the world moving away from the user, is strongly correlated with feelings of

12

cybersickness. The features of primary concern in sensory mismatch theory are accurate
tracking and navigation.
Postural instability theory posits that motion sickness will be induced if a person is unable to
maintain a stable posture given a stimulus from the outside environment. The stability of a
posture is typically defined as the amount of sway in the posture. Using the picture example,
the individual's posture may be slanted to correct for the orientation of the visual stimulus, but
because gravity is felt straight down, this posture becomes unbalanced. The more unbalanced
the posture becomes, the more ill the participant feels. Postural instability is not as
straightforward as having sway versus no sway. Seating participants, which should eliminate
sway, does decrease illness but does not completely remove the effect [29,30]. In addition, well
participants often have more sway along certain axes than sick participants [31]. Postural
instability can be viewed as a more restricted form of sensory mismatch, because it focuses on
the vestibular system being unable to cope with a stimulus. Postural instability theory is gaining
popularity in the cybersickness research community because it is subject to objective
evaluation.
Rest frame theory is based on the direction the user perceives or assumes is “up” in relation to
the world. Theoretically, the degree of symptoms relates to how much this perception deviates
from actual gravity. It is similar to postural stability in that the symptoms are due to a perceived
"up" versus actual gravity. However, it is more general to when postural control may be less of
a factor, such as when participants are seated. Rest frame theory largely relies on the eye's
vestibulo-ocular reflex (VOR), or the eye's ability to track objects when the head is moving. Virre

13

explains that the body assumes a certain proportional gain between the vestibular and visual
systems within certain frequencies of movement [27]. If the gain of the VOR changes (such as
when a prescription in eyeglasses changes), the habitual level of movement is no longer
accurate, and the eyes must re-adapt until the objects are stabilized given the new VOR gain
(typically a few days for a 5% prescription change). Virre also mentions that two states of VOR
can exist and that the nervous system can quickly switch between these two states. The theory
is frequently used to explain why the frequent use of a virtual environment results in less
cybersickness and is the basis for the use of independent visual backgrounds [32]. There is
some argument against the two state VOR theory from Kennedy and Stanney, who mention
that the altered VOR persisted between their experimental sessions rather than simply
switching when the immersion began [33].

2.3 Cybersickness Detection
Cybersickness researchers have diverse means to determine the severity and incidence of
participants’ symptoms. The most commonly used measures are questionnaires, with the
simulator sickness questionnaire (SSQ) from Kennedy et al. being the de facto standard [34].
The SSQ does not readily permit the monitoring of symptoms as it takes too long to administer.
Thus, one-question scales that ask for a participant's current state are regularly used. Given
that questionnaires are subjective, objective measurements using postural instability and
physiological state are also under development. Postural instability is gaining popularity as
inexpensive and robust sensors become available. The physiological state measure has only
been studied in the last several years and is currently not reliable enough to replace other
methods of measurement.

14

2.3.1 Questionnaires
The most common questionnaire in cybersickness studies is the simulator sickness
questionnaire (SSQ), which was developed in 1993 by Kennedy et al. [34]. The questionnaire
and its scoring metrics are included in Appendix A. It remains the most commonly applied
measure in cybersickness research and is frequently used as the target value during the
development of new detection measures. The SSQ consists of 16 questions requesting the
current

degree

of

severity

of

each

symptom.

The

severity

scale

is

“none, slight, moderate, or severe” or “0, 1, 2, 3”, respectively. The participants’ responses are
grouped into three categories, each with its own weighting. A score below 10 is traditionally
considered to be normal. The three categories are nausea, oculomotor, and disorientation, and
are abbreviated N, O, and D, respectively.
The SSQ does have some notable disadvantages. There is a degree of cross-correlation in the
categories. For example “general discomfort” appears in both the nausea and oculomotor
categories, and there has been some concern over its generalization because it was developed
based on a military demographic. Bouchard, Robillard, and Renaud refer to this redundancy in
the categories as loading. They attempted to remedy both issues by refactoring the SSQ using a
nonmilitary populace to avoid cross-loading and created a new questionnaire with two
categories: nausea and oculomotor [35]. The participants consisted of 71% females, a third of
whom had anxiety disorders. This demographic is dissimilar from the general populace as well
as the populations reported in most nonmilitary cybersickness studies, and the usage of the
refactored SSQ is limited.

15

Following the SSQ in popularity are numerous verbal one-question surveys that analyze
symptom severity over time. One-question scales vary from study to study, but typically involve
asking the participants to report their current level of discomfort as a single number at regular
intervals. The disadvantage of these scales is that they disturb the participants on a regular
basis and increase their awareness of their physical state. This results in “priming” or “demand”
characteristics, or an increase in reported symptom due to expecting the symptom. There has
been little research on the level of this effect in cybersickness, but there is potentially a strong
effect. Young, Adelstein, and Ellis tested the effect of demand characteristics on the SSQ by
considering post-immersion scores with and without a pre-immersion test [36]. Not
surprisingly, the scores on the post-test were higher when a pre-test was given. In their results,
the nausea category had the largest priming effect with a change of symptoms resulting in the
post-test scores being only a third as high without the pretest. A related study by Nichols et al.
analyzed the correlation of the participant’s sense of the presence of an “interface” and its
effect on cybersickness, where the “interface” was defined as: “…whether the interaction
devices or display quality interfered with the virtual experience, and how well participants felt
they could concentrate on the tasks in the virtual environment [6].” They found negative
correlations between the “interface” and all the SSQ categories, but they also found no
correlation between enjoyment and sickness ratings, signifying that feeling ill did not stop the
participant from enjoying the experience. This suggests that if participants are sufficiently
distracted, they may not notice mild symptoms.

16

Ling et al. reported the opposite finding in a sense [37]. They found a positive correlation
between participants that were more likely to get involved with their virtual environments and
cybersickness.

2.3.2 Postural Instability
Given the subjective nature of questionnaires, there have been several attempts to develop
objective measurements for motion sickness over the past two decades. Postural instability was
posed as a low-cost objective measure that caused limited interference with the participant and
could yield continuous symptom levels. Postural instability detection relies on the postural
instability theory of motion sickness, which states that the more ill someone becomes, the
more unstable their posture becomes. In practice, postural instability detection is neither
continuous nor does it leave the participant undisturbed. Postural instability detection typically
requires that a specific standardized stance be taken every few minutes, resulting in greater
interference than one-question scales, and only the position data recorded during the stance is
used for analysis. Postural instability researchers initially had difficulty finding consistent
correlations with cybersickness scores. The postural instability stances were then monitored for
"stance breaks" or "time-till-failure." Stance breaks occur when a participant can no longer
maintain a stance and is typically recorded as the number of breaks required per minute. Timetill-failure is the amount of time between when the participant first enters a stance to when a
break occurs. As Cobb explains, there are several flaws with these methods [38]. Cobb found
that many of the definitions of stance breaks were sufficiently vague, and that the same
description could yield different results. For example, “loss of balance” could mean that the
foot moved from its original position by any measurable amount, the foot moved more than

17

two centimeters from the original position, or several other movements. In addition, since the
results generally rely on special postures, there is a potential for learning effects.
Kennedy and Stanney [33] bridged postural stance and sway by examining sixteen different
stances with the intent of identifying a stance that is sensitive to postural instability. After
determining a stance (the tandem Romberg stance, which is comprises of a heel to toe position
of the feet, arms folded across the chest, and eyes closed), they analyzed the correlations
amongst time-till-failure, extent of head movement, level of movement as described by an
observer, and horizontal velocity mean and standard deviation. They found the most reliable
result with the horizontal velocity mean. The correlation of the horizontal velocity mean with
illness was preliminary, but promising.
With the development of robust postural sway hardware, postural stability measurement
studies gained reliability and accuracy. Both Cobb and Regan examined postural sway in their
studies and reported correlations with cybersickness. The primary disadvantages of postural
sway methods continue to be that a specific stance is often required that disturb the
participant, the specific stance may result in a learning effect, and the analysis is sensitive to the
sampling procedure used. This means that an analysis may not be applicable to different
stances, sampling periods, or sensor positions. For example, Ehrlich showed no effect on
postural sway while their SSQ scores showed an increase in symptoms with a sampling period
of 0.54 seconds [39], while researchers that have found correlations employed far smaller
sampling periods, for example, Chang et al. [40] used a 40 Hz rate, Smart, Otten, and Stoffregen
[20] used a 40-50 Hz rate, and Villard et al. [31] used a 60 HZ rate. For these reasons, while

18

postural stability has shown good correlations with cybersickness, it does not yet have a reliable
predictive method (although Smart, Otten, and Stoffregen discuss the use of potential
discriminate functions).

2.3.3 Biometrics and Physiological State
To address the fact that postural instability interferes with virtual environment interactions and
has the potential for learning effects, there have been attempts to develop objective detection
measurements that do not disturb the participant. In recent years, there has been an
exploration of “bio-signals” or “physiological signals” measurement of cybersickness.
Physiological signals such as the heart rate (known as the R-R interval) and blood pressure are
analyzed to determine their correlations with cybersickness. This method has the advantage of
having a very limited interference with the participant's usage of the virtual environment and a
lower potential for learning effects than that of postural instability methods. The disadvantages
of the method are that wearing the sensors is often uncomfortable and determining the
correlations with cybersickness is challenging. Some of the more common detection sensors
include electrocardiogram (ECG), blood pressure, electrogastrogram (EGG), and respiration
(RSP).

ECG

monitors

the

heart

signal

and

EGG

monitors

stomach

movement.

Two methods have shown promise thus far: ECG/blood pressure ratios and the EGG power
spectrum. Roberts and Gallimore attempted to use the EGG as an exclusive detection
measurement in 2005 [41]. They analyzed EGG data for gastric shifts, magnitude changes, and
power spectrum changes. They focused their attention on tachygastria (an EGG with a power
spectrum of four-nine cycles per minute rather than the standard three cycles per minute).
They found that tachygastria was present during participants’ exposure to stimuli and that it

19

was correlated with duration. They developed a neural network that used the EGG signal as the
input and the questionnaire rating as the target for each individual, which converged to attain a
sum-squared error of less than one. The development of a general model was not discussed.
A list of the sensors used in cybersickness research is presented in Table II.
Table II: Physiological Sensors Used in Cybersickness Research
Sensor

Definition

Blood pressure

Standard blood pressure
measurement

Electrocardiogram (ECG)

Detects the heart rate (R-R
interval)

Electroencephalogram(EEG)

Electrogastrogram (EGG)
Electrooculogram
(EOG)
Galvanic Skin Response (GSR)

Respiration (RSP)

Correspondence to
Cybersickness
Indirectly related to heart
rate
The ratio of low to high
frequency power has
shown a correlation

Detects brain activity

Preliminary results only

Detects stomach movement
and strength

The magnitude and high
power frequencies have
shown a correlation

Detects the movement of
the eye
Detects the amount of
moisture produced by the
skin. It is associated with
stress
Strength and frequency of
breath

None

None
The magnitude and high
power frequencies have
shown a possible
correlation

Also in 2005, Kim et al. examined a majority of the common physiological signals for any
correlations with cybersickness using a highly provocative stimulus (78.9% of the participants
experienced symptoms after only 9.5 minutes) [5]. They found a correlation between
cybersickness and EGG data, eye blink rate, and heart rate. The EGG displayed a gastric

20

tachyarrhythmia (a higher than normal spectral power) increase during the duration of the
exposure to the stimulus. The eye blink rate decreased at the beginning of the exposure period,
increased beyond the baseline within several minutes, and then leveled. The heart rate was
faster at the beginning and then slowly decreased down to the baseline, but returned quickly to
the baseline when the participant exited the virtual environment. Of these effects, EGG is the
signal that is most highly associated with cybersickness. The heart rate and eye blink rate can
be explained by the psychological effects of the participant’s initial interest in the new
environment and the process of becoming familiar with it.
The power frequencies of the physiological signals are often split into a lower frequency range
(LF), approximately 0.04–0.15 Hz, and a high frequency range (HF), 0.16–0.45 Hz. Kiryu et al.
examined ECG data, blood pressure, and respiration [42]. They reported a correlation between
duration and cybersickness using ECG, blood pressure, and respiration LF/HF ratios as
measurements [42]. Watanabe and Ujike also found that the LF/HF ratio obtained from the
heart-rate correlated with duration [43]. While the LF/HF ratio did not display a significant
correlation with the travel conditions in their experiment, the SSQ did display a significant
correlation. Because the LF/HF ratio between the two conditions was consistently different, the
LF/HF may not be as sensitive as the SSQ, but it is promising and warrants further analysis. The
underlying assumption of these studies is that a large LF/HF ratio occurs when the sympathetic
nervous system is very active. Biologically, this means that the heartbeat becomes stronger
and/or the blood flow has decreased turbulence.

21

The pupillary reflex (how much the pupil dilates compared with a participant’s baseline) has a
possible but poorly understood correlation with cybersickness [44]. Harvey and Howarth found
that skin temperature did not have a correlation with cybersickness [45]. The correlation of
EEGs with cybersickness have been more difficult to determine, but Ji et al. have proposed the
potential neural pathways that might be affected [46]. Chang et al. also examined to the
correlation between EEG data and cybersickness [47]. They found that lower delta and theta
powers were associated with fewer symptoms whereas higher alpha and beta powers were
associated with fewer symptoms. The usefulness of EEG data for the real time prediction of
cybersickness remains undetermined.

2.4 Factors
One major issue in cybersickness research is the number of factors and the flexibility of their
configurations. In 1995, Kolasinski listed over 40 possible factors [9], and Renkewitz and
Alexander updated this list in 2007 [48]. The combined list is shown in Table III. This large
number of factors necessitates the prioritization of experiments. In early simulator sickness
research, the studies focused on hardware, with the belief that hardware advances would
resolve the issue. However, although several of the hardware factors have improved
significantly over time, reports of cybersickness have actually increased. Research later shifted
with some success to focus on navigation direction and rotations.
Unfortunately, comparing results from different researchers and their labs has proven to be
difficult. The navigation paradigms, rendering parameters, and displays can generally be
selected independently of each other. The result is uncertainty as to whether the variance in
the results from different researchers is due to the factors being studied or the differences in

22

the configurations being used. For example, when examining the effect of field of view, van
Emmerik, de Vries, and Bos [49] argued that their results differed from those of Draper et al.
[50] as a result of screen size differences. Peli [51] argues that they found lower symptoms than
Howarth and Costello [52] when examining monoscopic versus stereoscopic viewing due to
their dissimilar survey procedures and a more active task.

2.5 Cybersickness Models
A general predictive model for cybersickness is currently lacking as each factor is independently
considered for a given configuration. This means that for each new configuration or application,
the only method to ascertain its expected cybersickness effects involves testing it with
participants. While there are no general models in use, experimental programs and theoretical
models have been proposed.
In the 1970's and 1980's, Simon proposed the empirical testing of several factors using a
central-composite experimental design, which is a factorial design that includes median and
median leave-one out experimental runs [11,12,13]. Median and median leave-one-out
experimental design is similar to mean and mean leave-one-out experimental design, but uses
the median rather than the mean as the primary statistic to compare results. The conditions
either include all the factors or leave one factor out. The relative magnitude of each effect is
then determined by regression analysis. Using these magnitudes, cybersickness could
theoretically be modeled and held below a specific threshold.
The Jones, Kennedy, and Stanney protocol proposed in 2004 modified Simon's approach [14].
Rather than creating a parameterized model of cybersickness, they proposed to extrapolate

23

new data using previous tests. Jones, Kennedy, and Stanney and Simon both considered the
configuration as a whole, but ignored individual factors.
Table III: Potential Factors in Cybersickness
Individual
Age
Concentration level
Ethnicity
Experience with real-world
task
Experience with simulator
(habitation)
Flicker fusion frequency
threshold
Gender
Illness and personal
characteristics
Mental rotation ability
Perceptual style
Postural stability
Eye dominance

Simulator (Display System)
Binocular view
Calibration
Color
Contrast

Task
Altitude above terrain
Degree of control
Duration
Global visual flow

Field of View

Head movements

Flicker

Screen luminance

Inter-pupillary distance
Motion platform

Method of movement
Rate of linear or rotational
acceleration
Self-movement speed
Sitting versus standing
Type of application
Unusual maneuvers
Vection

Phosphor lag
Position tracking error
Refresh rate
Resolution
Scene content (scene
complexity)
Time lag
Update rate
View region
Ambient brightness
Screen distance to the eye
Type of haptic feedback
Ambient temperature
Olfactory feedback
Audio feedback
Tracking method

So attempted to model cybersickness with the cybersickness dose value [10], which is modeled
after the Motion (sea) Sickness Dose Value (MSDV), which can predict sea sickness in
passengers. The basis of the metric is how much a scene changes over time along the vertical,

24

horizontal, and radial axes multiplied by display, task, and individual scaling factors. An update
on the progress of this “spatial velocity” or “cybersickness dose value” was provided in 2002 by
Chen, Yuen, and So [53]. This publication removed the display, task, and individual scaling
factors. A later study by So, Ho, and Lo expressed doubt on the sole use of their cybersickness
dose value when it was discovered that scene complexity did not significantly affect the level of
cybersickness. Because high scene complexity undoubtedly affects the cybersickness dose
value, additional factors in the scene must be considered [15].
One problem with all of the above models is that susceptibility is not addressed. Prior studies
have shown that single individual attributes can explain 10-18% of the variance in an
experiment and that several individual attributes may explain nearly half of the variance
[23,54]. The Simon and Jones, Kennedy, and Stanney experimental protocols require a large
number of empirical results, the consideration and reporting of several factors, and a uniform
spacing of data points. Cybersickness research is rarely able to meet such stringent
requirements.
Kolasinki proposed a linear model that explained 34% of the variance in an experiment. As
Kolasinki explains, cybersickness data is not normally distributed, which makes linear models
less reliable. In addition, certain factors, such as perceptual style, are time consuming to
perform and cannot be tested at home. This makes these factors unsuitable for commercial
products.

25

3

Analysis of Factors

A cybersickness model with over 40 factors would be difficult to understand and would be
highly prone to over fitting given the limited number of data points. Therefore, eliminating the
unlikely factors is critical for making the cybersickness modeling problem tractable. Using the
literature, we sought to find support, or the lack thereof, for the factors posed by Kolasinski [9]
and Renkewitz and Alexander [48]. The factors were rearranged into three categories:
hardware, software, and individual, as shown in Table IV.
The factors were further assigned to subcategories for ease of reference and to delineate
features that may be correlated with each other. Several factors were added to the list. These
include susceptibility features such as vision correction and video game play as well as new
topics in cybersickness such as independent visual backgrounds. These features are marked
with an (A).
Rendering parameters such as field of view were placed in the hardware category, although
software may affect them as well. While rendering parameters may be affected by software,
they are typically limited by hardware. The analysis of individual factors is more tentative, as
individual factors are often based on potentially inaccurate studies of motion sickness, as
shown in the age studies by Arns and Cern [55] and Park et al. [56].
Several features that were immediately removed from consideration are shown in Table V.
These features were ambiguous, near duplicates, or no longer applicable due to hardware
improvements. Some may argue that flicker fusion and phosphor lag remain applicable.
However, the displays that show noticeable flicker are usually deemed unfit for virtual reality

26

and have already been removed from use, and phosphor lag is directly related to blur because a
lengthy phosphor lag leads to visual blur with eye movements. “View region” was renamed to
"focus area" to include both the location of the screen and the locations of the objects within
the screen.
Table IV: Re-categorized Potential Factors in Cybersickness
Individual
Experience
Experience with real-world
task
Experience with simulator
(habituation)
Video game play (A)
Hats (A)
Duration

Hardware
Screen
Resolution/Blur
Horizontal and vertical field of
view
Weight of the display (A)
Display type

Physical Attributes
Eye dominance
Stereoscopic visual ability (A)
Postural stability
History of
headaches/migraines (A)
History of motion sickness
BMI

Tracking
Method of movement
Calibration
Position tracking error
Tracking method

Demographics
Age
Gender

Rendering
Stereoscopic rendering
Inter-pupillary distance

Ethnicity

Screen distance to the eye

Vision correction(A)

Update rate

Mental Attributes
Concentration level
Mental rotation ability
Perceptual style

Non Visual Feedback
Type of haptic feedback
Ambient temperature
Olfactory feedback
Audio feedback

Head movements

27

Software
Movement
Rate of linear or rotational
acceleration
Self-movement speed and
rotation
Vection
Altitude above terrain
Degree of control
Appearance
Screen luminance
Color
Contrast
Scene content or scene
complexity
Global visual flow
Orientation cues
Stabilizing Information
Focus Areas
Ratio of virtual to real world
(A)
Independent visual
backgrounds (A)
Sitting versus standing

Table V: Removed Factors

Illness and personal
characteristics
Flicker fusion frequency
threshold
Phosphor lag
Perceptual style

Dropped
Time lag

Type of application

Refresh rate

Unusual maneuvers

Ambient brightness

Flicker

The factors are labeled as untested, unlikely, possible, probable, confirmed, or conflicted.
Confirmed factors that had sufficient support to define a model were included when developing
the predictive model using all three categories. A few factors potentially required the use of
sub-components in order to separate them from their primary factors for further examination.
These factors have the additional label of partial.
Several factors deal with direction of movement and angle of rotation. These directional and
rotational axes are illustrated in Figure 3.

Figure 3: Rotational and Translational Axes.
28

3.1 Individual Factors
Individual factors range from standard demographics, to mental and physical characteristics, to
prior exposure to similar systems. Many of the individual susceptibility factors in cybersickness
draw heavily from motion sickness studies, with the assumption being that the effects are
identical for both types of sickness. Unfortunately, studies of factors such as age have
decreased the reliability of these assumptions [55]. We only list the cybersickness specific
studies for each factor if applicable studies exist. If no cybersickness specific studies exist, the
results from motion sickness studies are used and are indicated as such.

3.1.1 Demographics
Demographics are the easiest individual factor to determine, with gender being the most
commonly studied demographic feature. “Vision correction” was added as a new factor in our
experiments. Vision correction implies an additional level of refraction and can, in some cases,
shift display optics out of alignment with the eyes.

3.1.1.1 Age — Possible
Until recently, virtual environment researchers assumed that susceptibility to cybersickness
diminishes with age, as is the case with traditional motion sickness [57]. Motion sickness is
more prevalent in younger individuals, with the effects slowly decreasing until age 50, at which
time the effect plateaus. Arns and Cern demonstrated that this pattern curve does not apply to
cybersickness [55]. They had a system that was open to visitors to their virtual reality facility
and reported that the visitors who were under the age of 15 experienced less cybersickness
than the adult visitors. Park et al. also reported a smaller percentage of early withdrawals with
younger participants (13.7%) than older participants (37.3%) [56]. In contrast, Ujike, Yokoi, and
Saida, reported a higher incidence of cybersickness in participants under the age of 30
29

compared with participants over the age of 30, but no statistical analysis was performed and
their actual distribution of the ages was uncertain [58].
Because age is easily determined, it can be included in a general model of cybersickness.
Currently, its pattern of effect is indeterminate. Recruiting older participants can be difficult.
Most studies use college age participants, and most systems are not open to children or the
general public.

3.1.1.2 Gender — Conflicted
There have been numerous studies regarding the effect of gender on cybersickness. Many
studies report that females report more severe cybersickness symptoms, although there is
some evidence that males under-report the severity of their symptoms [23,59,60]. However,
some studies have found no statistical differences with respect to gender and the severity of
symptoms in some cases [1,37,61].
Harm, Taylor, and Bloomberg [62] and Clemes and Howarth [63] offer two possible
explanations for these disparate results. Harm, Taylor, and Bloomberg reported that while
females reported their symptoms sooner than males, they also recovered faster than males.
This means that the experimental procedures of duration and breaks can affect the
cybersickness scores relative to gender demographics. Clemes and Howarth considered
hormone levels as a possible source of explanation for the disparate results. They reported that
a female participant’s hormonal levels over the menstrual cycle correlated with cybersickness
and that her susceptibility was higher on day 12 of a 28 day cycle. While this phenomenon may
affect a female participant’s cybersickness on day 12, it also means that it might not affect her
cybersickness the majority of the time. Another possible explanation is that females generally
30

have a wider field of view [64]. Since a wider field of view influences what is seen, it may affect
cybersickness.
Currently, the variation in gender studies is too high to warrant the inclusion of gender in a
cybersickness model. Moreover, this variation may be due to other factors such as selfreporting reliability, general motion sickness susceptibility, visional field of view, and hormonal
levels. Graeber and Stanney's study supports this hypothesis [65]. After balancing gender with
susceptibility, there was no significant difference in their resulting SSQ-T values.

3.1.1.3 Ethnicity — Untested
No ethnicity-specific studies were found for cybersickness. Studies in motion sickness have
reported that participants of certain ethnic backgrounds report higher levels of motion
sickness. Klosterhalfen and Kellermann reported that while Chinese participants reported a
higher susceptibility than Caucasian participants, the actual degree of symptoms reported by
the Chinese participants was significantly lower than that of the Caucasian participants [66].
However, many studies have reported that Chinese participants have a greater susceptibility to
motion sickness compared with other ethnic groups [67,68].

3.1.1.4 Vision Correction — Untested
Because virtual environments deal primarily with sight, any individual characteristic affecting
the eyes is of interest. Stereoscopic ability has been previously examined as a cybersickness
factor, but visual accuracy has not been studied. Contacts and eyeglasses require another layer
of refraction before visual stimuli can reach the eye. Glasses may shift the optics of a system
out of alignment. If a system does not permit glasses, as is the case with some models of Oculus

31

Rift, then the scene may be blurred. No cybersickness studies were found regarding vision
correction, but it was included in the susceptibility survey in our experiments.

3.1.2 Experience
The effect of order, or habituation, is well documented as decreasing cybersickness with
repeated use. Duration of immersion is also well documented as increasing cybersickness.
However, other sources of habituation such as real-world tasks, video game play, and other
virtual reality configurations remain largely unstudied.

3.1.2.1 Experience with Real-World Task — Conflicted
Because the repeated use of a virtual environment decreases reports of cybersickness,
repeated use of the real world counterpart may also affect cybersickness. Participants who are
familiar with the real object may be less tolerant of discrepancies, or they may have already
developed a resistance to the stimuli. This is a feature of concern in training simulators.
Kolaninski reported past research on simulation sickness in flight simulators indicated no clear
correlation between a participant’s experience with a real world task and cybersickness [9].

3.1.2.2 Experience with Simulator or Habituation — Confirmed, Partial
Cybersickness studies consistently report a decrease in symptoms through habituation
[69,70,71,72,73]. However, how long this effect lasts is less certain. Stanney and Kennedy
suggest having sessions every two to five days to encourage habituation [69]. Howarth and
Hodder had participants return ten times, and spaced the sessions two to seven days apart [72].
All the conditions resulted in a significant decrease in symptoms, with 50% of the participants
reporting no symptoms at the end of the experiment. Regan's study implies that this effect may
be limited for participants that return four times [74]. In Regan’s study, the spacing between

32

the first, second, and third sessions was four months, while the spacing between the third and
fourth sessions was one week. The participants reported a decrease in their symptoms between
the first and second session, which was likely due to the reduced novelty of the system and any
coping strategies that may have been learned. We noticed a similar effect in our own
experiments where some participants in their second sessions would display markedly different
interactions with the system compared with their first sessions. There was almost no change in
participants’ symptoms between Regan's second and third sessions, but there was an
additional decrease in their symptoms between the third and fourth sessions. This implies that
loss of habituation can occur over time periods between one week and four months.
One limitation of this approach is that it is unknown whether other virtual environment systems
also help to develop habituation. This includes stereoscopic movies, which contain similar visual
stimuli. Here, the habituation effect is difficult to determine, as the change in display may also
affect the level of cybersickness. For habituation to be examined with stereoscopic movies, the
effect of the display would need to be normalized.
The habituation value reports from the above studies are shown in Table VI. We include our
results if we had a sufficient number of participants within a given time period. For example,
less than 5 of our participants had 20+ days between sessions, which was not a large enough
sample to reliably calculate statistical significance. Our experiments include display weight (Our
HMD), size of the screen (Our Screen), and stereoscopic rendering (Our Render) and are
described in more detail in Section 4. Studies that only reported the number of people who did
not become ill, rather than an estimate of symptoms, are not included in Table VI.

33

The values in Table VI seem erratic upon initial inspection. For example, the decrease in the
absolute values of the SSQ differs substantially between the weight and screen experiments.
But upon closer inspection, the percentages of decrease are more consistent, and Howarth and
Hodder's study implies that the percentage of decrease remained fairly consistent even with
multiple repeat sessions. The literature habituation values are combined with our own in Figure
4. While the constancy has improved over the original SSQ-T values, the data points are not
well grouped besides the fact that they are generally less than 0. Part of this lack of consistency
is due to a difference in applications and initial symptoms. Our study had the highest reported
initial SSQ-T, and therefore had the greatest potential to decrease symptoms though
habituation. The average decrease in symptoms was 11.8%. Therefore, using an 11.8% or 0.882
decrease per session, the formula to predict habituation with the same virtual environment
configuration is:

1

Where A is the virtual environment configuration factor vector and aorder is the number of the
current session. This formula results in an exponential decay of symptoms over time. The
duration of habituation remains unknown, partially because habituation resulting from a
participant’s first use of a system seems to last longer than habituation resulting from
subsequent sessions, possibly due to the participant’s learning of coping mechanisms.
Therefore, a minimum of three sessions are required to study long term habituation: the first to
learn coping mechanisms, the second to create the baselines for a participant, and the third to
analyze the decrease in symptoms.

34

Table VI: Habituation Values

Measure

Study
Bailenson
and Yee [73]

Toet et al.
[71]

Howarth and
Hodder [72]

MISC Dif.
0-10 (MISC
% Change)

SSQ Dif.
(SSQ %
Change)

SSQ Dif.
(SSQ %
Change)

-1.44
(-16.87)
-1.44
(-6.65)
-1.78
(-14.02)
-1.5
(-16.25)
-1.16
(-10.45)
-1.33
(-11.8)
-1.33
(-10.64)

-0.01
(-0.674)
-0.02
(-1.53)
-0.05
(-3.07)
0.025
(2.93)

Regan
[74]

Ours1

SSQ Dif.
(SSQ %
Change)

HMD, Screen, Rendering
SSQ Dif.
(HMD, Screen, Rendering
SSQ % Change)

Time
1
2
3
4
5
6
7
13, 14

-0.575
(-19.54)

0.0
(2.017)
-5
(-27.7)
-0.02
(-1.71)

1-7
8-12
13-19
-9, 2
(-36, 12.5)

125
1

-7.06, -0.4, -12.15
(-11.54, -32, -36.3)
-10.9, 7.48, -2.49
(-17.09, 28.57, -36)
-3.74, -1.36, -5.236
(-14, -10.74, -13.21)
-7.95, NA, -18.7
(-16.56, NA, -11.19)
-8.58, -0.63, -3.74
(-31.9, -11.72, 8.33)

We had a few participants that started with a score of 0 and later increased, resulting in division by 0 errors. To

compensate for this, small SSQ-T changes (< 4) were ignored, and large changes were assigned a 100% increase.

35

Figure 4: Habituation Data

3.1.2.3 Video Game Play — Untested
Since the repeated use of a system decreases reports of cybersickness, it is possible that visual
stimuli that are similar in nature to the virtual environment might also decrease cybersickness.
Many console video games and some computer games provide visual stimuli that are similar to
virtual reality stimuli. While motion sickness has been reported with video game play [30], no
studies were found that examined if video game play could help create habituation to virtual
environments. This factor was included in a background survey in our experiments.

3.1.2.4 Hats — Untested
The inclusion of hats in our initial susceptibility survey was included for three reasons: Tight
bands may cause headaches [75,76], some participants dislike any additional weight on their
heads, and the center of mass on the head is altered. Because head mounted displays are worn

36

on the head, participants that wear hats may be more tolerant of head mounted displays. No
prior cybersickness specific studies were found for this factor.

3.1.2.5 Duration — Confirmed
It is well known that longer durations of virtual environment use correlate with a higher
frequency and severity of cybersickness [23,29,74,77,78,79,80]. Stanney et al. reported on the
proportion of participants that withdrew early at various duration times [81]. Of the
participants that had a 15 to 60 minute session, 20% of them did not finish their session. Of
those that withdrew, 50% withdrew after 11 to 20 minutes of exposure, 20% withdrew after 21
to 30 minutes of exposure, and 20% withdrew after 31 to 40 minutes of exposure.
The task of comparing the SSQ values from different configurations with different exposure
times is error prone, and there is currently no method available to normalize the effect of such
configuration changes. Therefore, cybersickness models must be based on a percentage change
in symptoms. The values and percent changes in Table VII summarize the above systems. Regan
mentions that participants withdrew due to dizziness, which occurred in our experiments as
well. It can be argued that nausea alone is not a sufficient monitoring measure and therefore
the values were transformed into an SSQ-T estimate for comparison.
The results have been combined and fitted to the linear models shown in Figure 5. The average
SSQ-T effect, where A is the virtual environment configuration factor vector and aminutes is the
number of minutes, is:

2

37

Table VII: The Results of Several Duration Studies
Study

Measure

Stanney et
al. [23]
SSQ-T
So, Lo and
Ho [15] 1
Nausea
ratings
(0-6)
So and Lo
[78]
Nausea
ratings
(0-6)
Stanne et al.
[79]
SSQ-T
Ours
Immersion
ratings (010)
Immersion
ratings (010)
Moss and
Muth [82]
SSQ-T
SSQ-T

SSQ
Estimate

Time in Minutes
15

30

45

60

20

26

29

32

5

10

20

30

0.1, 0.3,
0.1, 0.4, 0.7

0.25, 0.75,
0.3 ,0.8,
1.75

0.9, 1.4,
1.2, 1.6, 2.7

1.75, 2.2,
2.1, 2.4, 3.6

5

10

15

20

.78, .75, .8,
0

1.25, 1.1,
1.4, 0

1.5, 1.25,
1.5. 0.25

1.55, 1.6,
1.5, 0.4

15

30

45

60

16

22

33

35

3

6

9

12

0.25, 0.47,
0.34

0.64, 0.74,
0.86

1.27, 1.02,
1.58

1.98, 1.29,
2.18

15

18

21

24

2.27, 1.51,
2.84

2.57, 1.77,
3.14

3, 2.02, 3.5

3.34, 2.31,
4.02

0

1.5

5

10

1.79

7.38

9.79

13.88

15

20

25

18.85

24.49

32.33

38

30, 34,
26, 48,
47

30, 34,
38, 15

27.26,
21.22,
32.97

Table VII (cont'd)

Keshavarz
studies
HMD,
PowerWall
[82] (pitch,
pitch/roll,
pitch/roll/ya
w) [83] [real
3d, real 2d,
sim 3d, sim
2d][84]

FMS (0-20)

FMS (0-20)

FMS (0-20)

FMS (0-20)

FMS (0-20)

1
0.5, 0.3
(0.4, 1.5)
[1.1, 0.4,
0.9, 1.5]

2
1, 1.9
(0.7, 2)
[1.2, 0.4,
0.9, 1.7]

3
1, 2.8
(1.1, 2)
[2.9, 0.9,
1.5, 1.7]

4
1.9, 3.5
(1.5, 2.7)
[2.8, 0.8,
2.1, 2.1]

5

6

7

8

2.1, 4.1
(1.8, 2.8)
[4.5, 1.5,
2.7, 2.8]

2.1, 4.2
(2.1, 3.3)
[5.1, 1.2,
2.5, 3]

2.5, 4.3
(2.2, 3.8)
[5.5, 2.2,
2.4, 3.4]

3, 6
(2.5, 3.5)
[5.4, 2.1,
2.9, 3.1]

9
3.5, 6
(3, 3.6)
[5.1, 2.8,
3.1, 2.8]

10

11
3, 5.5
(3.1, 3.7)
[5.2, 3.4,
3.9, 3.1]

12
3.5, 6.3
(3.5, 3.4)
[6.1, 3.5,
4.2, 3.5]
16

3.5, 6.3
(2.8, 4)
[5, 2.9, 3, 3]

13
3.4, 6.4
(3.8, 3.5)
[6, 4, 3.6,
3.0]

14
4, 7
(3.9, 4)
[7.3, 3.9,
4.4, 3.8]

15
4.1, 7.2
(4, 3.7)
[7.2, 3.9,
4.7, 3.8]

16
4.1, 8

17
5, 8.8

18
4.8, 9

(20, 21,
30) 2

[38, 40,
44, 37] 2

4, 6.5

1

So, Lo, and Ho suggested that their results after 10 meters per second may be inaccurate due to early withdrawals.

2

The multiplier was averaged from these two studies for the graph.

3.1.3 Mental Attributes
Mental attributes are one of the most difficult individual features to include in susceptibility
tests. They generally require additional tests that are not readily available for use at home.
However, they may beneficial for identifying the limits on virtual environment design and
explaining other poorly understood attributes such as gender.

39

Figure 5: Fitted Duration Values

3.1.3.1 Concentration Level — Probable
Concentration level refers to both the participant’s sense of "being" in the virtual world rather
than firmly identifying it as a simulation, and the participant’s concentration on the task in a
virtual environment. Demand characteristics state that if a person expects a particular
response, that response is more likely to occur. Concentration may diminish demand
characteristics by distracting a participant from undesirable effects such as cybersickness, and
Young, Adelstein, and Ellis [36] and Nichols et al. [6] support this hypothesis. Young, Adelstein,
and Ellis studied participants’ post-immersion scores with and without the administration of a
pre-immersion test. The SSQ scores of the post-test where higher when a pretest was given,
with nausea being the most affected symptom. The change in SSQ-N was only a third as high
without the administration of a pretest.

40

Nichols et al.'s second experiment tested the correlation between participants sensing the
presence of an “interface” and cybersickness where the “interface” was defined as: “…whether
the interaction devices or display quality interfered with the virtual experience, and how well
participants felt they could concentrate on the tasks in the virtual environment.” They found a
negative correlation between the “interface” and cybersickness. Yang and Sheedy reported that
although a stereoscopic movie produces more symptoms, participants also reported a greater
focus on their tasks. [85]
The effect of a participant’s concentration on a task is less certain. On one hand, concentration
may decrease participants' monitoring of their physical state, which would decrease
cybersickness. However, if the task causes physical strain it could increase cybersickness. For
example, if the virtual content is difficult to read, participants may strain their eyes to read it.
This eye strain might then increase cybersickness.

3.1.3.2 Mental Rotation Ability — Conflicted
Because virtual environments are three dimensional, a more intuitive understanding of three
dimensional spaces may affect cybersickness in some participants. Kolasinski reported that
improved mental rotation ability increased cybersickness in females, but this effect was
reversed for males [1]. However, mental rotation ability was a secondary variable under
examination and statistics were not reported. In addition, mental rotation ability differs
between the standard two-dimensional pen and paper test and three-dimensional virtual
environments tests. Parson et al. reported that while males performed better on twodimensional mental rotation tests, there was no difference between males and females when

41

they were tested in a virtual environment [86]. The impact of this factor remains uncertain and
requires further study.

3.1.3.3 Perceptual Style — Possible
Perceptual style includes how a participant views a system. Traditionally, it is how a person
determines what is vertical or "up" in a given situation. Participants who are field-dependent
determine what is "up" from external visual sources. Participants who are field-independent
determine what is "up" using internal vestibular sources. This is frequently tested with the "rod
and frame test" wherein subjects attempt to adjust a rod to true vertical inside a frame that has
been pivoted away from true vertical. Barret and Thornton reported that 18 of their 39 of fieldindependent subjects withdrew early while 5 of their 7 field-dependent subjects withdrew early
[87]. Initially, this implies that being field dependent increases the likelihood of cybersickness
symptoms. However, all the extremely field-independent participants withdrew early. The
authors suggest that this discrepancy may be due to the consistency of the participants’
responses to the rod and frame test rather than field-dependence. Due to the requirement of
special equipment to test perceptual style, it is difficult to include in a cybersickness prediction
model.

3.1.4 Physical Attributes
Physical attributes are somewhat more easily identified than perceptual styles. Eye dominance
is a ten second test that can be easily conducted without any special equipment. However,
factors such as stereoscopic visual ability require longer tests and measures of postural
instability require special equipment. Of the physical attributes, a past history of motion

42

sickness and headaches/migraines hold the greatest promise for inclusion in a cybersickness
prediction model.

3.1.4.1 Eye Dominance — Unlikely
Most people have one eye that is favored over the other eye. Since menu options or interface
controllers are often placed on one side of the screen, one eye may witness a larger proportion
of important details than the other eye. This is particularly problematic for single eye displays,
which may decrease the accuracy for participants with a dominant opposite eye [88]. Most
displays engage both eyes, and Ling et al. found no correlation between cybersickness and eye
dominance [37]. However, due to their use of independent T-tests, their statistical accuracy is
questionable. T-tests are not ideal for cybersickness studies because the results of
cybersickness studies are highly dependent on the individual.

3.1.4.2 Stereoscopic Visual Ability — Unlikely
Stereoscopic systems mimic the natural disparity in images that each eye sees. However,
certain individuals are more sensitive to stereoscopic rendering and other individuals may lack
this ability completely. The ability to process stereoscopic rendering is referred to as
stereoscopic acuity. Hale and Stanney tested each participant's stereoscopic acuity and then
gave them tasks to perform in a virtual environment [89]. While with the participants with poor
acuity tended to travel farther in virtual environments, the authors reported no statistical
difference in the symptoms of the low and high stereoscopic acuity groups. While the Hale and
Stanney study reported no difference, one study is not conclusive regarding the effect of
stereoscopic acuity. However, testing for stereoscopic disparity requires special equipment and
takes time, which makes it less desirable for use in a cybersickness prediction model.

43

3.1.4.3 Postural Instability — Confirmed
Postural instability has been correlated with cybersickness in numerous studies
[31,40,61,90,91,92,23]. As mentioned earlier, postural instability is gaining favor as a
cybersickness symptom measurement. Current postural instability research typically measures
changes in the X and Z axes as shown in Figure 3, with the Z axis being the axis that is most
highly affected by cybersickness. The summary statistics typically used in postural instability
research include the average variability and velocity along a given axis. Typical locations for the
position sensors used in postural instability studies include the base of the neck and torso,
although Bos et al. employed a Nintendo Wii Balance Board [92].
Unfortunately, postural instability's use as a predictive measurement has not been developed,
with only Smart, Otten, and Stoffregen discussing the potential use of discriminate functions
[20]. Different authors report diverse values for their results, and the sampling rate, data
filtering, and position of the sensors vary between the postural instability studies reported in
the literature.

3.1.4.4 History of Headaches/Migraines — Possible
Recently, researchers studying illness during the viewing of stereoscopic movies have found a
correlation between a past history of headaches/migraines and illness [93]. Headaches are
included as a symptom in the SSQ, and there are many similarities between stereoscopic
movies and virtual environments. Therefore, a history of headaches may increase the likelihood
of cybersickness. This factor was included in our updated usability survey.

44

3.1.4.5 History of Motion Sickness — Probable
Cybersickness researchers often assume that the same factors that affect motion sickness also
affect cybersickness, with a potentially different symptomology. Therefore, a susceptibility to
motion sickness may imply a susceptibility to cybersickness. The inclusion of a past history of
motion sickness is not particularly common amongst cybersickness researchers, and the only
researchers that have employed it are Stanney et al. [23], Toet et al. [71], Bos et al. [94], and
Van Emmerik, de Vries, and Bos [49]. Stanney et al. found a correlation between cybersickness
and a past history of motion and carnival ride sickness. Bos et al. and Van Emmerik, de Vries,
and Bos included a past history of motion sickness as a factor, but did not test for its correlation
with cybersickness. Graeber and Stanney tested both past history of motion sickness and
gender, and found a correlation that was due to a past history of motion sickness only [65].
However, Stanney et al. did not find a correlation between their motion sickness history
questionnaire (MHQ) and SSQ-T [17]. This factor was included in our experiments, and we
found a correlation with cybersickness that will be discussed in Section 5.5.2.

3.1.4.6 Weight and Body Mass Index — Possible
Taha et al. [95] and Stanney et al. [23] tested the effect of a participant’s weight on
cybersickness. Taha et al. examined raw weight and found no effect, but Stanney et al. did find
a correlation between a participant's body mass index (BMI) and the SSQ-O [23]. Given that
BMI may not be readily available and has a minor effect on overall cybersickness, it may be
difficult to add BMI to a predictive cybersickness metric.

3.2 Hardware Factors
Virtual environments have a wide variety of configurations due to application and hardware
limitations. There are multiple rendering, tracking, display, and environment options. Of these,
45

the display has become the most recent hardware factor of interest to cybersickness
researchers. The field of view of a screen has shown some of the strongest effects on
cybersickness in the hardware category of factors.

3.2.1 Rendering
Rendering factors relate to how a scene is created and displayed. Stereoscopic rendering, field
of view (FOV) and interpupillary distance have generated the most interest for cybersickness
studies. Many cybersickness researchers who study stereoscopic rendering use eye function
rather than the SSQ as their measurement. Thus, their research is inconclusive regarding how
these features affect cybersickness, because eye function only relates to the oculomotor
category of the SSQ.

3.2.1.1 Stereoscopic Rendering — Conflicted
Stereoscopic movie participants have an anecdotally higher reporting of illness than
monoscopic movie participants. Therefore, a stereoscopic virtual environment may also induce
more incidents of cybersickness than a monoscopic virtual environment. Stereoscopic, biocular, and monoscopic displays all cause strain on the eyes that may affect their normal
function. However, many studies on stereoscopic rendering use eye function rather than the
SSQ as their measurement. How stereoscopic rendering actually affects overall cybersickness is
unknown, but the measure of choice in the literature is heterophoria.
Heterophoria is defined as a deviation of eye position that is normally suppressed with stereo
fusion. This means that one eye moves from its resting position if the other eye is covered or if
stereo fusion has been inhibited by other means (see Figure 6 ). Heterophoria can be measured
by several different methods and is often clinically measured with a prism and a cover-uncover

46

test. For more details on the techniques used to measure heterophoria, see Rainey et al.
(Rainey et al. 1998). Most individuals have a minor amount of heterophoria, but this may be
affected by near or far sightedness. All the following heterophoria results are for horizontally
shifted heterophoria.

Figure 6: Heterophoria
Mon-Williams, Wann, and Rushton found that after 10 minutes of stereoscopic HMD use, all
but one of their participants had a significant increase in heterophoria, while a bi-ocular display
had little effect after 30 minutes of use [16]. Howarth found the opposite result, i.e., no change
in heterophoria with stereoscopic bi-ocular displays [96]. Unfortunately, the stereoscopic interpupillary distance (IPD) used in Howarth’s study was set to an average IPD while the bi-ocular
displays were set to be either lower or higher than the average IPD. Therefore it is uncertain if
their results were due to IPD or rendering. Karpicka and Howarth reported an increase in
heterophoria with a stereoscopic display, but not with a monoscopic display [97]. However,
heterophoria did not affect the level of discomfort of the participants.

47

Mon-Williams and Wann demonstrated the importance of the task when studying the effect of
stereoscopic displays [98]. They found that shifting a focus point along the Z axis resulted in an
increase in heterophoria when focusing at a distance. Yang and Sheedy demonstrated that a
stereoscopic movie produces a greater variance in vergence and accommodation than a
monoscopic movie, and that the vergence and accommodation are farther from ideal in a
stereoscopic versus monoscopic setting [85].
There have been a few studies using the SSQ as a measurement, but the results are unclear.
Hakkinen, Vuori, and Paakka reported higher cybersickness scores for participants playing a
stereoscopic game using an HMD than for those watching a bi-ocular movie using an HMD or
watching a movie with a normal screen [61]. Unfortunately, due to the different visual stimuli
between the conditions, the SSQ measurement effect was uncertain. Ehrlich reported only that
the SSQ-N was higher during stereoscopic versus bi-ocular rendering [39]. Kershavarz and Hecht
compared real stereoscopic video of a rollercoaster, a stereoscopic rendering of a threedimensional model of the same rollercoaster, and a monoscopic version of the real and
modeled rollercoasters [84]. They reported no significant differences between the conditions,
although the real stereoscopic rollercoaster cybersickness scores trended higher.
Experimental results support that there is no difference in cybersickness symptomology using
bi-ocular vs. monoscopic rendering. Kershavarz, Hecht, and Zschutschke reported that there
was no difference between bi-ocular and monoscopic rendering with respect to cybersickness
[82]. They specifically tested whether the differences reported between using a large
monoscopic screen and a bi-ocular HMD were due to rendering or some other factor. They

48

created a device called a "synopter" that converts a large screen into a bi-ocular arrangement.
They found no difference in cybersickness between the participants using the monoscopic and
bi-ocular conditions.
Kolasinski examined IPD correlations with the SSQ-O, SSQ-D, and eyestrain [1], and only found a
correlation with eyestrain.

3.2.1.2 Inter-Pupillary Distance — Possible
To correctly render a stereoscopic virtual environment, the spacing between the rendering
frustums should be aligned with the participant's eyes. The spacing between the pupils of the
eyes is called the inter-pupillary distance (IPD). If the software and hardware alignment differs
from the IPD, the stereoscopic presentation will be flawed. In practice, near-perfect alignment
rarely occurs, as both the software and hardware must align the images correctly to each
individual. Many systems do not provide this option, and when they do, special equipment is
often needed to accurately measure the IPD. The average IPD is 64 millimeters.
Howarth measured heterophoria during his investigation of various HMD configurations [96].
Three different HMDs were tested. The first was bi-ocular with a 60 millimeter interpupillary
distance (IPD), the second was bi-ocular with a 70 millimeter IPD, and the third was
stereoscopic with an IPD of 64 millimeters. Although all three configurations caused
heterophoria, they found no correlation between the level of IPD disparity and heterophoria.
However, there was a trend inward for the first and third HMDs and outward for the second
HMD, which were the smaller and larger IPDs, respectively. Kolasinski and Gilson reported that
the strain on the eye became more severe the more dissimilar the display's IPD was with the
participant's physical IPD [99].
49

3.2.1.3 Screen Distance to the Eye — Untested
When focusing on an object, the eye's vergence and accommodation adjust to that object. In
physical settings, the focus depth for accommodation is the same as the depth for vergence. In
virtual environments, the distance to the screen remains fixed, which means that vergence will
resolve content at a different depth than the focus depth. Theoretically, the more the distance
between vergence and accommodation differs, the greater the potential for cybersickness.
Differences between the physical screen and virtual object locations lead to differences in
vergence and accommodation. There are no published studies regarding this phenomenon, but
this is a factor under consideration in our experiments.

3.2.1.4 Update Rate — Conflicted
Lag is a delay in visual stimuli relative to tracking or other physical measures. Lag is the sum of
such delays due to the acquisition time of tracking sensors, the rendering time for computer
graphics, and the delay of the presentation of the next frame. Because it impacts response
time, lag has been of significant interest in simulator research.
As lag increases, visual and vestibular information begin to deviate from each other. According
to sensor mismatch theory, this should increase cybersickness. However, the research results
have been conflicting. Draper et al. used a panorama with a 48, 125, or 250 millisecond delay
[50]. They found no statistically significant difference between the conditions, but there was a
trend of more cybersickness in the 125 millisecond condition. Kinsella argues that a constant
length of delay is not the issue, rather, it is that the length of the delay can vary over time [100].
Although Kinsella reported that significance was not reached with a regular ANOVA, a faster
variance (1 Hz) in the length of the delay induced more symptoms then a slower variance (0.2

50

Hz). Dizio and Lackner tested 67, 100, 200, and 300 millisecond delays [101]. They reported
symptom increases with the delays. However, a comparison across the delays was not
reported.
The degree to which lag affects cybersickness is uncertain, but a minimum update time of 60 Hz
is a standard requirement for display systems. If all aspects of tracking, rendering, and
displaying are not accomplished within that time, there is too much lag. Even the smallest lag in
the above studies is unacceptable by current standards.

3.2.2 Tracking
There are many different options for tracking an individual. Some researchers choose not to
track participants at all and use a chin rest to guarantee identical visual stimuli. Other
researchers use full 6 degree-of-freedom tracking, as shown in Figure 3, although any one of
the six degrees of freedom may be used independently. Correct tracking is necessary to render
the appropriate image, but all types of tracking add an unavoidable delay due to the
transmission of position data. Tracking also includes the controllers used by a system, which
means that movement can be due to the physical motion of a participant, a controller such as a
joystick, gestures, and/or forced movement due to the application.

3.2.2.1 Method of Movement — Probable
Sensor mismatch theory proposes that a better match between the real world and virtual world
will result in lower levels of cybersickness. Theoretically, a more natural movement should
induce lower levels of cybersickness. There are three sources of movement: movement due to a
participant's body, movement due to a controller, and movement forced by the application.
Most interactive systems use all three sources of movement. For example, the direction a

51

participant faces may control the orientation of the viewpoint, a joystick might specify how fast
a participant moves in space, and collision control may force the viewpoint away from a wall.
Research comparing movement methods supports the use of more natural movement. Jaeger
and Mourant compared the forward motion controlled by a treadmill versus a 2D mouse [59].
The treadmill condition exhibited lower sickness scores than the mouse condition. Similarly,
Chen, Plancloulaine et al. compared two interfaces: a joystick and head/body position [102].
They found that the head/body condition exhibited lower cybersickness scores and performed
better than the joystick condition.

3.2.2.2 Calibration — Untested
Rest frame theory postulates that cybersickness should decrease with virtual environment
movement predictability. What is expected may vary from individual to individual, and a virtual
environment can be calibrated to a specific individual. For example, Jaekl, Jenkin, and Harris
reported that participants typically misjudge their distance traveled in a virtual environment,
reporting that they traveled 1.2-1.4x farther than they were estimated to travel in these
systems [28]. A one-time calibration sequence can be performed to determine individual
estimation bias. No studies directly testing the effect of calibrations on cybersickness could be
found in the literature.

3.2.2.3 Position Tracking Error —Possible
In general, virtual environments try to match real world positions and orientations as closely as
possible. An incorrect alignment can cause sensory mismatch, and therefore cybersickness. One
study indirectly tested tracking errors. Kinsella tested different latency frequencies and
amplitudes [100]. Although the effects were not significant, varying the latency, and therefore

52

the tracking error, consistently increased cybersickness symptoms. Because cybersickness is
highly individualistic and frequently non-normalized, Kinsella’s conventional ANOVA analysis
may have hidden the effect of position error. In addition, position error may interact with
certain characteristics of the application, as some applications are for more tolerant than
others of positional errors as long as the degree of error remains relatively stable. This is
particularly true for orientation as it is used in applications that over-rotate the scene to
prevent people from walking into physical walls.

3.2.2.4 Tracking Method — Conflicted
Some cybersickness studies use chin rests rather than head tracking, so it is reasonable to
consider the advantages and disadvantages of head tracking. Howarth and Finch examined the
nauseagenic effects of HMDs with and without head tracking [103]. They found that head
tracking caused a greater sickness rating within eight minutes of exposure, and that the ratings
gap between the two conditions continued to increase with time. McGee also reported
increased illness with head tracking [104]. These results are in conflict with both movement
studies and sensor mismatch theory, and further study is warranted. In addition, the axes that
need to be tracked are unknown. For example, tracking just the orientation axes may be
sufficient. In addition, the tracked y-position, or the participant’s height, is rarely altered in
many virtual environments.

3.2.2.5 Head Movements — Untested
There are anecdotal reports in the literature that some participants use a coping mechanism of
slowing their head movement when they feel ill. This implies that if a system requires a large
amount of head motion, cybersickness may increase. Tracked and non-tracked studies have

53

been performed, but no study was found comparing systems requiring large versus small
amounts of head movement.

3.2.3 Screen
Screen quality and size have been topics of concern in virtual environments for decades. Early
head mounted displays were large, heavy, and unwieldy, and often required the use of
counterweights. The newer head mounted displays are lighter and more comfortable, but often
have lower-quality optics. Large screens, whether they are projected, utilized across multiple
small screens, or simply consist of a very large television screen, are an alternative to head
mounted displays. However, they restrict the direction in which a participant may face and may
require a participant to stand to achieve the benefits of head tracking.

3.2.3.1 Resolution/Blur — Possible
Low resolution or highly blurred scenes may decrease participants’ use of virtual reality stimuli,
causing them to rely on real world stimuli instead. This would likely cause a decrease in
cybersickness symptoms as there is less of a potential mismatch between the two types of
stimuli. Additionally, most virtual reality systems do not include motion blur, which may result
in a flickering or lag type effect that might increase cybersickness. Conversely, some
researchers state that low refresh rates results in blur due to eye movements and will increase
cybersickness. These factors have not been explored in detail, although Mon-Williams, Wann,
and Rushton cited resolution improvements that potentially increased cybersickness in 1995
[16].

54

3.2.3.2 Horizontal and Vertical Field of View — Confirmed, Partial

Figure 7: Virtual Field of View and Real Field of View Ratios
The field of view (FOV) factor includes the actual screen's field of view and the ratio of the real
screen to the virtual rendering plane. The actual screen's field of view and the virtual rendering
plane are called the real field of view and virtual field of view, respectively. The effect of
variations in the real field of view is well supported, but research on the ratio of the real versus
virtual field of view is conflicted. Figure 7 provides examples of a magnifying one-to-two
condition (a), a one-to-one ratio (b), and a minifying two-to-one condition (c).
Generally, only the physical horizontal or diagonal field of view dimension is given. This is
because the vast majority of displays have 4:3 or 16:9 ratios. Seay et al. reported that the SSQ-T
score was higher for a near 180° field of view than for a 60° field of view, even when not
controlling for level of user control and stereoscopic/monoscopic viewing [105]. Duh et al.
found similar results when testing 30°, 60°, 90°, 120°, 150°, and 180° conditions in two virtual
environments (a city and a black and white radial pattern) [106]. For both scenes, increases in
the size of the field of view generally resulted in significant increases in cybersickness. There
was no statistical difference in the ratings between the scenes. Harvey and Howarth examined

55

this effect with small (39 inch), medium (70 inch), and large (230 inch) screen sizes with
identical luminance for each condition [45], and found that sickness ratings increased with
screen size. Dizio and Lackner reported that halving the field of view also halved the symptoms
of cybersickness, although the exact values were not reported [101]. Ujike, Yokoi, and Saida
also reported that increasing screen size increases illness, but no statistical analysis was
performed [58]. Stoffregen et al. studied video games, and reported a similar doubling effect
[30].

An examination of the above data shows that doubling the field of view approximately doubled
the severity of cybersickness in each study. The actual values are shown in Table VIII. Harvey
and Howarth's study is not included because there was no method to convert their malaise
score, which ranged from 1-4, to an SSQ-T estimate. All the non-SSQ values were transformed
to SSQ-T values for comparison in the chart in Figure 8. To use the results of Toet et al. and Duh
et al., the MISC and difficulty scores had to be converted to SSQ-T values. Because they did not
report SSQ values, it was impossible to calculate a direct estimate using their results. Keshavarz
and Zschutschke's fast motion sickness (FMS) scale is very similar to the MISC scale, but has a
different constant that ranges from 0 to 20 [82]. Therefore, to convert the MISC scale the MISC
values were multiplied by two and then by the average FMS to SSQ-T scaling factor. The
difficulty scores were given the same multiplier since difficulty scores has the same scale as the
MISC. Estimating field of view effects presents the same problem as habituation in that the
starting values may vary dramatically and have a strong effect on the results due to changes in
the field of view.

56

Table VIII: FOV Values
Study

Measure

Seay et al. [102]

Duh et al. [106]

Toet et al. [69]

Stoffregen et al.
[73]

SSQ-Horizontal
FOV

Difficulty
maintaining
posture(1-10)

MISC (0-10)
(real to virtual
FOV)

SSQ (standing
only)

2, 1.5

3.25 (1-1), 1.5 (1-2)

FOV
30
35
43
60

60
13.2 (N), 17.1 (O),
11.14 (T for mono)

77
90
120
150
180

2.5, 2.5

4.25 (1-1), 1.1 (2-1)
100

3.5, 4.5
4.5, 6.1
5, 7
30.8 (N), 31.7 (O),
28.19 (T for mono)

5.3, 7.5

Study

Measure
FOV
13
19
28
35
43
60
77
90
100
180

Ujike, Yokoi, and
Saida [104]
SSQ

Lin et al. [105]

Kershavarz,
Hecht, and
Zschutschke [96]

Ours 1

SSQ non
transformedhorizontal FOV

FMS (0-20)

SSQ

4

27.268

3
3.9
4.5
8.9
0.47
0.82

9

1.49
1.65
1
The results are based on stereo rendering and screen size experiments.

57

37.24

Therefore, when developing the model, the target data used was the percent change in the
effect given the initial field size and the difference in the screen size. The data points within a
study were paired with every other data point within the study. The resulting plane is shown in
Figure 9 and can be expressed using the following formula:

3

Where A and B are virtual environment configuration factor vectors, and

are

their diagonal fields of view. For clarity, the points above the plane are shown in red, while the
points below the plane are shown in blue.

Figure 8: Real Field of View Data

58

Figure 9: Percent Change in SSQ-T with Change of FOV
The effect of the ratio between the real and virtual fields of view is less clear. Until recently,
researchers had assumed that a one-to-one ratio was ideal. Draper et al. had reported that a
two-to-one real-to-virtual field of view ratio resulted in increased cybersickness symptoms
when compared with the standard one-to-one ratio. Toet et al. were the first to propose that a
ratio that is not one-to-one might improve cybersickness symptoms [71]. They altered the field
of view ratio of a projection screen by altering the viewing distance, and reported that a one-toone ratio was associated with a higher severity of cybersickness versus other ratios. They found
a significant increase in sickness ratings using both minifying and magnifying conditions. Moss
and Muth reports a conflicting result, having found no cybersickness effects that were due to
the ratio between the real and virtual field of views [29]. Van Emmerik, de Vries, and Bos argue
that divergent results on the correct ratio may be primarily due to variations in the real field of

59

view. Their smallest field of view (40°) was larger than Draper et al.’s largest field of view (25°).
Given the demonstrated doubling effect of the real field of view, the effect could have been
masked. In addition, Draper et al. employed a panorama image for their environment that lacks
the foreshortening and object occlusion of most virtual environment experiments. Moss and
Muth's ratios were smaller than those of Van Emmerik, de Vries, and Bos or Dizio and Lackner
and included different peripheral occlusions. They did not find an interaction between the
rendering ratio and peripheral vision, but their results may have been impacted by a high
dropout rate and a change in the procedure midway through the study.

3.2.3.3 Weight of the Display — Possible
The weight of the display has been assumed to be a source of increased symptoms in an HMD,
but very few empirical investigations of its effects on cybersickness have been conducted. Only
one study was found that considered the weight of the display, and the conclusion was that it
had no effect on cybersickness [101]. However, it was used as a secondary component and no
statistics were reported. Therefore, the weight of the display was included in our experiment
set. Subsequent analysis showed that the display weight had no effect on cybersickness, so the
possible label was changed to unlikely after experimentation.

3.2.3.4 Display Type — Probable
The literature and anecdotal evidence suggests that the display or screen used has an effect on
cybersickness. In contradiction, the effects in one system are also assumed to occur in a
different configuration. Cybersickness researchers have consistently reported increased effects
in the progression from desktops to large screens to HMDs, although the differences may not
be significant. Anecdotally, although desktop applications are associated with decreased

60

cybersickness, Vinson et al. showed that cybersickness can occur with desktop virtual reality
applications [107].
Sharples, Cobb et al. compared four display systems: HMD, desktop, projection (large screen),
and reality theatre [108]. All four systems resulted in measureable incidences of cybersickness.
They found that the HMD produced higher SSQ-N scores than the other three systems, and that
the HMD’s SSQ-T and SSQ-D scores were higher than those of the desktop system. The
projection screen and reality theater systems had similar results. Although there was not a
significant difference between them, their SSQ-T values were distributed between the desktop
and HMD SSQ-T values. Liu and Uang compared a standard monitor, a stereoscopic monitor,
and an HMD with the addition of varying depth cues [109]. Depth cues are pieces of visual
information that allow participants to estimate the distance and size of objects. The HMD
produced higher cybersickness scores than a conventional monitor.
Smart, Otten, and Stoffregen reported on the percentage of participants that become ill using
different displays [20]. In a moving room, 23% of them became ill, in a space travel simulator,
43% of them became ill, with a projector system, 17% of them became ill, and with an HMD,
42% percent of them became ill. All the experiments included the same sinusoidal motion, but
the visuals used were variable.
One difficulty that is encountered when analyzing the results of different displays is that other
factors tend to shift with the different displays, which confounds the analysis. A projection
screen allows a participant to see the real world next to a display which mimics an independent
visual background. Most projection systems also require a person to stand, while most HMDs

61

and most monitors do not have this requirement. HMDs also typically have smaller real fields of
view and different resolutions. Kershavarz, Hecht, and Zschutschke held many of these factors
constant [82]. Initially, participants viewed the same stimuli on an HMD and a large screen with
the same visual angles and chin rest. The large screen induced higher symptoms than the HMD,
which is in direct conflict with several of the studies that were presented earlier. To try to
determine the source of these differences, they mimicked an HMD's view by masking the view
to just the projection screen, so that the external environment could not be seen. This
experiment revealed no difference in cybersickness symptoms between the masked screen and
the HMD. Therefore, if external factors are held constant, there is no difference in
cybersickness symptoms between displays if no tracking is provided. Our experimentation later
supported Kershavarz, Hecht, and Zschutschke’s claim that the display itself has little effect, and
instead the differences lie elsewhere. Therefore, this factor was relabeled as unlikely after
experimentation.

3.2.4 Non Visual Feedback
The non-visual stimuli in virtual environments are rarely studied. Sound, olfactory, and haptic
feedback may increase the realism of a system, but there has been a lack of research on these
topics. Ambient temperature may be a concern as it impacts the comfort of the participants.

3.2.4.1 Type of Haptic Feedback — Untested
Haptic feedback can increase the realism of a virtual environment. Because more natural types
of movement ease cybersickness symptoms, the inclusion of haptic feedback may also help
decrease cybersickness. No studies directly relating haptic feedback and cybersickness were

62

found in the literature. However, studies in position (sitting versus standing) suggest that
increased tactile feedback decreases symptoms.

3.2.4.2 Ambient Temperature —Untested
Cybersickness can cause sweating, so participants are often more comfortable in cooler
environments. Virtual reality labs are typically air-conditioned, but HMDs produce a small
amount of warmth. Those with motion sickness often desire cool air upon the onset of their
symptoms, but ambient temperature has shown to have no effect on the incidence of motion
sickness [110]. However, sweating often proceeds motion sickness and may therefore be an
effective warning symptom [111].

3.2.4.3 Olfactory Feedback —Untested
Olfactory feedback is rare in virtual environments, but there are olfactory hardware under
current development [112]. In all likelihood, including scents that would make people
nauseated in the real world would make them nauseated in the virtual world, but no studies
have been performed based on this assumption.

3.2.4.4 Audio Feedback —Untested
Sound can be easily added to most virtual environments. Headache-inducing sounds would
likely increase cybersickness, but no research regarding this factor was found in the literature.

3.3 Software
The application may have a substantial effect on cybersickness. An application that places all of
the required objects immediately in front of a participant is very different from one that
requires the participant to wander through the environment. The SSQ values reported solely for
head mounted displays range from 6 [36] to 120 [113]. Software factors tend to be harder to

63

quantify. If a participant has free control over their interaction in the environment, the visual
stimuli may vary considerably in the same application. Therefore, many studies of software
aspects use a chin rest to ensure consistent visual stimuli. Unfortunately, this also creates some
uncertainty of the effect of the software when the results are applied to systems that allow free
movement.

3.3.1 Movement
Movement is the most studied aspect of virtual software as most applications require some
interaction with the environment. The speed of this interaction is well correlated with
cybersickness, although its limits are less certain. The expectation of movement is also
correlated with cybersickness, where participants having more control over their motion exhibit
less cybersickness.

3.3.1.1 Rate of Linear or Rotational Acceleration —Untested
Acceleration in a virtual environment is either controlled by the application or by the
participant through the use of some sort of controller. When the acceleration is controlled by
the application, it is easy to adjust the starting and stopping accelerations. When it is controlled
by the user, the acceleration may be very dramatic, and may even contain instantaneous
stopping and starting. The effects of frequent changes in acceleration have been well studied,
and are discussed in greater detail below. The amount of acceleration may also have an effect
on cybersickness. There are currently no cybersickness studies in the literature on the
acceleration factor; however, studies on seasickness have presented models based on
frequency and acceleration. Increased illness is reported with moderate frequencies, but

64

increasing the acceleration increases illness at all levels [114,115,116]. This effect is illustrated
below in Figure 10.

Figure 10: Seasickness Model from McCauley et al. [116]

3.3.1.2 Self-Movement Speed and Rotation —Confirmed, Partial
Although the acceleration within a virtual environment has not been tested in cybersickness
studies, movement velocity has generated substantial interest in the field. Simply showing a
picture to a participant in a virtual environment does not cause cybersickness [117,118].
Therefore, it is clearly movement through the world that induces symptoms. There have been
several studies on translational and rotational movement that have attempted to ascertain the
speed boundaries that might be associated with a significant decrease in cybersickness
symptoms. Anecdotally, cybersickness decreases when the translational speed is sufficiently
slow or sufficiently fast to blur the scene. However, due to limits on refresh rates and

65

implementations of motion blur, no studies have found a maximum speed associated with a
decrease in cybersickness.
To assure consistent speed through the environment, and therefore the presented visual
stimuli, most studies of navigation speed are done with a chin rest. Studies have shown that it is
not necessarily the direction of movement that is the source of the problem. Chen, Chen, and
So presented a cave scene to their participants with identically textured walls, and then moved
the viewpoint along the vertical, horizontal, and fore-aft axes [119]. No single axis was found to
be worse than the others.
So, Lo, and Ho considered the effect of navigation speed in the Z axis [77]. They tested 8
conditions of forward navigation: 3.3, 4.3, 5.9, 7.9, 9.5, 23.5, 29.6, and 59.2 meters per second.
The same video was displayed to participants at different speeds. The speeds of the other axes
were proportional to the forward translation speed. The resulting SSQ-T values are presented in
Figure 11. Increasing the forward speed steadily increases cybersickness, but duration
eventually masks this effect. A high dropout rate likely affected the results in Figure 11. Using a
repeated-measures analysis, they showed that duration had an effect from 10 m/s to 30 m/s.
Mourant and Thattacheny tested the effects of driving at 25 mph in a city, 60 mph in a rural
setting, or 60 mph on a highway [120]. Although they used an actual vehicle as a control, the
visuals were rendered with an HMD. All the conditions exhibited a significant increase in
cybersickness, but the 60 mph condition was associated with a smaller increase than the 25
mph condition. Unfortunately, this difference may have been due to the either a difference in
speed or visuals. Upper bounds on speed have not been firmly established. The visual blur that

66

is assumed to decrease cybersickness could not be analyzed in the above studies due to the
refresh rate used, as a refresh rate of 60 Hz or higher is theoretically required for blurring to
occur without implementing motion blur in the program. With actual video, the time span that
is required to create each frame creates motion blur, while in computer rendering it is
instantaneous, which prevents blur in the absence of additional rendering features. Motion blur
can be implemented in the applications, but it is rarely done.

Figure 11: Nausea versus Navigation Speed from So, Lo, and Ho [77]
So proposed the Cybersickness Dose Value (CSDV), which is defined as the integral of “spatial
velocity” over time multiplied by display, task, and individual scaling factors [10]. The scaling
factors were later dropped [15]. So, Ho, and Lo subsequently defined their measure of scene
complexity using the luminance frequency along the horizontal, vertical, and radial axes, and
navigation speed [15]. They found a correlation between spatial velocity and cybersickness

67

symptoms when they increased the scene velocity in the Z axis with yaw rotation. In our own
studies, the participants commented that using the run button induced cybersickness.
Rotation is often considered independently to limit possible correlations with other navigation
modalities. In cybersickness research, there is no agreed upon “dominant” single axis of
rotation that causes an increase in illness. So and Lo considered pitch, yaw, and roll axis
rotations [78]. While the SSQ-N scores increased for all the axes versus no rotation, no one axis
was worse than the others.
Duh et al. examined roll axis rotation with oscillating frequencies of 0.05, 0.1, 1.2, 0.4, and 0.8
Hz in two virtual environment scenes [121], a tropical hillside scene and a black and white radial
pattern. The cybersickness score increased as the frequency decreased in both virtual
environments, and the 0.8 Hz rotation condition showed no statistical difference from the
baseline. They then extended the experiment to include physical motion. In their second
experiment, they rotated the individual in the same axis as the visual stimulus, although not at
the same frequency. The resulting perceived frequency was 0.2 Hz or 0.06 Hz on average. They
found the sickness ratings to be significantly higher for the 0.06 Hz average rotation. Chow et al.
examined oscillating frequencies of 0.1, 0.2, 0.4, 0.6, 0.8, and 1.0 Hz and found that
cybersickness symptoms decreased with increasing frequency, but no formal statistics were
reported [122]. Chen et al. presented a checkerboard tunnel to participants and oscillated the
scene to imply forward motion at specific luminance frequencies [123]. They tested 0, 0.05, 0.1,
0.2, and 0.8 Hz frequencies with the velocity held constant and found a significant difference

68

between 0.05 and 0.8 Hz for vection ratings, with 0.8 Hz having a lower score. This is similar to
navigation speed in that excessive movement decreases cybersickness.
The effect of rotation is worsened if the rotation occurs over more than one axis.
Bonato, Bubka, and Palmisano examined simultaneous multiple axis rotations [124]. A checker
board room was presented to their participants along with a stationary virtual frame. They then
oscillated the room around one or two axes. They found higher cybersickness ratings under the
two axis rotation condition. Keshavarz and Hecht further explored the effects of multiple
rotation axes [83]. They created a virtual rollercoaster with pitch, pitch and roll, or yaw, pitch
and roll conditions. The cybersickness symptoms increased during the pitch-only condition, but
slowed after approximately five minutes. The multiple axes conditions resulted in more
cybersickness.
In all likelihood, translational and rotational speed affect cybersickness, but more studies are
needed to firmly establish the patterns of this effect. Also, while rotation studies and sea
sickness studies are fairly consistent, most applications do not present continual rotation. The
variance, axes, and average speed may be needed for a complete model, but most applications
have most of their motion in 1 to 3 axes. Therefore, using only the axes with the highest
amounts of motion may be an acceptable estimate for modeling. The models created here are
tentative.

The

values

reported

from

the

above

studies

are

shown

in

Table IX.
Mourant reported their results as SSQ-O rather than SSQ-T values. To allow a model to use their
results, SSQ-O must be converted to SSQ-T values. Because they used a visual similar to So, Lo,

69

and Ho, their overlapping conditions should have similar SSQ-T values. Therefore, Mourant's 25
mph condition was estimated as the average of So's studies at 9.5 m/s (21 mph). The remaining
SSQ-O values were scaled proportionally. The resulting model is shown in Figure 12, where amph
is the forward translation speed in miles per hour. As there are only a few data points, this
formula is tentative:

4

Table IX: Speed Values

Measure
Speed
3.3 m/s
4.3 m/s
5.9 m/s
7.9 m/s
9.5 m/s
25 mph
60 mph

So, Lo, and Ho [77]1

Study
Mourant and Thattacheny [120]

So, Ho, and Lo [15]2

Nausea ratings (0-6)

SSQ-O

SSQ-T

1.75
2.2
2.1
2.4
3.6

31
33

48
15
7 highway, 9 rural

1

Only up to 10 ms were recorded because dropout and duration may have affected the results above that threshold.

2

This is from the same experiment as So, Lo, and Ho [77], but uses a different measure.

Because rotation studies have examined frequency rather than velocity, only general
suggestions can be inferred from them. Rotation speed should be either extremely slow or
moderately fast. Very high speeds are unlikely to have an effect. The caveat is that jittering the
display also tends to anecdotally increase cybersickness symptoms. These frequencies would be
well above what was tested, but are inferred to have an effect in Kiryu et al.'s study [125].

70

Figure 12: Fitted Speed Values

3.3.1.3 Vection —Conflicted
Vection is the participant’s feeling that the environment is moving away from oneself. Previous
research has suggested that vection is correlated with cybersickness, and vection is commonly
cited as the reason why wide screen displays increase cybersickness [126,127]. This also
explains why static imagery has little effect on cybersickness [117]. Hettinger, Berbaum, and
Kennedy indicated that participants who reported vection were much more likely to report
feelings of motion sickness [126]. They did not find a linear effect, and therefore used a
dichotomous "yes or no." Webb and Griffin reported a less clear relationship [127]. The illness

71

ratings of the high and low vection conditions were similar. Ji, So, and Cheung reported that
illness can still occur in the absence of vection [128]. The main issue with vection in
cybersickness research is that it relies on participant reports because there is no objective
measurement of vection. At best, this implies that backwards motion should be limited to
decrease cybersickness symptoms.

3.3.1.4 Altitude above Terrain —Probable
In flight simulator research, the altitude above the terrain can dramatically affect what is seen.
In cybersickness research, Watanabe and Ujike compared walking on the ground (a hilly
condition) with a planar fly-over of a hilly virtual environment [43]. Watanabe and Ujike found
an increase in the SSQ-D score for the height-varying mode over the planar fly-over condition,
but the SSQ-T scores were statistically identical. The fly-over condition had the effect of
flattening the environment, which has been suggested to decrease cybersickness. Golding,
Doolan et al.’s initial environment conditions were flat environment with an upright, inverted,
and abstracted scene with no orientation cues. There was no cybersickness effect among these
conditions until they included vertical cues in a second experiment [129]. Taha et al. performed
a pilot study of 10 individuals and reported that taller (higher altitude) participants experienced
fewer cybersickness symptoms [95].

3.3.1.5 Degree of Control —Confirmed
Research supports the seemingly obvious notion that shaky motion is more likely to induce
cybersickness than smooth motion. Shaky motion has been used as a stimulus in cybersickness
research [42]. In addition, to intentionally increase the likelihood of illness, a virtually swayed
walk may be added [71]. Theoretically, this means that applications should use natural motion

72

whenever possible, and if applications need to move the camera independently, they should
move it in a smooth manner.
Stanney and Hash tested three movement conditions: navigation control with active (six
degrees of freedom) control, active-passive control (three degrees of freedom with the chosen
axes dependent on the given task), and passive (no) control [130]. They reported that all the
SSQ category scores for the active-passive condition were lower than the passive control
condition. The active condition also produced lower SSQ-O and SSQ-T scores than the passive
control condition. Stanney et al. considered navigation control alongside the effects of
duration, scene complexity, and performance [79], and found that six-degrees of freedom
induced higher levels of sickness.
Dong, Yoshida, and Stoffregen [91], Chen et al. [4], and Dong and Stoffregen [90] also
considered control in their environment in the most extreme form. That is, if the participant
had any control at all. An HMD and a chin rest offered the advantage of guaranteed identical
visual stimuli so that recorded scenes from a “driver's” game play could be shown to a
"passenger." In all instances, the passive “passenger” reported higher sickness ratings than the
“driver.”

3.3.2 Appearance
The appearance of an application is the hardest cybersickness factor to quantify, particularly if
participants are free to roam the virtual world. Global visual flow and scene content/complexity
have been the focus of most of the research on this factor. While both of these appearance
factors were found to have an effect on cybersickness, there have been few consistent results
reported in the literature.
73

3.3.2.1 Screen Luminance —Untested
Some researchers have controlled for screen luminance in their research [45] and found that
bright systems may cause more eyestrain, or worse, make flicker noticeable. The human flicker
threshold decreases with dimmer lighting and, to some extent, color [131]. However, research
on the screen luminance factor’s effect on cybersickness has not been published.

3.3.2.2 Color —Possible
The human visual system is more sensitive to certain colors. The most sensitive wavelengths are
530–538 nm (green), which are primarily sensed with the γ cone and partially sensed with the ρ
cone [132]. The least sensitive cone is the β cone, which primarily senses blue. In theory, having
a virtual environment primarily composed of a sensitive color could increase eyestrain or the
SSQ-O score. The only study on the cybersickness effects of color was done by So and Yuen
[133]. They found no effects, but only altered approximately one fourth of the scene with two
different colors, one of which was a neutral color. There have been reports of a color effect in
motion sickness studies, namely, that the inclusion of color imagery increases illness compared
with black and white imagery [134]. The human eye also has more difficulty with the extremes
of the visual spectrum and with excessively high contrast in text colors and backgrounds [135].

3.3.2.3 Contrast —Untested
Given the same illumination level, the contrast in natural settings is lower than that of manmade settings, and the contrast in man-made settings is lower than that of typical virtual
environments. For example, a park displays varying shades of green while a supermarket may
display purple, red, blue, and yellow within a span of a few feet. This increase in contrast can
cause extra eye strain; however, the relationship of contrast to cybersickness is unknown.

74

3.3.2.4 Scene Content or Scene Complexity —Probable
Scene content may also affect sickness ratings, but exactly how particular content affects
cybersickness is poorly understood. Most studies consider one particular aspect of the content
without addressing its relation to the larger system.
Jaeger and Mourant considered the level of detail and found a trend of lower SSQ-Ts with lower
levels of detail [59]. Liu and Uang report conflicting results [109]. They used planar or three
dimensional models on shelves with different displays. Their three dimensional model condition
produced an increase in cybersickness regardless of the display used, but there were
interactions in some of their data.
The fact that cybersickness symptoms decrease with lower levels of detail is similar to the
impact of resolution and blur. If a person cannot determine what an image is, there is too little
information to cause sensory mismatch. So et al. present references for low, medium, and high
degrees of scene complexity, as shown in Figure 13 [15]. Low complexity has no textures and
presents information about the surrounding environment only. Applications that have only a
few objects centered in a blank room also have low scene complexity. In medium complexity
systems, there are some textures, and the surrounding environment is clearly 3D. High
complexity scenes are mostly textured, but lack shadows and use mostly generic textures.
Realistic systems, as illustrated by van Emmerik, Vries, and Bos [49], either use actual video or
are fully textured with non-generic textures and include shadows.

75

Figure 13: Scene Complexity (a-c) from So et al. [15] and (d) van Emmerik, Vries, and Bos [49]

3.3.2.5 Global Visual Flow —Probable
Global visual flow is closely related to scene complexity. Global visual flow is the change in
luminance or color of a scene over time. Global visual flow is implicit in rotational and
translational speed studies and in realism studies, but only a slight difference between detailed
versus black and white imagery has been reported [121]. The Cybersickness Dose Value (CSDV)
by So is, in one sense, a global visual flow metric [10][15]. So found a correlation between the
CSDV measure and cybersickness symptoms.

3.3.2.6 Orientation Cues —Possible
One aspect of most virtual environments is that the implied "up" direction agrees with visual
and vestibular cues. Orientation cues also affect global visual flow. More highly realistic
environments provide more clues about orientation. Golding et al.'s pilot experiment used a

76

panorama that was tilted, inverted, or abstracted to remove any horizontal or vertical cues, and
reported no effects based on these conditions [129]. They restructured the conditions so that
the visuals consisted of a real scene from Westminster Bridge rather than a perfectly flat
landscape, and found that the upright condition induced significantly higher cybersickness
symptoms than the inverted condition. They suggest that the inverted condition was likely
considered to be so unlikely by the participants that it was not used to determine the "up"
direction. Their study also implies that vertical cues are of greater importance than horizontal
cues.
Given that their study examined a completely unlikely versus a real environment, their research
is inconclusive regarding which orientation cues affect cybersickness.

3.3.3 Stabilizing Information
Stabilizing information has been gaining interest as a cybersickness factor due to some
promising results of pilot studies. Independent visual backgrounds, the ability of participants to
see part of the real world, or the inclusion of screen stable elements have consistently been
shown to decrease cybersickness.

3.3.3.1 Focus Areas —Probable
The focus areas factor relates to where a participant should focus in the scene. Mon-Williams et
al. measured heterophoria during the testing of five different gaze angles, or the vertical angle
of gaze relative to the ear-eye line (drawn from the center of the ear to the center of the eye)
[136]. The angles were 20° above, and 0°, 20°, 40°, and 60° below the ear-eye line in an HMD.
They found that post-immersion heterophoria varied with the gaze angle, with the smallest
change occurring at 34° below the center ear-eye line on average, which corresponds closely

77

with looking directly forward. While there is a fair amount of individual variability, in general,
the farther the gaze was from this line, the greater the measured heterophoria.
Diels, Ukai, and Howarth also considered whether the focus location on a screen could affect
cybersickness symptoms [137]. They displayed a scene consisting of an expanding-contracting
random dot pattern oscillating around the center of a screen along the Z axis. When the focus
point shifted over time, cybersickness values increased when compared with a static centered
focus point or the lack of a set focus point.

3.3.3.2 Ratio of Virtual to Real World —Untested
Augmented reality has had almost no reporting of cybersickness, as most of the virtual world is
real. The degree to which the world needs to become virtual before cybersickness becomes an
issue is unknown. This factor is directly related to independent visual backgrounds, which
include parts of the real world in virtual environments.

3.3.3.3 Independent Visual Backgrounds —Confirmed
Independent visual backgrounds (IVB) are elements of the visual field that remain stable
relative to the participant. They may be arranged by either having an object fixed relative to the
participant’s position on the display or by having an object that is external to the virtual
environment yet visible though the display, as presented in Figure 14. IVBs provide a real world
expectation that is enforced in relation to the virtual world due to their configurations. The
results have been very promising with respect to correlations between IVBs and cybersickness.
Duh et al. placed a grid at one of three different depths in front of the participant and at two
different intensities: bright or dim [138]. The grid improved cybersickness symptom ratings, and
there was no difference between the grid conditions except for brightness. In another study by

78

Duh et al., they tested dim, bright, or invisible grids overlaid on a rotating virtual environment
scene at two different frequencies [139]. The grid decreased cybersickness, but there was
significant interaction between the grid and the roll frequency. Moss and Muth reported that
simply removing the peripheral occlusion molding from HMDs decreased cybersickness [29].

Figure 14: Independent Visual Background
Prothero et al. created an IVB through the use of a partially occluded HMD, in other words, a
virtual environment superimposed on the real world [32]. Their first experiment reported lower
SSQ-T scores and fewer postural stance breaks under the IVB condition. Their second
experiment, which added an identification task to the independent background condition,
showed no difference in the SSQ-Ts, but fewer stance breaks were noted under the IVB
condition. These conflicting results could mean that the additional focus on the background or
the elimination of a participant from part of the analysis had an effect. The median scores on
the SSQ were identical in the two studies, but their standard deviations differed.
Kershavarz, Hecht, and Zschutschke indirectly tested an IVB by masking a projection screen so
that the external environment could not be seen, similar to an HMD [82]. Eliminating the real

79

world stimuli resulted in the same level of cybersickness using a projection screen and an HMD.
These studies imply that even less obtrusive IVBs decrease cybersickness. Kershavarz, Hecht,
and Zschutschke reported the rather unexpected result that the PowerWall they used induced
higher cybersickness symptoms without masking. There are a few possible explanations for this
result. An IVB enforces expected stimuli. This is normally the real world, but in very familiar
settings such as a virtual car, the IVB enforcing the real world opposes expectations. In addition,
different participants were employed in testing the two displays, and they may have had
different susceptibilities. The PowerWall had a 3:4 ratio of male to female participants while
the HMD had a 15:16 ratio of male to female participants, and females frequently posses higher
susceptibility.
Since the IVBs vary dramatically it is difficult to create a model using the IVB factor. The above
studies’ reported values are shown in Table X. Since an IVB is meant to enforce real world
expectations, the values from Kershavarz, Hecht, and Zschutschke are "yes" for the HMD and
"no" for the PowerWall because the HMD hides the unexpected stable room in a car video.
The average percentage drop was calculated directly without transforming all the measures to
the same scale. Since Duh et al.'s [103] studies showed no difference across grid conditions, the
grid conditions were averaged for the sake of comparison. Since Prothero et al. reported the
median, which has a different meaning than the mean, only the stance breaks are used in the
calculation. In the future, improved models are likely if IVB studies can be combined with
"amount of the real world" studies. The average percentage drop in the studies resulting from
enforcing the expectation to not enforcing the expectation is 30.6%. If we assume that the

80

differences in our screen size and HMD studies are due to the IVB effect of seeing the real
world, this average drop becomes 30.1%. More formally, where aIVB is whether configuration A
uses an IVB or not:

5

Table X: IVB Values
Study
Duh et al.
[138] Bright
(Dim)

Measure

Duh et al1
[139] Bright
(Dim)

Difficulty
Difficulty
Maintaining
Maintaining
Posture (1Posture (1-10)
10)

Prothero et
al. [32]

Moss and Muth
[29]

Kershavarz,
Hecht, and
Zschutschke
[82]

SSQ-T
median
(Stance
breaks)

SSQ-T Handrail
(SSQ-T, no
handrail)

FMS (0-20)

15, 15
(4.1, 1.6)
18.7, 18.7
(9.1, 2.6)

6, 4.8, 4.4, 6.1
(5, 3.9, 7, 2.1)
5.1, 6.6, 8.5, 6.9
(5.5, 7.5, 5.4, 8.8)

Expectation

1

Yes

3.2, 2.7, 3.1
(3.3, 3.1, 3.8)

0.85
( 0.9)

No

5.1

1.26

3.8
9

A frequency of 0.8 Hz has consistently shown little to no effect on cybersickness in rotation studies, therefore these

values were dropped from the model.

3.3.3.4 Sitting versus Standing - Confirmed
Some virtual environments are typically used with a standing position while others are typically
used in a seated position. This may mask the differences between two configurations. This is
one theory behind why monitor configurations are anecdotally associated with less
cybersickness. Merhi et al. reported a strong cybersickness effect under seating versus standing
conditions [140]. All of the standing participants withdrew, but there was no difference in the
cybersickness ratings between those that withdrew early under the sitting versus the standing

81

conditions. Therefore, seating participants should decrease the frequency of cybersickness.
Increased tactile information seems to help decrease cybersickness in general. In a study by
Moss and Muth, the procedure was altered midway to include a hand rail in order to reduce a
high dropout rate [29]. Stoffergen et al. studied the effect of video games on postural stability
and reported the opposite results; their standing participants had lower cybersickness
symptoms [30]. Peculiarly, other measures such as mean time to illness and percentage of
participants stating illness were slightly worse under the standing condition. Although there are
few experiments that specifically examine the sitting versus standing factor, the results are so
dramatic that the factor is confirmed, however, the level of experimentation done on sitting
versus standing conditions does not permit the modeling of the effect.

3.4 Summary of Factors
The factors that have been shown to be reliable and consistent predictors of cybersickness are
listed in Table XI. The factors in this table do not all contain sufficient data to allow for a fitted
model, but the literature and experimental results are consistent. Habituation and field of view
are special cases in that part of their effect is established, but other components are uncertain.
Novel sub-factors have been tentatively relabeled as possible. The effect of the horizontal or
diagonal screen field of view is well established, but the virtual to real ratio and the vertical
fields of view are labeled conflicted and untested, respectively. Habituation has only been
confirmed under the same configurations and applications. Table XIV lists the factors that have
no clear indication of effects, and no claims at all can be made of the untested measures listed
in Table XV. Of these, the tracking factors are likely to be the easiest to control and merit

82

further study. While gender has received much attention, its effect may be found though other
factors that may have better correlations with cybersickness [65].
The probable and possible factors are listed in tables Table XVII and Table XVIII, respectively.
They lack sufficient data to determine their levels of cybersickness effects. Some of these
factors are difficult to determine or control, which limits their usefulness in prediction
measures. These factors have been removed and placed into the “elimination” Table XVI. The
factors that are difficult to determine or too time consuming for practical commercial
configurations are also placed into Table XVI.
Following our experimentation, several factors had their labels altered. A few untested factors
were included as experimentation helped to explain conflicting results. For example, the effect
of displays was found to be due to other underlying factors. These factors have been placed
into their respective tables according to their labels after experimentation.
Table XI: Confirmed Factors
Individual
Experience

Hardware
Screen

Habituation

Horizontal field of view

Duration
Video game play
Postural stability
Physical Attributes
History of
headaches/migraine

Software
Stabilizing Information
Independent visual
backgrounds
Sitting versus standing

Tracking

Movement

Method of movement

Self-movement speed
Degree of control

Demographics
History of motion sickness

83

Table XII: Probable factors
Individual

Hardware

Software
Movement
Altitude above terrain
Appearance
Scene content or scene
complexity
Global visual flow
Stabilizing Information
Focus area

Table XIII: Possible Factors
Individual
Experience
Experience with other virtual
environments

Demographics
Age

Hardware
Screen

Software
Appearance

Resolution/Blur

Color

Ratio of real to virtual field of
view
Vertical field of view

Self-movement rotation

Tracking
Position tracking error

Stabilizing Information
Ratio of virtual to real world

Orientation cues

Rendering
Inter-pupillary distance

Table XIV: Conflicted Factors
Individual
Experience
Experience with real-world
task
Vision correction

Hardware
Tracking
Tracking method
Stereoscopic rendering

84

Software

Table XV: Untested Factors
Individual
Demographics

Hardware
Tracking

Ethnicity

Calibration

Software
Movement
Rate of linear or rotational
acceleration

Head movements
Non Visual Feedback
Type of haptic feedback
Ambient temperature
Olfactory feedback
Audio feedback

Appearance
Screen luminance
Contrast

Individual
Experience
Hats

Hardware
Screen
Weight of display
Display type

Software
Movement
Vection

Physical Attributes
Eye dominance
BMI

Rendering
Update rate
Screen distance to the eye

Table XVI: Eliminated Factors

Demographics
Gender
Mental Attributes
Mental rotation ability
Concentration level
Perceptual style

85

4

General Procedure and Hypotheses

A detailed examination and summarization of the literature allowed us to prioritize
cybersickness factors. However, it also revealed some discontinuities in the existing
cybersickness knowledge. Individual susceptibility has not been well considered in
cybersickness research, and there are several conflicting or ambiguous factors. For this reason,
we conducted a series of experiments, including a search for individual factors and a
clarification of display factors, while holding known factors constant. These results were then
combined to create predictive models as a means to compare the results across different virtual
reality systems and develop future guidelines. The sets of experiments we performed include
HMD weight (Our HMD), screen size (Our Screen), and stereoscopic rendering (Our Render).
Anecdotally, the weight of an HMD is assumed to increase cybersickness. Modern displays
range in weight from 220 grams (eMagin 3 Visor) to 440 grams (Oculus Rift). The center of
gravity of the head is altered, since the weight is primarily in front of the display, and the
tightness of the bands may cause headaches [75,76]. Only one study was found in the literature
that took HMD weight into consideration [101]. The study found no effect on cybersickness, but
given that its focus on HMD weight was secondary and the lack of reported statistics, the effect
of this factor remains uncertain. The HMD weight experiment we conducted directly considered
this factor.
Conventional desktop monitor displays have been associated with a trend of decreased
cybersickness compared with other types of displays [108,109]. At a fundamental level, there is
little difference between a monitor and a large screen display. Both are flat, stationary displays

86

that can use a nearly identical interface. In practice, there are several differences. Large screens
usually have larger fields of view and typically engage standing participants, while participants
are usually seated in front of monitor displays. Other common differences include ambient
lighting, angular momentum, resolution, head tracking, and refresh rate. The screen size
experiment directly considered this factor.
From the first two experiments, we concluded the differences reported between display types
are not due to the weight of the display or the size of the screen. The remaining question was if
there would be a change in cybersickness between a large screen and an HMD when
participants are standing and had the same average field of view in both. There are a few
fundamental differences between these displays. An HMD blocks out the outside world making
the scene appear to float in space, while the room is still visible to the participant with a large
screen. This may yield an IVB-like effect. The interaction also differs slightly. An HMD alters the
rotation of the view with head movements, while head tracking for large screens adjusts the
display frustum, but not the rotation. This was considered in the display type's analysis.
Most studies of stereoscopic rendering use eye strain or heterophoria as a measurement.
However, the stereoscopic rendering effect is highly individualized, and the extent to which
heterophoria encompasses cybersickness is unknown. A quick Internet search for "movie
motion sickness" reveals many people complaining of difficulty with 3D movies, so it is
reasonable to assume that issue may be transferred to stereoscopic virtual reality. The
stereoscopic rendering experiment directly considered this factor.

87

Cybersickness is known to be highly individualistic and therefore paired statistical tests are
preferred in cybersickness studies. Despite this fact, individual demographics and susceptibility
are rarely included in cybersickness research. Moreover, the degrees to which individual factors
influence cybersickness are largely unknown. If a few questions can determine a participant's
susceptibility, they can be used to normalize the cybersickness results across populations, and
give a means to dynamically adjust an application to a participant's susceptibility. Several
surveys were utilized throughout our HMD weight, screen size, and stereoscopic rendering
experiments. These surveys asked for background information from the participants and are
explained in further detail below.

4.1 Individual Factors
To examine individual factors, background surveys were given to all the participants. The
surveys are presented in Appendix A. The "usability survey" included both demographic factors
and usability factors. A past history of motion sickness, age, and gender were included in
Golding's MSSQ [141]. Several questions were specifically designed to address usability. These
questions were primarily used to avoid the participants’ discomfort due to the display, and to
gain insight into the participants’ interest in these systems. The self-limiting immersion rating
(original usability survey question 9 or revised survey question 7) was included to gain an
understanding of a participant’s tolerance for cybersickness
An early analysis of the original usability survey showed that several factors did not meet
statistical significance. Thus, they were removed from the revised usability survey. Video game
use displayed a trend, and this question was reformatted to gain specific information regarding
this trend in the revised usability survey.

88

Table XVII. Individual Factors Question Location, Reason, and Hypotheses
Factor
1

Hat Use (6 O)

2

Video game use
(general) (1 O)

3

Video game use
(specific) (1 R)

4

Headache/migraine
history (2 R)

5

Vision correction (7 O
and 3 R)

6

7

8

9

10

11

Prior use of head
tracked virtual reality
(2 O and 3 R)

Reason for Inclusion
Tight band around the head can
cause headaches, which may
increase cybersickness
Video games have similar visual
stimuli as virtual environments and
therefore create a habituation
effect
Video games displayed a trend and
interacted with the MSSQ so
clarification was required
Solimini et al. [93] reported a
correlation between theater
sickness and a past history of
headaches and migraines
Vision correction adds another layer
of refraction and may shift optical
placement, increasing eye strain
Prior use of VR, even in different
configurations, may lead to a
habituation effect

3d display use (4 )

3D displays have similar visual
stimuli as VR and may lead to a
habituation effect

Prior 3d display
discomfort (5 O)

Those that have a prior history of
issues with 3D displays may be
more likely to have trouble with
cybersickness

O

Prior history

Age

M

Gender M

M

Being sensitive to one type of
motion may increase sensitivity to
other types of motion
Prior studies by Arns and Cern [55]
and Park et al. [56] displayed a
correlation with age
Prior studies include conflicting
reports about females having a
higher incidence of cybersickness

89

Hypothesis
Those that wore hats would
have a decreased incidence
of cybersickness
Increased game play will
decrease cybersickness
Increased game play in
certain genres will decrease
cybersickness
Those with a history of
headaches will be more
susceptible to cybersickness
Vision correction will affect
cybersickness
Those that have previously
used VR will have a lower
incidence of cybersickness
Those that have previously
often used 3D displays will
have a lower incidence of
cybersickness
Those that have had a prior
difficulty with 3D displays
will have increased
cybersickness
Those that are more
sensitive to motion sickness
will be more sensitive to
cybersickness
Younger participants will be
less susceptible to
cybersickness
Gender will have no effect
when holding MSSQ
constant

During experimentation, Solimini et al.'s [93] released a study that examined the relationship of
headache and migraine history and symptoms while watching 3D movies, so questions related
to those factors were added. The factors under consideration, their reason for inclusion in the
surveys, and their associated hypotheses are listed in Table XVII. The original usability survey is
labeled with O and the revised survey is labeled with R. The information from the MSSQ survey is
labeled with M.

4.2 Hardware and Software Factors
The most problematic of the conflicting factors is the effect of using different displays. All
virtual reality systems require the use of a display. Therefore, knowing which displays are more
likely to induce cybersickness is critical for creating guidelines. Prior studies have shown trends,
but not consistent results relating to the effect of displays on cybersickness. One issue is that
several other factors tend to shift with the display. For example, a monitor display is almost
always used by a person who is seated, while an HMD is almost always used by a person who is
standing. An HMD weighs more than the glasses that are required for 3D monitors or large
screen displays. A monitor has a smaller field of view than a large screen display. Standing
versus sitting and field of view are both known cybersickness factors with sizable effects.
Stereoscopic rendering has been the subject of considerable study, but most of the results have
been reported in terms of eye function rather than cybersickness, and the results are
inconsistent. We do not know if heterophoria correlates with cybersickness. Therefore, studies
using heterophoria as a measure cannot be directly applied to cybersickness studies.

90

We completed three sets of experiments to analyze the following factors: HMD weight, screen
size, and stereoscopic rendering. The HMD experiment was included to confirm or deny the
effect of HMD weight on cybersickness. Anecdotally, the additional weight of HMDs has been
blamed for their association with increased cybersickness. In the screen size experiment, the
field of view was held constant, and the participants remained standing for both screens, but
the average screen size was changed. The field of view was also kept the same as that of the
HMD experiments, so the large screen versus HMD analysis could be performed. The final
experiment compared bi-ocular with stereoscopic viewing. The field of view was increased and
the application was rendered for either stereo or bi-ocular viewing. Bi-ocular viewing was
chosen over monoscopic viewing to keep the illumination level constant. Bi-ocular viewing has
shown no difference with respect to cybersickness compared to monoscopic viewing [82], so
this choice is unlikely to alter the effect. The factors under consideration and their respective
hypotheses are available in Table XVIII.
Table XVIII: Experiments Rational and Hypotheses

1

2

3

4

Factor
HMD
weight
Screen Size

Display
Type
Stereo
versus biocular

Reason
Anecdotally, HMD weight has been
blamed for increasing cybersickness
Monitors show a trend of decreased
cybersickness versus large screens, but
other known factors tend to shift with it
HMDs are often associated with increased
cybersickness. By holding the application,
interaction, field of view, and standing
position constant, that may or may not
remain true
Stereo displays tend to cause eye strain,
but we do not know how that affects
cybersickness

91

Hypothesis
Heavier displays increase
cybersickness
Screen size will have no effect
HMDs will increase
cybersickness, holding
interaction, field of view, and
standing position constant
Stereo displays will increase
cybersickness

4.3 Environment
Testing utilized a virtual environment in the Media and Entertainment Technologies Laboratory.
The virtual environment was presented using Vizard 3.0 with 3-sample antialiasing and a 4:3
aspect ratio. The computer had a 2.58Hz Intel Core 2 Duo E4700, 4GB RAM, and a Quadro FX
3700 graphics card. Tracking was done with an Intersense IS900 which has a specified latency of
4ms. Formal tracker-to-display latency calculations were not performed. If stereo was used, the
software IPD was 6 centimeters. Two different display technologies were employed during
testing. A Glasstron LDI-D100B HMD, with an 800X600 resolution, fixed interpupillary distance,
and a 35 degree diagonal FOV and an stereoscopic capable InFocus DepthQ projector with a
maximum resolution of 1600 X 1200 and a maximum of a 120 Hz refresh displayed using a rearprojection screen. The active shutter glasses had a maximum refresh of 100 Hz, and therefore
the project was set to a 100 Hz refresh as well.
The test virtual environment consisted of a set of five to nine rooms, two of which were mazes,
with objects placed on pedestals or hung on the walls as shown in Figure 15. The game was
essentially a treasure hunt for the objects listed on the left-hand side of the screen. Objects and
pictures were set to scale, if possible. Participants were timed, but locating a star object
removed 30 second from their total time and was counted as an item found. Certain doors
were locked and the participants were required to locate a matching colored key to open the
door. To ensure progression, the doors unlocked automatically after a specific time period had
elapsed.

92

Figure 15: Screenshots from the Application

93

To provide navigation help in the mazes, a float button was added to allow the participants to
rise 2.5 meters to look over the top of a wall. They could not move forward while they were
floating. The normal walk speed was a maximum of 2 m/s and the rotation was a maximum of
π/3 radians/s. A run button doubled the maximum forward/backward speed. The arrangement
of the buttons is illustrated in Figure 16. The head-tracked position and orientation was added
directly on top of the controller position and orientation. This means that “forward" was always
the same direction in the real world despite how the participant's body was oriented. This
encouraged the participants in the HMD conditions to face primarily in one direction and aided
in keeping the interactions similar between the HMD and screen experiments.

Figure 16: X-Box Controls

4.4 Factor Limitations
Certain factors could not be held constant across the studies. The most problematic factor in
this regard was habituation. Because cybersickness is highly individualized, different conditions
should be tested with the same individual. Unfortunately, due to habituation, a participant's
score will naturally decrease from session to session.

94

The same application was used for all the experiments, but the set of rooms changed from
session to session to minimize boring the participants. To decrease the variance across the sets,
each set was required to have one rectangular maze, one curved wall maze, one large room
with a generic texture, and one large room with a high orientation cue texture (e.g., bricks).
Fortunately, when the effect of the rooms was tested with a nonparametric version of ANOVA,
the set of rooms statistically displayed no effect.
A large screen display's glasses weigh less than an HMD, but as our HMD weight experiment
demonstrated, headset weight is not a source of cybersickness. In the screen size experiment, a
participant stood at different distances to achieve the same field of view. This changes the
angular momentum and the accommodation of the eye.
The HMD blocked out most of the outside world while the large screen still allowed 100° of the
real world to be seen despite a dimming of the room. This was in effect an independent visual
background, the use of which has been shown to consistently decrease cybersickness.
Lastly, although the navigation paradigm encouraged the same type of interaction, there were
differences between the HMD and large screen displays. When participants turned their head
with the HMD, the application changed its view accordingly. When participants turned their
head under the large screen condition, they were allowed to view more of the real world.

4.5 Procedures and Measures
All the experiments followed the same general procedure and employed the same virtual
reality "treasure hunt" application. The participants signed a consent form before the
experiments began, and all the participants were over the age of 18. The repeat sessions were

95

separated by a minimum of one week, and two weeks were encouraged to try to limit the
habituation effect. The general set up was explained to the participants and then they put on
the display, and a postural sensor was clipped to the back of their collars. They were then given
an X-box controller for navigation. In the first session, the participants were given a tutorial so
that they could learn how to work the controls and play the game. They were then placed into
the full environment and instructed to locate menu items as quickly as possible.
The participants were instructed that they could stop at any time for any reason, and were
monitored for symptoms every three minutes. They were asked, "On a scale of zero to ten,
where zero is how you felt coming in, and ten is that you want to stop, where you are now?"
The 0-10 scale is taken from Bos et al. who showed good correlation with the SSQ-T [94], but
the question was rephrased to specifically avoid any references to illness or nausea to avoid
demand characteristics [36]. This proved to be important as only a third of the participants that
withdrew early specified nausea as the reason for stopping. The values used in this question are
called "immersion ratings." The highest value of a participant’s immersion rating during a
session is called the "max immersion rating."
Immediately following the session, the participants were given the SSQ and one additional
survey. In the initial session, the additional survey was the background survey. In the second
session, the participants were given the MSSQ. The surveys were split to avoid survey fatigue. If
participants attended both the weight and screen size experiments, a smaller background
survey with just the display-specific question was administered after they used the large screen
display for the first time. Many of the participants also gave informal feedback throughout the

96

virtual reality sessions and many of the early participants were asked if there was anything that
could be done to improve the system. The SSQ was primarily used to compare the experiments.

97

5

Experiments

Three experimental sequences were performed as to further characterize the effects of several
factors on cybersickness that had not previously been addressed in sufficient detail or with
statistically significant results. These experiments addressed:
1. The effects of HMD weight.
2. Variations in screen size.
3. Impact of stereoscopic rendering.
The HMD weight experiment was performed first, followed by the screen size experiment. As
these experiments were later used to compare the display types, the participants were asked to
attend both the HMD and screen size experiment sets. However, due to incomplete
experimental sets where participants did not complete both conditions, not all subjects
participated in both the HMD weight and screen size experiment sets. The stereoscopic
rendering experiment was performed last, while data for the individual factor analyses was
collected during all experiments.
One issue with cybersickness data is the nature of the distribution. The SSQ is one of the
standard measurements used in cybersickness research, but it is not a linear scale. A typical
distribution is shown in Figure 17. Many standard statistical models assume the existence of a
Normal, or Gaussian, distribution. When data is not normally distributed, it is typically
transformed so that the normality condition can be met. Examples of these standard
transformations include

and,

Cybersickness data is often skewed to the extent

98

that even with these transformations, the normality condition cannot be met. Accentuating this
issue is the SSQ's oversensitivity, with SSQ-T values of 0 to 10 traditionally signifying no effect,
or non-susceptibility. Non-susceptible participants encompass over 50% of the general
population. Occasionally, existing studies have run analysis only using participant data with
SSQ-T values greater than 10, but the percentage of non-susceptible participants is as
important as the severity of symptoms in susceptible participants. As shown by Rebenitsch and
Owen [142], incorrect models can lead to unstable tests and faulty significant results.
Therefore, paired non-parametric tests were used.

Figure 17: Sample SSQ-T scores

5.1 HMD Weight Procedures
The participants were recruited to test two different weight conditions. We used an HMD with
a base weight of 340 grams. The weighted condition added 150 grams to the front of the
display as seen in Figure 17. To relieve the pressure from the original narrow strap, the strap

99

was replaced with a padded version and an over-the-head band was added to further reduce
the pressure on the participant’s head. These changes made the display more similar to modern
displays. We recruited 24 participants for both weight conditions. Their average age was 19.8
years with a standard deviation of 2.5 years. There were 5 females and 19 males. We had 4
early withdrawals, all males, out of 48 sessions.

Figure 18: The Head Mounted Display
The participants were requested to participate in the experiment twice, at least a week apart:
once for the base condition and once for the weighted condition, in random order. The
participants were also given one of two room sets in random order, although the Kruskal test
(the Kruskal test is a non-parametric variant of ANOVA) analysis displayed no effect based on
the choice of room set (p < 0.75).

5.1.1 Specifications
The application frame refresh rate was 60 Hz on average, but because it was rendered in
stereo, the effective frame rate was 30 Hz per eye. The HMD was a Sony Glasstron LDI-D100B
with an 800 X 600 resolution, a non-adjustable interpupillary distance, and a 35 degree diagonal

100

FOV. The application was rendered with a 33.3 degree diagonal FOV. The participants were
permitted to take one step in each direction.

5.1.2 Results
The SSQ-T scores were analyzed using a paired Wilcoxon test. This is a non-parametric test and
has shown to be robust with respect to outliers in cybersickness data [142]. Paired tests were
used since the same participant was in each condition, and paired tests are more sensitive to
changes within a subject. The SSQ-T scores are used since they are the standard measure of
total cybersickness, while the immersion scores were designed for monitoring participants.
There was no effect of weight on cybersickness, with p < 0.88. The mean for the non-weighted
condition was 23.12 with a standard deviation of 21.7, and the mean for the weighted
condition was 29.6 with a standard deviation of 31.3, among those that completed sessions
under both conditions. The mean and standard deviation are parameters for a normal
distribute so they lose some of their meaning with non-Gaussian distribution of SSQ data.
The high p value means that the range of modern HMD weights is unlikely to affect
cybersickness, and judging from the high p-value, even extremely lightweight HMDs are unlikely
to have an effect. Therefore, the difference between HMD and other displays must lie
elsewhere. This is of benefit to HMD developers because it means that additional hardware can
be added without much cause for concern, except for one caveat. Although the HMD weight
did not show an effect on cybersickness, the participants were vocal about the discomfort of
the display. The heavier display placed more pressure on the bridge of the nose, and this
additional pressure on the nose was nearly universally disliked. Therefore, while displays can be
made heavier, they also need to disperse the weight away from the bridge of the nose. Heavier

101

displays may be less comfortable and, potentially, less competitive as a product, but they do
not cause increased incidence of cybersickness.

5.2 Screen Size Procedures
Participants from the HMD weight experiment were asked to attend the screen size experiment
as well. A few participants attended only the screen size experiments. We recruited 22
participants for both screen experiments with an average age of 19.9 years and a standard
deviation of 2.6 years. There were 5 females and 17 males. We had 2 early withdrawals out of
44 sessions, consisting of one male and one female participant.
In the screen size experiment, the participants were presented with a 113 centimeter screen or
a 70 centimeter screen in random order. Two new room sets were created for the screen size
experiment, which were presented to the participants in random order. Under both conditions,
the participants were standing, had head tracking, experienced approximately the same
ambient brightness, and were positioned so that they would have the same starting field of
view. Since a temporary step in each direction was permitted, only the average field of view
was identical for all the participants. Hardware limitations required the smaller screen to have
80% of the resolution of the larger screen. Angular momentum changes were associated with
the displays, as the smaller displays had faster rotational velocities due to their larger relative
changes in position. Similar to the HMD weight experiment, a Kruskal analysis of the room set
order displayed no effect (p < 0.37).

5.2.1 Specifications
The application frame refresh rate was 100 Hz on average, but because it was rendered in
stereo, it was effectively 50 Hz per eye. The refresh rate was a limitation of the active shutter

102

glasses. The projector was a stereoscopic capable InFocus DepthQ projector connected with a
VGA cable. The maximum resolution was 1600 X 1200, and the image was displayed using a
rear-projection screen. The screen sizes were 113 centimeters or 70 centimeters on the
diagonal, and the participants were positioned so that they had a 35 degree average diagonal
FOV. The 210 centimeter screen was rendered at an 800 X 600 resolution due the masking half
of the display. Due to hardware limitations, the small screen was rendered at 80% resolution.
Participants were permitted to take one step in each direction, but were required to return to
their starting positions.

5.2.2 Results
The SSQ-T was analyzed using a paired Wilcoxon test. Paired tests with the SSQ-T were used for
the same reasons as the HMD experiment. There was no effect on cybersickness, with p < 0.66.
The mean for the smaller screen was 18.5 with a standard deviation of 19.3, and the larger
screen had a mean of 18.8 with a standard deviation of 15.7. As this is substantially different
from a statistical trend, this means that the differences reported earlier between using
monitors versus large screens are likely due to differences in the field of view, head tracking, or
the use of standing versus seated participants. This also means that the distance to the screen
likely has little effect. Given the low amount of physical movement in our study, the effect of
angular momentum is still uncertain. Because the resolutions we used were so similar,
resolution also remains an uncertain factor. This benefits cybersickness researchers because it
means that the results from monitor experiments can be compared directly with the results
from large screen experiments, assuming the remaining factors are held constant.

103

5.3 Display Type Procedures
This analysis used the results from the HMD weight and screen size experiments. Both
experiments used the same average field of view. The interaction between the two systems
was also very similar. By adding the head tracked viewpoint directly on top of the controller
position, the “forward” direction remained the same in both conditions. This encouraged the
HMD participants to always face in the same direction as they would in a large screen
environment. In addition, both the HMD and screen size conditions permitted only one step in
any direction.
Since the HMD and screen size sessions showed no effects, a participant's scores within each of
the two experiments were averaged. This has the added benefit of decreasing the effect of
habituation on the data because the screen size experiments were performed after the weight
experiments. Some habituation effect was expected, although there is no data showing that the
habituation effect extends to other display types. If a participant chose not to attend both HMD
sessions and both screen size experiments, their single score was left alone.
We had 24 participants with data for at least one HMD and one screen size experiment. These
participants had an average age of 19.7 years with a standard deviation of 2.5 years. There
were 5 females and 19 males. We had 7 early withdrawals out of 92 sessions, consisting of 6
males and 1 female. We had four participants miss either one HMD session or one screen
experiment, either due to the participant’s request or scheduling issues.

5.3.1 Results
The SSQ-T was analyzed using a paired Wilcoxon test as before. There was a significant effect
on cybersickness, with p < 0.02. This suggests one or more of the following. There was an IVB

104

effect, the slight change in interaction had an effect, there was sufficient transfer of
habituation, or a combination of these effects. The study by Kershavarz, Hecht, and Zschutschke
supports the IVB theory [82]. When they saw a difference between a large screen and HMD
while holding the application, interaction, and field of view constant, they performed another
experiment to clarify the effect. They masked the outside imagery in the screen condition so it
would resemble that of an HMD, therefore eliminating the IVB effect. This change in
configuration rendered the screen and HMD cybersickness results statistically equivalent. If the
model found for IVB is used to convert the HMD scores to the projection screen IVB setting,
there is no longer a statistical difference (p = 0.77).
Our participants’ HMD cybersickness scores were on average SSQ-T 9.7 higher than the large
screen experiments. In Kershavarz, Hecht, and Zschutschke’s study, the SSQ-T average
difference was 5.2 lower. This opposite effect may be due to different modes of interaction and
applications. We used head tracking and participant controlled motion, while Kershavarz,
Hecht, and Zschutschke used a chin rest with no participant control with the visual stimulus of a
moving car. There are two theories for these conflicting results. The effect of head tracking and
user-controlled application interaction overshadows the effect of the display. Alternatively, the
use of a chin rest made the large screen system seem more unnatural than the HMD system,
which is expected to increase cybersickness according to rest frame theory.
Habituation might still be transferred to other displays, as later analyses of video games have
shown. The overall effect of the minor difference in interaction is unknown and further studies
are needed to clarify this effect.

105

5.4 Stereoscopic Rendering Procedures
This experiment used the treasure hunt game application and used a large screen as before.
Therefore, the same statistical tests were used as before. However, the participants were
permitted to move freely, and the screen field of view was increased. A participant was
presented with the application using stereoscopic or monoscopic viewing in random order, at
least one week apart. Two new sets of rooms were created for the experiment, which were
presented to the participants in random order. New participants were recruited for this study,
although four participants from the earlier studies asked to attend, but only one participant
completed both sessions. Fortunately, there was an average gap in time of one month between
the screen and stereo experiments.
We had 28 participants, but 6 could not return due to time constraints, as this experiment was
run near final exam time. The remaining participants were on average 21.4 years old with a
standard deviation of 3 years. There were 8 females and 14 males. We had 8 early withdrawals
out of 50 sessions, consisting of 5 males and 3 females. Two male participants withdrew twice.

5.4.1 Specifications
The application frame refresh rate was 100 Hz on average, but because it was rendered in
stereo it was effectively 50 Hz per eye. The refresh rate was a limitation of the hardware. The
screen size was 225 centimeters on the diagonal, and participants stood wherever they pleased.
The field of view was typically set between a 60-90 degree diagonal FOV. The screen was
rendered at a resolution of 1600 X 1200. The participants were permitted to freely move in a 2
by 1.5 meter area.

106

5.4.2 Results
The SSQ-T was analyzed using a paired Wilcoxon test, and showed no cybersickness effect, with
p = 0.22. There was still no effect after normalizing for the effect of habituation using the
developed model (p = 0.19). Several participants only showed a temporary effect and returned
to their baselines within 10 minutes. The mean for the monoscopic condition was 28 with a
standard deviation of 27.3, while the stereo mean was 33.3 with a standard deviation of 24.1
among those participants that completed experiments under both conditions. The results are
tentative. Stereoscopic rendering may interact with the application. The human visual system
only relies on stereo fusion out to several feet, and primarily within arm's reach. The visual
system uses other visual cues to estimate the distance of farther objects. Mon-Williams and
Wann also mentioned an effect on cybersickness if the focal distance changes frequently [98].
There also may be an underlying trait that makes someone susceptible or not to stereoscopic
rendering. While more study is needed, one can conclude that the effect is likely to be small.

5.5 Individual Susceptibility Procedures
Immediately following a session, each participant was asked to fill out a usability survey after
their first session and the MSSQ survey after their second session. These surveys are available
in Appendix A. The survey questions included both usability and a broad range of background
demographics. The reasons for the inclusion of the factors are available in Table XVII. An early
analysis revealed that several of the factors under consideration were not viable or required
further clarification. The usability survey was revised midway through the experiments. The
original usability survey was administered to 20 participants from the HMD weight experiment
set. There were 3 females and 17 males recruited for the study. The participants' ages ranged
from 18 to 31 years, with all except one under the age of 22. Unfortunately, 4 participants
107

withdrew due to scheduling difficulties, and therefore these participants lack MSSQ data. The
updated survey was given to 33 participants. One participant was removed from the analysis
with the updated survey. This participant had very inconsistent results. On the MSSQ, he circled
moderate susceptibility, but reported a MSSQ score of 0.
The MSSQ asks for the frequency of illness during several modes of travel in the last 10 years,
and the history of motion sickness as a child under the age of 12. The scale used is 0, 1, 2, 3, 4,
or “never”, ”rarely”, “sometimes”, “frequently”, and “always”, respectively. Past research has
suggested that specific modes of travel may be better correlated with cybersickness, specifically
carnival ride sickness [23]. To ascertain a participant’s susceptibility to specific modes of travel
values based on the MSSQ, the following function was used:

6

In this equation,

is the MSSQ score of the subcategory,

frequency of the subcategory,

is the motion sickness

is the vomiting frequency of the subcategory, and

is (1) if

the participant has used that mode of travel, or (0) if the participant has never used that mode
of travel.
The Kendall non-parametric test was used to test for correlations with the non-normalized
data, and the Spearman test was used with ordered data (e.g., the frequency of 3D exposure).

5.5.1 Initial Results
Initially, we intended to include age and gender in the analysis. However, due to the
unbalanced nature of the data, this analysis could not be performed. Age, in particular, is

108

difficult to model with an appropriate distribution, and studies typically require a primary focus
on age or a publicly available system to gain an appropriate age distribution [55,56]. Only one
participant had used an HMD before, so past display experience was eliminated from the
analysis. Several categories naturally fell into bins. Game Play was binned into “rarely” (0-2
hours), “sometimes” (2.5-6 hours), and “regularly” (6.5 and over hours) per week. 3D Display
Exposure was binned into “never”, “monthly”, and “yearly.” Hat usage was binned into “rarely”
(less than 6 hours), “sometimes” (6-15 hours), and “frequently” (over 15 hours) per week.
We found significant correlations between SSQ and MSSQ (p < .03), SSQ and MSSQ carnival (p <
.01), SSQ and max immersion rating (p < 0.001), and SSQ and Vision correction (p < .01). We
found trends between SSQ and Game Play (p < 0.08) and SSQ and MSSQ train (p < 0.055). The
adjusted coefficient of determination, or the adjusted variance, is provided for each individual
factor in Table XIX. These values mean that Hats, past 3D Discomfort, VR Experience, and 3D
Exposure can be eliminated from future consideration.
Table XIX: Coefficient of Determination due to Characteristics
Adjusted Coefficient of Determination
10.8%
9.2%
-3.8%
-5.7%
-5.4%
17.9%
0.3%

MSSQ
Video Games
Hats
Past 3D Discomfort
VR Experience
Vision
3D Exposure

The adjusted variance, or the adjusted coefficient of determination, is better suited to small
data sets. It considers the effect of each new item and only improves the variance if the effect is

109

better than random. This helps prevent over fitting, but the resulting values are always lower
than normal variance and can become negative. If the adjusted variance becomes negative, the
mean of the values is better correlated than the actual values.
While there was a correlation with vision correction and the SSQ, we had few participants that
wore glasses exclusively, which made the results inconclusive. In addition, a subsequent
analysis using multiple factors revealed a possible interaction between the MSSQ and Game
Play. Greater detail on models using multiple factors is available in Individual Variation in
Susceptibility to Cybersickness (2014) [54]. The following two linear models explained 43% and
58% of the adjusted variance, respectively. However, the second model’s estimated terms
failed to meet significance. Therefore, the first model has a better fit to the data.
7
8

Suscept1 and Suscept2 are estimates of susceptibility given a set of individual background
demographics. P is the vector of background demographics for the participant, pCarnival is the
participant’s carnival motion sickness score,
0 otherwise,

is 1 if the participant wears contacts and a

is 1 if the participant wears glasses and 0 otherwise, and

is 1 if the

participant wears both contacts and glasses and a 0 otherwise.

5.5.2 Additional Results
We examined the data for additional indications of correlation with the factors of video game
genre, vision correction, and headaches. The MSSQ correlation followed the same trends as
before. Correlations with the MSSQ modes of travel were also reanalyzed, and the additional
modes of travel found to correlate to the SSQ-T are shown in Table XX.
110

The revised usability survey asked subjects to indicate not only their degree of game play, but
also their game play in specific gaming genres such as first person shooters, platformers, etc. A
wide variety of games was reported by the participants. Several other games were placed into
different genres. These included moving "Grand Theft Auto" to first person shooter, MMOs to
RPGs, and "Minecraft" to platformer. Sports, rhythm, and strategy games remained in the other
genre. The Spearman correlation was selected for video game play over the Kendal correlation
because the Spearman reacts stronger to the degree of distribution. Therefore, the Spearman
correlation is more appropriate when the degree of difference is important, such as with game
play. This is less true with the MSSQ and SSQ, which are non-linear.
Separating video game play into separate genres helped clarify the effects, as seen in Table XXI.
Overall game play no longer had a correlation to cybersickness, but specific genres did. First
person shooters and platformers have the strongest correlations with cybersickness. Both of
these genres have large amounts of forward motion and are the most similar to many virtual
environments. Game play does not correlate with the MSSQ, as seen in Table XXII.
The effect of vision was no longer statistically significant after including participants from the
screen size and rendering experiments. The average SSQ-T was found for each category per
experiment set, as seen in Table XXIII. The numbers of participants in each category were very
unevenly distributed (e.g., there were only two participants that wore contacts for the
rendering experiment). The standard deviation for all the values was quite high, with most
conditions having a standard deviation greater than 20. No grouping was found to reach

111

statistical significance, with the lowest p values equal to 0.19 which occurred when people who
wore both glasses and contacts were compared to all the other participants.
Therefore, we conclude that vision correction is either unlikely to have more than a small
effect, or only has an effect with HMDs. The uneven distribution may mask the effect, but the
high standard deviation signifies that any effect would be minor. The revised survey was given
primarily to the projection screen participants and glasses were more likely to be an issue with
the HMDs. There were not a sufficient number of participants in each condition to test for any
differences between them.
Headaches correlated to the SSQ but did not correlate to the MSSQ or the primary genres of
game play as shown in Table XXIV.
Table XX: MSSQ Correlations
Self-Limiting
Immersion Rating and
MSSQ
Correlation
-.07
P -value
0.51
Test
Pearson

MSSQ and
SSQ
0.25
0**
Kendall

MSSQ
MSSQ
MSSQ Train
Airplane and
Playground
and SSQ
SSQ
and SSQ
0.16
0.2
0.16
0.02*
0.003*
0.016*
Kendall
Kendall
Kendall

MSSQ
MSSQ Playground and MSSQ Small
Carnival and
SSQ
Boat and SSQ
SSQ
Correlation
0.16
-0.02
0.17
P -value
0.016*
0.82
0.008**
Test
Kendall
Kendall
Kendall

112

MSSQ Big
Boat and
SSQ
0.07
0.27
Kendall

MSSQ Cars
and SSQ
0.17
0.009**
Kendall

Table XXI: Binned Game Play and SSQ Correlations
Game Play Game Play
Game Play
shooter and platformer
and SSQ
SSQ
and SSQ
Correlation
-0.11
-0.34
-0.29
P -value
0.19
0.002
0.011
Test
Spearman
Spearman
Spearman
Binned

Game Play
Game Play
Game Play
driving and
fighting and
RPG and SSQ
SSQ
SSQ
0.09
0.28
0.23
0.46
0.02
0.05
Spearman
Spearman
Spearman

Table XXII: Binned Game Play and MSSQ Correlations
Game Play Game Play
Binned
forward
shooter
and MSSQ and MSSQ
Correlation -0.08
-0.09
-0.19
P -value
0.55
0.62
0.31
Test
Spearman Spearman Spearman
Game Play
and MSSQ

Game Play Game Play Game Play Game Play
platformer driving and RPG and fighting
and MSSQ
MSSQ
MSSQ
and MSSQ
0.09
0.29
-0.001
0.13
0.64
0.12
0.99
0.5
Spearman Spearman Spearman Spearman

Table XXIII: SSQ-T Values for Vision

HMD
Screen
Render
Total Average

None
31.56
24.5
30.3
28.6

Contact
27.42
16.8
62.6
28.6

Glasses
24.93
19.4
30.4
27.6

Both
12.45
16.2
29.9
19.5

Table XXIV: Headache Correlations
Max
Headaches Headaches
Immersion Headaches Migraines Headaches Migraines
and Binned And Binned
Rating and And SSQ And SSQ And MSSQ And MSSQ
platformer Shooters
SSQ
Correlation
0.5
0.266
0.021
0.244
-0.20
0.05
0.2
P -value
0**
0.02*
0.86
0.21
0.31
0.78
0.255
Test
Kendall
Spearman Spearman Spearman Spearman Spearman Spearman

113

We can conclude that there are three individual susceptibility factors correlated with
cybersickness, which are independent of each other. For an individual susceptibility model, the
effects of the experiments would need to be normalized. To convert from the screen size
experiment to the HMD experiment, the values are multiplied by 1.431 to remove the IVB
effect. To convert from the rendering experiment, the values are multiplied by 1.431 to remove
the IVB effect and by 0.53 to counteract the increase in screen size. The best linear model of
the MSSQ to the SSQ, which explains 30.5% of the adjusted variance, is where

represents

the participant's MSSQ score in the following equation:
9

The best model for game play used binned shooters and platformers. Fighting games were
dropped due to a poorer model. In the P demographic vector, a participant had a 1 in their
assigned video game bins, and a 0 in all the other bins. For example, if a participant played
shooters rarely and platformers regularly, their video game subset would be <1, 0, 0, 0, 1, 0>,
where the values represent rare shooter (shootRare), regular shooter (shootReg), frequent
shooter (shootFreq), rare platformer (platRare), regular platformer (platReg), and frequent
platformer (platFreq), respectively. The following equation explains 6% of the adjusted variance
in the data:

10

114

The best model found for headaches explained 15.8% of the adjusted variance where pHeadache
represents the number of headaches per month:
11

Simply using the MSSQ, headache, and video game play functions, 37.2% of the adjusted
variance in the data was explained. This function can be written as:
12

The low variance is not surprising as it only considers the individual portion of the results and
not the variation due to different configurations and other external factors. There are two
further limitations to this model. As expected, the values within the model do not follow a
Gaussian distribution, which decreases the accuracy of linear models. The video game term
only had a trend of p = 0.8 and the headache term did not reach significance (p = 0.11) in the
model, which indicates a weaker fit to the data. This is not surprising, given the relatively lower
variance that can be explained by their factors.

5.6 Participant Preferences
The participants gave both informal and formal feedback on their preferences. This feedback
can be used to identify potential factors, to build better systems, and to predict how
participants will use virtual reality.
The self-limiting immersion rating originally showed a significant correlation with the MSSQ, but
increasing the number of participants in the analysis made this no longer true, with p=0.6. The
last question in the survey was open-ended, and asked about other potential uses of virtual

115

systems. The uses suggested by the participants varied according to the display used. The
participants’ write-ins for other included:
1. Would be cool to use virtual reality for school
2. Google street view
3. Training and target practice
4. Virtual tour of Campus. Cities. Possibly concerts.
5. Campus tours or building tours
6. For some sort of interactive data visualization type thing it could be really useful
7. Virtual reality travel (vacation)
The participants also stated that they would like to see virtual reality in wider use, but only in
certain contexts. The percentage of participants that would use each device is shown in Table
XXV to Table XXVII.
Table XXV: HMD Preferences

TV
VR
Write
In

Watching TV

Movies

Video games

90%
10%

90%
10%

0%

0%

25%
70%
1) Depends on the
game

116

Landmark
exploration
20%
70%
1) Real life
2) Depends

Table XXVI: Large Preferences

TV
VR
Write
In

Watching TV

Movies

Video games

44
2
1) Depends on the
show

40
7

13
33
1) Depends on the
game

Landmark
exploration
5
41
1) Both

Table XXVII: Cumulative Preferences

TV
VR
Write
In

Watching TV

Movies

Video games

73
4

67
11

22
54

Landmark
exploration
9
65

1

0

2

3

5.6.1 Informal Feedback
The participants made comments during their use of the virtual reality system, and we
observed certain commonalities. A discussion of information feedback and world design
recommendations are provided in much more detail in World and Object Designs for Virtual
Environments (2014)[143]. Minor difficulties relating to game play were fixed during the
experiments. These difficulties included minor texturing problems, object orientation, and
menu readability. Early participants commented that the menu text was difficult to read at
times. Therefore, the text size was slowly increased. Correcting for object orientation was more
difficult. Flat objects (e.g., plates), small objects (e.g., pencils), and low objects (below 80 cm)
were difficult for participants to identify. Flat and low objects were easily fixed by rotating an
object upwards and raising the shelves several centimeters, respectively. The small object issue
could not be resolved because the objects were kept to-scale if possible. The smallest readily
identifiable size appeared to be an 8 centimeter cube, and it was only identifiable if there was

117

good color variation. Participants made several “feature requests”. Though not added, these
included sounds, true jumping, blowing up walls, animation, adjustable controls, monsters, and
more action in general.
Navigation through the virtual environment proved to be difficult for many of the participants.
Restricting the field of view made participants forget where they had been and made them
more likely to miss nearby objects. Direction markers were added to the maze room to help
with navigation and were specifically referenced the first time the participant entered a maze.
These markers only slightly improved navigation. In the stereoscopic viewing experiment, which
effectively doubled the field of view, navigation was much less of an issue. Many of the
participants frequently ran into walls. In the earliest rooms, the hallways were rendered toscale and were 1.85 meters wide, but these hallways proved to be much too narrow. The
hallways sizes were slowly increased in each new room model. The smallest acceptable hallway
size was 3 meters for a small field of view and 2.75 meters for a large field of view. Another
unexpected participant behavior was that certain corridors were ignored during navigation. In
two rooms, there was a hallway with four corridors branching off one side. For unknown
reasons, many of the participants frequently skipped the second corridor.
Although we were not testing the virtual environment visuals directly, there were some
consistent comments about problematic areas. Generic, low orientation textures, as shown in
Figure 19 f-g were problematic, with g being nearly universally disliked. The high orientation
textures in a-e, particularly those with non-repeating patterns such as “a” and “e,” were less
likely to cause issues. As mentioned earlier in the literature review, faster navigation is

118

associated with increased cybersickness. Judging from their comments, the participants seemed
to be in agreement that the run button should be used sparingly, with one participant
commenting "the run button is the sickness button."

Figure 19: Sample Textures

5.7 Cumulative Analysis and Results
A set of models was derived from the literature review and our experiments. Because
cybersickness values are relative, direct modeling is not viable. Thus, these values must be
estimated from the percent change of SSQ-T value based on prior studies. The results in the
following models estimate the percentage that B's SSQ-T changes from A, where A is a prior

119

study's configuration factor vector and B is a new configuration with a change in only one
factor:

13

14

15
16
17
18
19
20

There are diminishing returns on habituation over time and the model would be more accurate
if it included the smaller decreases associated with longer time periods, but there have been
few studies that have spaced their sessions over one week apart.
Using these results, two models were tested for their predictive capability. Direct linear
modeling was used for its simplicity and ease of use. Linear modeling permits direct use of the
data from the literature, but potentially suffers from the highly skewed and individualized
nature of cybersickness. The second model tested was a zero-inflated negative binomial
distribution model (ZINB). This model readily permits the modeling of individual effects and
takes into account the large percentage of people who do not become ill. The ZINB model is not
120

widely supported, and when it is supported, it has restrictions on the format of the data, which
will be discussed in greater detail in Section 5.9.

5.8 Linear Model using Reported Configurations
Linear models using individual data have been attempted before. Kolasinski also had a linear
model which explained 34% of the variance of the data [1]. As the author mentioned, the highly
skewed data of cybersickness makes a linear model unreliable. In our model, the results of
previous studies are used rather than the individual results, which should slightly lessen the
detriment on its reliability. This approach does, however, have the unwanted effect of making
any predications general to the population and unusable for dynamic adjustments to
individuals. However, it does provide a method to estimate the effect of a new virtual reality
system before time and capital are invested in creating the system.
The experimental configurations and their resulting SSQ-T values were recorded from the
literature and are available in Appendix B. Only those studies that reported an SSQ-T value or a
means to convert to an SSQ-T value, as well as a sufficient number of hardware and software
factors, were included. Because study conditions are normally random or use different
individuals, habituation was dropped from the analysis. For the analysis, if stereoscopic
rendering was not specified, monoscopic rendering was assumed. While some of the factors did
not have enough support to create a model, they were reported frequently enough to be
included in the analysis. These factors include realism, seated, head tracking, and the controller.
The levels of the controller were limited to joystick (gamepad), simulator, mouse, head, and
none. Although we found no cybersickness effect with stereoscopic rendering, others have
found such an effect, so it was kept for analysis. Factors such as head tracking are likely

121

affected by the number of orientations that are supported and interaction with the controller,
but these effects are currently unknown. Therefore, head tracking was represented as “yes or
no” if head tracking was provided and head movement was permitted.
Many variables were undefined in their respective publications. If realism was not given, age
and implied content were used to estimate its value. Displays were converted into those that
were likely to have an IVB effect or not. However, there are many ways to achieve an IVB effect.
Simply increasing tactile feedback, as was done in Hakkinen, Vuori, Puhakka [61], can achieve
an IVB effect. The chair used in their movie condition had more postural support, yielding an
IVB like effect. Jaeger and Mourant’s treadmill was considered to be a simulator for the
purpose of the analysis. The missing head tracking and sitting values were replaced by the
implied method in their publications. Sharples et al. did not report the field of view in several of
their conditions [108]. The projection screen was assumed to have a smaller field of view than
the reality theater system and the monitor was assumed to be viewed at a comfortable
distance. The converted values are available in Appendix B. To test for direction, the y/n factors
were converted to 1/-1, respectively. The realism factor range was converted to 0, 1, 2, and 3
(low, medium, high, and realistic, respectively). The percent changes were used directly in the
model, rather than having the original SSQ-T values first adjusted by the earlier created
equations.
A linear model using all the factors with a 2 way interaction explained 70.2% of the adjusted
variance. While this is a great improvement over the 34% reported earlier, our model is too
complicated to be of practical use. It includes 396 terms, and the vast majority of these terms

122

did not reach significance. This means that the model was over fitted, and the majority of the
coefficients did not reach statistical significance. The best model with exclusively significant
terms is Equation 23, below. All terms except the initial SSQ-T (aSSQ), which had a p = 0.03, had a
statistical value of p < 0.0001, which represents a good fit. The Predict(A,B) model explained
54.5% of the adjusted variance where xtracking is 1 if the participant was head-tracked and 0
otherwise, xrealism is the 0-4 value for realism, xcontrol is the controller used, xsitting is 1 if the
participant was seated and 0 otherwise for configuration X.

21

22

23

Improved models can be expected if multiple nearby points are used to estimate the effect.
Some of the interactions are unsurprising. An IVB deals with the enforcement of participants’
expectations, so the realism of the scene could affect the amount of presence, or the
participant’s belief of “being there.” Head tracking and the controller are directly related, and
the presence or absence of head tracking will greatly affect how the controller works. What is
surprising is the small effect of

, or the SSQ-T value from configuration A that is used as a

123

starting point. Stereoscopic rendering by the FOV could be added to the model to explain
another 0.4% of the adjusted variance, but such a model has a slightly poorer fit to the data.
Some of the issues with the model are that it does not account for the individual, and the
individual alone can explain 30-40% of the variance. Because the results that were used for the
analysis are summary statistics, this effect is expected to be reduced, but another 10-20% of
the variance is still likely due to individual differences in the study populations.

5.9 Zero Inflated Model using Experimental Results
Rebenitsch and Owen found that zero-inflated negative binomial models showed promise as
future predictive models [142]. The zero-inflated negative binomial (ZINB) model includes a
probability parameter for a zero-class (well group). If a data point falls into the non-zero class
(ill group), the distribution follows a negative binomial distribution. A negative binomial
distribution is essentially an over-dispersed Poisson distribution. However, it is not well
supported in statistical packages and requires an adjustment of the data if it is to be used. The
negative binomial portion of the model requires integer count values. Therefore, all the SQQ-T
data must be rounded to the nearest integer, making the results more similar to a count of
symptoms multiplied by their severity. Since our reported SSQ ranged from 0-150, this minor
decrease in accuracy is acceptable. The original ZINB model also requires that the well-group
have the same zero value. Therefore, all the values below a certain threshold must be set to 0,
and the remaining values must be decreased by that threshold to maintain continuity. From the
literature review, the standard SSQ-T interpretation of a rating that is less than 10 as "well"
does not seem to be true. Many of the well participants reported values of up to 15. Therefore
the cut off for the well group was changed to 15.

124

We dropped two participants from the analysis. One participant "could not remember" if he
became sick and the other participant had the extremely inconstant MSSQ as mentioned
earlier. The original usability survey did not include the game genre or headache factors, but
removing the original study values would eliminate most of the HMD participants from
consideration. Setting a value of zero for all the participants would likely mask the results, so
their answers were estimated. For game play, the average number of hours was known, and
most participants tended to favor one genre. For example, someone who played 10 hours of
video games would have 7 hours of playing “shooters” and 3 hours of playing another category.
Therefore, we could estimate their genre game play by the frequency of the genre and their
hours of game play.
Estimations of the headache and MSSQ responses were less accurate. The headache response
had to be estimated using random sampling, and weighted directly from the correlation
estimate from their SSQ-T scores and the frequency of each answer. This has the unfortunate
potential effect of forcing the data to fit the model for these participants. The MSSQ had to be
estimated directly with random distribution from its standard deviation. Fortunately, there
were only a few participants that did not fill out an MSSQ.
The total number of sessions was 147. Of these, 140 sessions had an associated MSSQ, 75 had
completed revised usability surveys, and 71 that generated complete data sets.
Because the ZINB model does not use standard linear models, calculating its variance is highly
questionable. Therefore, the goodness of fit models is estimated from the residuals from a
fitted model, with a lower average absolute residual representing a better model. To reflect the

125

fact that stereoscopic viewing and IVB can affect the direction of the results in both directions,
these values were transformed to “-1” for “no” and “1” for “yes.”
When creating the ZINB model, both the raw data and the data that were converted using
single factor models were tested. Using all the factors and the raw data, the ZINB regression
created a model with an average residual of 0.69 and a standard deviation of 0.88. The best
model found had statistical significance p<0.05 for all its terms, an average absolute residual
mean of 0.66 and a standard deviation of 0.85. Two terms were found for the negative binomial
portion of the model:

by field of view and IVB by Headache by Video(A). The

zero/well class prediction terms were: the MSSQ, binned first person shooters, headaches,
and Stereo. The Stereo term in the negative binomial portion could be dropped with
little effect on the residuals, but this caused the term to lose its statistical significance. The
coefficient values are shown in Table XXVIII.
Table XXVIII: ZINB Model Terms

Intercept**
2.69

Intercept**
7.95

MSSQ**
-0.03

Negative Binomial
by
IVB by headache by Video
FOV**
(A)*
0.001
-0.002

Log(Theta)*

Zero-Inflation Model Coefficients
Video(A)*
Headache(A)*
Habit(A)*
-0.06
-0.04
-4.74

1.59

24

Stereo*
-0.75

This model implies a few traits about the nature of the data. Stereo is still a small term, but it
greatly affects other factors, which explains the conflicting results on this factor in the
literature. Most of the terms are part of the “well” class predictor or concern items that are
126

individualistic. This means if the individual is factored out, there is a moderate, rather than
large, effect imparted by the equipment. If the “well” grouping terms are placed in a linear
model, they explain 23% of the adjusted variance of the system.
This model provides a better means to determine what will happen to a particular person in a
particular virtual environment. The prior models either predict how sensitive a participant is to
cybersickness, or predict what would occur to a general population with a particular
configuration. Models using hardware, software, and individual terms could be utilized to give
better guidance to a person. The ZINB model is also more sensitive to susceptibility effects on
statistical results. If a resistant individual becomes ill, it has a stronger meaning than if a highly
susceptible person will be ill.

127

6

Conclusion and Future Directions

Cybersickness remains a complex issue to describe and study. Individual models yield an
explanation of adjusted variance of 37%, while cross configuration models developed from the
literature explained an adjusted variance of 55%. If these results were purely additive, 92%
percent of the variance in a system could be explained by these variables. It is more likely that
the explained variance is approximately 75-80%. Determining the actual variance is not possible
without including individual participant data from other experiments. This is a large
improvement over Kolasinski's model that explained 34% of the variance [9]. The models
suggested by So [10], Simon [13], Jones, Kennedy, and Stanney [14] do not account for the
individual. Simon and Jones, Kennedy, and Stanney both require a large number of
configurations to be tested in a uniform fashion using the same measures. The cybersickness
literature does not meet these requirements. The zero-inflated negative binomial (ZINB) model,
as developed, mixes raw data and modeled results. The ZINB model also follows the
distribution of the data more closely. Using both individual and configuration data results in
higher accuracy, with residuals that are less than one on average.
A literature review of the factors proposed by Kolasinski [9] and Renkewitz and Alexander [48]
revealed only a few factors with sufficient evidence that would be suitable for use in potential
virtual system guidelines. These include the real field of view, independent visual background,
habituation, and duration. Several factors have shown consistent and strong results, but others
would greatly benefit from additional data points, including the navigation speed and sitting
versus standing factors.

128

Several other factors displayed a potential to greatly affect cybersickness, but these factors
have limited data. These factors include scene realism, initial postural stability, change in
postural stability, the ratio between the virtual and real field of view, method of movement,
and head tracking. Originally, the type of display showed promise as a cybersickness factor, but
the recent results of IVB studies and our own results suggest that the differences associated
with them are due to other factors.
One major difficulty in studying the effects of cybersickness is the variability in individual
susceptibility. Within subject experiments are more likely to detect effects, but using the same
participants over multiple sessions suffers from the effects of habituation. There are also
questions of how the results will generalize to a normal population as participants tend to be
self-selecting. Ideally, the individual aspect should be normalized before any comparisons are
made, but studies on individual susceptibility to cybersickness are limited. Moreover, results
from motion sickness studies may not necessarily apply to cybersickness studies [55].
Therefore, far more research on individual susceptibility to cybersickness is needed. However,
several of the individual factors that have been suggested are time consuming or difficult to
assess. These include interpupillary distance (IPD), mental rotation, perceptual style, and body
mass index (BMI). One individual characteristic that is in critical need of study is whether
habituation with different configurations transfers to new virtual reality systems.
Several models were developed during our experimentation for different purposes. These are
single factor models that can be used for virtual system guidelines and normalizing results.
Individual susceptibility prediction models can be used to dynamically adjust a virtual reality

129

application to a participant. The Individual(A, B) from Equation 17 may be used to normalize a
population to allow different groups of individuals to be tested. To normalize, a standard target
value is necessary. The motion sickness susceptibility questionnaire (MSSQ) average is
approximately 40, and may be the best choice for generalization. Cross configuration prediction
models are useful in that they can estimate the effects of a new virtual reality application
before any time or capital is spent developing it. The cross configuration models also permit the
comparison of results from different publications.
Based the single factor models, the following guidelines are proposed:
1. If a participant is highly susceptible, shrink the field of view with its effect modeled by
Equation 3.
2. Use an independent visual background (IVB) whenever possible. The expected change in
effect is modeled by Equation 5
3. Keep forward movement slow. If there is some user interface akin to a run button that
allows a participant to move more quickly than would be natural, limit the amount of
time the run button can be used. The expected change in effect is modeled by Equation
4.
4. Participants generally develop tolerance to cybersickness-inducing stimuli if they are
subjected to repeat sessions. Habituation also can develop through playing video games
with a large amount of forward movement such as that found in first person shooters
and 3D platformer games. If a system needs to be used over a longer time period,

130

shorter repeat sessions can build tolerance. The expected change in effect is modeled
by Equation 1.
5. Longer sessions are more likely to cause illness. The expected change in effect is
modeled by Equation 2.
6. Have different configuration schemes for low, moderate, and highly susceptible
participants. Administer a short susceptibility test prior to the participant’s first use of a
system to incorporate suggestions. The model provides 12 questions as an estimate, but
Golding has proposed a much shorter susceptibility test that is likely to be better suited
to cybersickness studies [144]. Since the tolerated symptoms differ between individuals,
permit the scheme to be set by the user.
Our statistical results have implied the following suggestions for researchers and
developers:
1. Estimate a new configuration and application effect. Use the model Equation 23 before
building a system.
2. If the population and configuration are known, Equation 24 can provide an estimate of
the effect.
3. The effects of habituation can be normalized, and within subject design is better at
detecting effects.
4. Cybersickness data is not normally distributed. Nonparametric statistical tests are
required.

131

5. For ease of comparison with other publications, the field of view, the display/IVB, the
controller, duration, sitting or standing conditions, history of motion sickness, amount of
motion, and scene complexity should be reported.
Design suggestions are available in Rebenitsch, Owen, and Coburn [143]. The individual
susceptibility model in Equation 12 could be used to allow commercial virtual reality
applications to make suggestions on the rendering appearance for an individual. For example, if
this formula yields a low 5, the participant is unlikely to become ill, so high speeds, wide fields
of view, and realistic visuals are permitted. If a participant scores a high 40, the system can
narrow the field of view, limit the periods of high speed, and simplify the visuals.
The cross configuration results were tested using only paired data. Improved estimates can be
expected if multiple similar configurations are used to estimate the results. The models are
missing a few important factors, such speed and realism. So's cybersickness values are likely a
good measurement of these two factors, but are time consuming to produce and require a prebuilt application. Using average speed, rotation, and scene complexity directly are likely to be
good alternatives to using So’s cybersickness values, but these models will need to be
developed. The data used to create the ZINB model has fewer configurations to test: just 6
configurations in total with 147 sessions. However, for this type of model to be reliable, a joint
pool of data from many different researchers and studies may be required.

6.1 Future Directions
Our cybersickness experimentation and models have clarified the effects of several factors. An
HMD’s weight has little effect, assuming that there is a proper fitting display. This is of benefit

132

to developers as it permits them to introduce heavier hardware. Our results also support the
fact that the display itself has little effect on cybersickness, and that any changes associated
with the display are likely due to different factors and/or an IVB effect. This is of benefit to
researchers and developers as it simplifies their experimentation. Unexpectedly, stereoscopic
rendering did not show any effect on cybersickness, even when accounting for habituation. The
results from the literature also present conflicting results with respect to stereoscopic
rendering. The ZINB model implies that stereoscopic viewing may affect the likelihood of a
participant becoming ill, but it does not predict how severe the symptoms might be if they did
occur. Most likely, stereoscopic viewing has a very minor effect on cybersickness, or possibly
interacts with the distance of the virtual object in the application.
While our experiments and models explain several factors, many factors remain that require
further study. A list of some possibilities in this regard is available in Appendix B. Another
necessary component of future cybersickness research is to find quick and easy ways to
determine individual susceptibility. At two pages, the MSSQ survey used for our results is too
lengthy for non-academic use. Fortunately, Golding has proposed a much shorter susceptibility
test with two question grids (see Appendix A), which is associated with only a minor decrease in
accuracy [144]. Surveys remain the standard measurement of cybersickness, but they are not
the most accurate. We had two participants that reported clearly unusable survey results.
Surveys are also likely to be skipped by participants if they delay game play. Postural stability is
a viable objective alternative, but a predictive model is needed.

133

While the simulator sickness questionnaire (SSQ) and background demographics were the
primary areas of our analysis, participants’ symptoms over time, game play movements, and
postural stabilities were also recorded. These data sets can be used for subsequent analysis. A
participant's game play movement can be analyzed to determine how different people
approach and navigate 3D environments. For example, which areas are traveled frequently
versus which areas are skipped? While no formal analysis was conducted, participant
immersion scores over time show one of three behaviors: no symptoms, a linear increase, or an
exponential increase, with an immersion rating of 4 seeming to be the critical point. If these
values could be satisfactorily correlated to a predictive postural stability model using recorded
data, dynamic warning systems could be developed to warn people to exit a virtual
environment several minutes before the predicted onset of cybersickness symptoms.
Cybersickness is only one aspect of a virtual environment. There is still much to be learned
about navigation and interaction. Participants regularly ran into walls and became lost in our
system. We had several suggestions from participants during the course of our experimentation
on how to improve the system. One item of primary importance is how to include head
movements into navigation. Using head movements alone is unnatural, but the participants in
our study forgot which way was forward when their head movements were completely
separated from the joystick’s movements. The best solution to this problem is likely to include
using a mix of these two types of movement.

134

APPENDICES

135

APPENDIX A

SURVEY INSTRUMENTS
Table XXIX: Simulator Sickness Questionnaire Symptom Categories
Weight
SSQ Symptom

Nausea

Oculomotor

Disorientation

General discomfort

1

1

0

Fatigue

0

1

0

Headache

0

1

0

Eyestrain

0

1

0

Difficulty focusing

0

1

1

Increased salivation

1

0

0

Sweating

1

0

0

Nausea

1

0

1

Difficulty concentrating

1

1

0

Fullness of head

0

0

1

Blurred vision

0

1

1

Dizzy (eyes open)

0

0

1

Dizzy (eyes closed)

0

0

1

Vertigo

0

0

1

Stomach awareness

1

0

0

Burping

1

0

0

Column Weighting for Category Scores

9.54

7.58

13.92

Total Score = 3.74 * (Sum(Nausea) + Sum(Oculomotor) + Sum(Disorientation))

136

How many hours per week on average do you play console video games?
Have you ever used virtual reality before (e.g. head-mounted displays, large screen where
your viewpoint is tracked, or moving platform rides, not games such as “Second Life” where
the virtual reality is only experienced on a conventional computer screen)? Y / N
Have you ever used virtual reality with the type of device you just used? Y / N
How frequently do you view 3D movies or 3D TV?
Have you had any discomfort with viewing 3D movies, if so what?
How many hours per week on average do you wear hats or headbands?
Do you use glasses? Contacts? Neither? (circle)
Was there anything in particular about this virtual reality system that was causing any
discomfort?
On a scale from 1 to 10, 1 being how you felt upon arrival and 10 being, 'I want to stop,' at
what score would you choose to avoid using this virtual reality system?
Would you prefer a regular TV/monitor or this virtual reality system (if money was not an
issue) for the following (please circle):
a. Watching TV
TV
VR
b. Movies
TV
VR
c. Video games
TV
VR
d. Landmark exploration (e.g. Maps, museums, archeological sites)
TV
VR
e. Other (please list)
Figure 20: Original Usability Survey

137

1. How many hours per week on average do you play console video games? ____________
a. If yes how many hour do you play:
First Person Shooters (i.e. Halo)
_______
Platformers (i.e. Super Mario)

_______

Driving (i.e. Mario Cart)

_______

RPG (i.e. Final Fantasy)

_______

Fighting (i.e. Street Fighter)

_______

Other (Type:___________)

_______

2.
3.
4.
5.

How many day per month, on average to you have Headaches:____ Migraines:____
Do you use Glasses? Contacts? Neither? (circle)
Was the nose peice comfortable: Y / N
Was there anything in particular about this virtual reality system that was causing any
discomfort?
6. Were there any pressure points from the display? Y / N
a. If so, where:______________
7. On a scale from 1 to 10, 1 being how you felt upon arrival and 10 being, 'I want to stop,' at what
score would you choose to avoid using this virtual reality system? _______
8. Would you prefer a regular TV/monitor or this virtual reality system (if money was not an issue)
for the following (please circle):
a. Watching TV
TV
VIRTUAL REALITY
b. Movies
TV
VIRTUAL REALITY
c. Video games
TV
VIRTUAL REALITY
d. Landmark exploration (e.g. Maps, museums, archeological sites)
TV
VIRTUAL REALITY
e. Other (please list)

Figure 21: Revised Usability Survey

138

Motion Sickness Susceptibility Survey
This questionnaire is designed to find out how susceptible to motion sickness you are and what sorts of motion are most effective in causing that
sickness. Sickness here means feeling queasy or nauseated or actually vomiting. After some background questions, the questionnaire consists of
two sections: Section A is concerned with your childhood experiences of travel and motion sickness, that is, before the age of 12 years. Section B
is concerned with your experiences of travel and motion sickness over the last 10 years.
The correct way to answer each question is explained in the body of the questionnaire. It is important that you answer every question. Thank you
for your help.
Background Questions
1. Please State Your Age ____
2. Please State Your Sex (circle) Male / Female
3. Please State Your Current Occupation _________
4. Do you regard yourself as susceptible to motion sickness? (circle) Not at all / Slightly / Moderately / Very much so

Section A: You childhood experience only (before 12 years of age). For each of the following types of transport or
entertainment, please indicate:
5) As a child(before age 12), how often you travelled or experiences (tick boxes):
Never 1-4 trips 5-10trip 11 or more trips
Cars
Busses or Coaches
Trains
Aircraft
Small Boats
Ships, e.g. Channel Ferries
Swings
Roundabouts: playgrounds
Big dippers, carnival rides
6) As a child(before age 12), how often you felt sick or nauseated (tick boxes):
Never Rarely Sometimes Frequently Always
Cars
Busses or Coaches
Trains
Aircraft
Small Boats
Ships, e.g. Channel Ferries
Swings
Roundabouts: playgrounds
Big dippers, carnival rides
7) As a child(before age 12), how often you vomited (tick boxes):
Never Rarely Sometimes Frequently
Cars
Busses or Coaches
Trains
Aircraft
Small Boats
Ships, e.g. Channel Ferries
Swings
Roundabouts: playgrounds
Big dippers, carnival rides

Always

Figure 22: Full Golding's Motion Sickness Susceptibility Questionnaire (MSSQ) [141]

139

Figure 22 (cont'd)
Section B: You experience over the last 10 years (approximately). For each of the following types of transport or
entertainment, please indicate:
8) Over the last 10 years, how often you travelled or experiences (tick boxes):
Never 1-4 trips 5-10trip 11 or more trips
Cars
Busses or Coaches
Trains
Aircraft
Small Boats
Ships, e.g. Channel Ferries
Swings
Roundabouts: playgrounds
Big dippers, carnival rides
9) Over the last 10 years, how often you felt sick or nauseated (tick boxes):
Never Rarely Sometimes Frequently Always
Cars
Busses or Coaches
Trains
Aircraft
Small Boats
Ships, e.g. Channel Ferries
Swings
Roundabouts: playgrounds
Big dippers, carnival rides
10) Over the last 10 years, how often you vomited (tick boxes):
Never Rarely Sometimes Frequently
Cars
Busses or Coaches
Trains
Aircraft
Small Boats
Ships, e.g. Channel Ferries
Swings
Roundabouts: playgrounds
Big dippers, carnival rides

140

Always

Motion sickness susceptibility questionnaire short-form (MSSQ-Short)
This questionnaire is designed to find out how susceptible to motion sickness you are, and what sorts of
motion are most effective in causing that sickness. Sickness here means feeling queasy or nauseated or
actually vomiting
Your childhood experience only (before 12 years of age), for each of the following types of transport or
entertainment please indicate
1. As a child (before age 12), how often you felt sick or nauseated (tick boxes)
Your experience over the last 10 years (approximately), for each of the following types of transport or
entertainment please indicate

2. Over the last 10 years, how often you felt sick or nauseated (tick boxes)

Figure 23: Golding's [144] Short Motion Sickness Susceptibility Questionnaire

141

APPENDIX B

DATA TABLES
Table XXX: Literature Experiment Configurations
Study

Condition

Sickness

FOV
(Diag.)

Ours

HMD

27.268

35

24

N

Y

Screen Size 21.22043
Large
33.3
Stereo
Large Mono
28.5

35

24

N

Y

Gamepad
and head
Gamepad

80

24

N

Y

80

24

N

Low speed

31

60

30

med speed
High speed
Low
complexity
Med
complexity
High
complexity

33
48

60
60

8

So, Ho, and
Lo [15]

Watanabe
and Ujike[43]

Duration Sitting Stereo

Controller

Head
Tracked

Display/IVB Complexity

Y

HMD

high

Y

Flat Screen

high

Gamepad

Y

Flat Screen

high

N

Gamepad

Y

Flat Screen

high

Y

-

none

N

HMD

high

30
30

Y
Y

-

none
none

N
N

HMD
HMD

high
high

60

30

Y

-

none

N

HMD

low

29

60

30

Y

-

none

N

HMD

medium

45

60

30

Y

-

none

N

HMD

high

Hill

45

360

10

Y

Y

none

N

CAVE

-

Plane

30

360

10

Y

Y

none

N

CAVE

-

142

Table XXX (cont'd)

Stanney and Kennedy [145]

Short
Med
Long

Duh, Parker, and Furness
[139]1

Moss and Muth [29] (no
handrail)

Jeager and Mourant [59]

29.44
(41.03)
20.57
(27.23)
21.93
(30.96)

60

15

Y

-

mouse

Y

HMD

-

60

30

Y

-

mouse

Y

HMD

-

60

45

Y

-

mouse

Y

HMD

-

Bright Posture Difficulty

.85

180

4

N

-

none

N Dome/IVB

-

Dim Posture Difficulty
None Posture Difficulty
0ms delay, 2 mag
Inclusion
0ms delay, 2 mag
occlusion
0ms delay, .88 mag
Inclusion

0.9
1.26

180
180

4
4

N
N

-

none
none

N Dome/IVB
N
Dome

-

5

50

10

N N

none

Y

HMD

realistic

5.5

50

10

N N

none

Y

HMD/IVB

realistic

3.9

50

10

N N

none

Y

HMD

realistic

0ms delay, .88 mag
occlusion

7.5

50

10

N

none

Y

HMD/IVB

realistic

Treadmill

15.24

60* ~20 N

-

Mouse

23.94

60* ~20 N

-

N

143

Treadmill N
mouse

N

HMD
HMD

mediumhigh
mediumhigh

Table XXX (cont'd)
3.25
30 50 Y None
N Projection /IVB realistic
(60)
Big
4.25 (100) 50 50 Y None
N Projection /IVB realistic
Vinson et al.[107]
25.87
96 16 Y N
varied
N Desktop/IVB medium
So and Lo [118]
Moving
35
60 20 Y none
N
HMD
high
Not moving
14
60 20 Y None
N
HMD
high
Ujike, Yokoi, and Saida [104]
Very small
3
13 20 Y N
None
N
TV
realistic
Small
3.9
19 20 Y N
None
N
TV
realistic
Med
4.5
28 20 Y N
None
N
TV
realistic
Large
8.9
43 20 Y N
None
N
TV
realistic
Lin et al [105]
Small
0.47
60 8 Y Y
None
N Car simulator
high
SSQ non transformed-horizontal FOV
med
0.82
77 8 Y Y
None
N Car simulator
high
Large
1.49
100 8 Y Y
None
N Car simulator
high
x-large
1.65
180 8 Y Y
None
N Car simulator
high
Draper et al [50]
neutral
5
30 30 - N
head
Y
HMD
realistic
2
Keshavarz et al. [82]
Full PW
9
60 18 Y None
N
PowerWall
realistic
HMD
3.8
60 18 Y None
N
HMD
realistic
Reduced PW
4
36 18 Y None
N
PowerWall
realistic
Masked PW
3.8
60 18 Y None
N
PowerWall
realistic
Stanney et al [60]
retraining
19.9
50 30 Y Y
mouse
HMD
Stoffregen et al. [30]
Seated close
101
77 50 Y N Game pad N
TV
realistic
Standing close
58.7
77 50 N N Game pad N
TV
realistic
Seated far
60.5
43 50 Y N Game pad N
TV
realistic
Mehri et. a; [140]
Seated
58.1
77 50 Y N Game pad N
HMD
realistic
Standing
63.6
77 50 N N Game pad N
HMD
realistic
Toet et al.[71] Misc (SSQ-T estimate)

Small

144

Table XXX(cont'd)
Hakkinen, Vuori, Puhakka [61]3

TV
HMD game
HMD movie

7
14.5
9.5

35.5
35.5

40
40
40

Y N
Y Y
Y N

Kolasinki and Gilson [99]

Game

21.22

30

20

Y Y

Mourant and Thattacheny [120]
SSQ-O (E. SSQ-T)

City

60

7

Y

60

Rural
Highway
Park et al.[56]4
Keshavarz and Hecht [84]

Sharples et al. [108]

1

Keshavarz and Hecht[146]

Session 3
Real 3d
Real 2D
Sim 3D
Sim 2D
HMD
Desktop
Theater
Projection
passive
Projection
active
Experiment 1

none
Car simulator
none
Mouse and
head

N
N
N

TV
HMD
HMD

realistic
realistic
realistic

Y

HMD

-

-

Car simulator

Y

HMD

high

7

Y Y

Car simulator

Y

HMD

high

60

7

Y Y

Car simulator

Y

HMD

high

135^^^
74
74
74
74
60
17in
156

20^^
14
14
14
14
30
30
30

Y
Y
Y
Y
Y
-

N
Y
N
Y
N
-

Car simulator
none
none
none
none
Mouse
Mouse
None

N
N
N
N
N
-

TV
Projection
Projection
Projection
Projection
HMD
Desktop
Projection

medium
realistic
realistic
high
high
medium
medium
medium

22.88 1725mm

30

-

-

none

-

Projection medium

11.5

1725mm

30

-

-

mouse

-

Projection medium

54.55

60

18

Y N

15
(52.1)
9
(31.2)
7
(24.3)
19.5^
38
40
44
37
29.92
15.35
26.18

none

N projection

Equipment specs were not actually reported in the study, but found in another publication presumably using the same equipment.

2

A IVB enforces expected stimuli. In very familiar setting such as a virtual car, the IVB enforcing the real world is opposite of expectations.

3

Although all conditions were seated, there was less physical support in the game condition.

145

realistic

4

SSQ-T scores calculated from reported SSQ-N, D, and O. Duration shorten to reflect 15 minute break between sessions Due to the use of monitor, this system

did not have the FOV-V that most systems of this width would have.

Table XXXI: Cross Configuration Data Set
Duration

Seated

Controller

35
35
80
80
60
60
60
60
60
60

24
24
24
24
30
30
30
30
30
30

N
N
N
N
Y
Y
Y
Y
Y
Y

joystick
joystick
joystick
joystick
none
none
none
none
none
none

Head
Tracked
Y
Y
Y
Y
N
N
N
N
N
N

45

360

10

Y

none

plane

30

360

10

Y

Short

29.44

60

15

Med
long
Bright Posture
Difficulty
Dim Posture
Difficulty
None Posture
Difficulty

20.57
21.93

60
60

0.85

Study

Condition

Ours

HMD
Screen Size
Large Stereo
Large Mono
Low speed
med speed
High speed
Low complexity
Med complexity
High complexity

27.26
21.22
33.3
28.5
31
33
48
8
29
45

Hill

So, Ho, and Lo [15]

Watanabe and Ujike
[40]
Stanney and
Kennedy [145]

Duh, Parker, and
Furness [139]

Sickness FOV

IVB

Stereo

Realism

N
Y
Y
Y
N
N
N
N
N
N

Y
Y
Y
N
N
N
N
N
N
N

high
high
high
high
high
high
high
low
medium
high

N

N

Y

high

none

N

N

Y

high

Y

mouse

Y

N

N

high

30
45

Y
Y

mouse
mouse

Y
Y

N
N

N
N

high
high

180

4

N

none

N

Y

N

high

0.9

180

4

N

none

N

Y

N

high

1.26

180

4

N

none

N

N

N

high

146

Table XXXI (cont'd)
Moss and muth [81]
(no handrail)

Jeager and Mourant
[57]

0ms delay, 2
mag Inclusion
0ms delay, 2
mag occlusion
0ms delay, .88
mag
0ms delay, .88
mag occlusion

5

50

10

N

none

Y

N

N

realistic

5.5

50

10

N

none

Y

Y

N

realistic

3.9

50

10

N

none

Y

N

N

realistic

7.5

50

10

N

none

Y

Y

N

realistic

Treadmill

15.24

60

20

N

simulator

N

N

N

medium

mouse

23.94

60

20

N

mouse

N

N

N

medium

Toet et al. [70] Misc
(SSQ--T estimate)

small

3.25

30

50

Y

None

N

Y

N

realistic

Vinson et al. [146]
So and Lo [118]

big
Moving
Not moving

4.25
25.87
35
14

50
96
60
60

50
16
20
20

Y
Y
Y
Y

None
joystick
none
None

N
N
N
N

Y
Y
N
N

N
N
N
N

realistic
medium
high
high

Ujike, Yokoi, and
Saida [104]

Very small

3

13

20

Y

None

N

Y

N

realistic

Small
Med
large
Small

3.9
4.5
8.9
0.47

19
28
43
60

20
20
20
8

Y
Y
Y
Y

None
None
None
None

N
N
N
N

Y
Y
Y
Y

N
N
N
Y

realistic
realistic
realistic
high

med

0.82

77

8

Y

None

N

Y

Y

high

Large
x-large

1.49
1.65

100
180

8
8

Y
Y

None
None

N
N

Y
Y

Y
Y

high
high

Lin et al [105]
SSQ non
transformedhorizontal FOV

147

Table XXXI (cont'd)
Draper et al [47]
Keshavarz et al. [83]

Stanney et al [58]
Stoffregen et al.
[74]

Mehri et. a; [141]
Hakkinen, Vuori,
Puhakka [59]

Kolasinki and Gilson
[100]
Mourant and
Thattacheny [119]
SSQ-O (E. SSQ-T)

Park et al. [54]

Neutral
Full PW
HMD
Reduced PW
Masked PW
retraining

5
9
3.8
4
3.8
19.9

30
60
60
36
60
50

30
18
18
18
18
30

N
Y
Y
Y
Y
Y

head
None
None
None
None
mouse

Y
N
N
N
N
Y

N
N
Y
N
N
N

N
N
N
N
N
Y

realistic
realistic
realistic
realistic
realistic
high

Seated close

101

77

50

Y

joystick

N

Y

N

realistic

Standing close
Seated far
Seated
standing

58.7
60.5
58.1
63.6

77
43
77
77

50
50
50
50

N
Y
Y
N

joystick
joystick
joystick
joystick

N
N
N
N

Y
Y
N
N

N
N
N
N

realistic
realistic
realistic
realistic

TV

7

30

40

Y

none

N

Y

N

realistic

HMD game

14.5

40

Y

simulator

N

N

Y

realistic

HMD movie

9.5

40

Y

none

N

Y

N

realistic

game

21.22

30

20

Y

head

Y

N

Y

high

City

15

60

7

Y

simulator

Y

N

Y

high

Rural
highway
Session 3

9
7
19.5

60
60
135

7
7
20

Y
Y
Y

simulator
simulator
simulator

Y
Y
N

N
N
Y

Y
Y
N

high
high
medium

35.
5
35.
5

148

Table XXXI (cont'd)
Keshavarz and
Hecht [85]

Sharples et al. [108]

Keshavarz and
Hecht [147]

Real 3d

38

74

14

Y

none

N

Y

Y

realistic

Real 2D
Sim 3D
Sim 2D
HMD
Desktop
Theater
Projection
passive
Projection active

40
44
37
29.92
15.35
26.18

74
74
74
60
30
156

14
14
14
30
30
30

Y
Y
Y
N
Y
N

none
none
none
mouse
Mouse
None

N
N
N
Y
N
Y

Y
Y
Y
N
Y
Y

N
Y
N
N
N
N

realistic
high
high
medium
medium
medium

22.88

80

30

Y

none

N

Y

N

medium

11.5

80

30

Y

mouse

N

Y

N

medium

Experiment 1

54.55

60

18

Y

none

N

Y

N

realistic

149

Table XXXII: Some Factors that Require Further Study
1. Long term
Habituation loss

2. Gestural Navigation

3. Orientation only versus
6DOF head tracking
7. Severity of the vision
correction

5. Color

6. Postural Pre-exposure

9. Percentage of
real world

10. Postural monitoring

11. Time of day

13. Contrast

14. Stopping speed
(acceleration)

15. Analytical/details
versus global/whole
perception

16. Sensor drift acceptance

17. Swivel chair
navigation

18. calibration

19. Rotation speed

20. Other IVB (e.g. icons)

21. Screen luminance

22. Vertical field of view

23. Individually adjusted
gain

24. Percentage of time actively
interacting with the system

150

4. Head movements
8. Vertical versus horizontal
cues
12. Time of day (group by day
versus night person)

APPENDIX C

TERMINOLOGY AND DEFINITIONS

Accommodation - the eyes' adjustment of the lens’ focal length to the perceived distance.
Augmented reality (AR) - any simulated environment whose visual content is partially produced
by a computer and the appearance of the environment changes according to the participant's
behavior.
Coefficient of determination (R2) - the proportion of variability or variance that a model
explains.
Computer Aided Virtual Environments (CAVE) - three or more large screens with head tracking.
What is sometimes referred to as a one-wall CAVE is considered to be just a large screen and is
sometimes called a PowerWall.
- The continuous variant of the Poisson function.
Degrees of Freedom (DOF) - the number of axes of translation and rotation when navigating a
virtual environment.
Field of view (FOV) - the subtended angle an object fills in a visual field.
Heterophoria - the extent that one eye moves away from vergence after the other eye is
covered.
Immersion - the time spent inside a virtual environment.
Independent Visual Backgrounds (IVB) - visual stimuli that enforce real world expectations.
Motion History Questionnaire (MHQ) - a questionnaire to determine a past history of motion
sickness.
Motion Sickness Susceptibility Questionnaire (MSSQ) – a questionnaire to determine a past
history of motion sickness and susceptibility.
Our HMD - refers to our experiment that compared a lightweight HMD to a heavy HMD.
Our Screen - refers to our experiment that compared a small screen to a large screen, but held
the field of view and standing position constant.

151

Our Render - refers to our experiment that compared a monoscopic application to a stereo
application.
Simulator Sickness Questionnaire (SSQ-T/N/D/O) - a standard cybersickness measurement. SSQ
refers to the questionnaire, SSQ-T refers to the total score, SSQ-N refers to the nausea score,
SSQ-D refers to the disorientation score, and SSQ-O refers to the oculomotor score.
Vergence - the eyes' movement to focus on the same object.
Virtual environment (VE) - any simulated environment whose visual content is at least partially
produced by a computer and the appearance of the environment changes according to the
participant's behavior. VE include both virtual and augmented reality.
Virtual reality (VR) - a simulated environment whose visual content is entirely produced by a
computer and the appearance of the environment changes according to the participant's
behavior.
Zero Inflated Negative Binomial (ZINB) - a statistical distribution model that includes a model .

152

BIBLIOGRAPHY

153

BIBLIOGRAPHY

1

Kolasinski EM. Prediction of Simulator Sickness in a Virtual Environment. Thesis. University
of Central Florida; 1996.

2

Ujike H, Ukaib K, Niheic K. Survey on motion sickness-like symptoms provoked by viewing a
video movie during junior high school class. Displays. 2008 March;29(2):81–89.

3

Oculus
Rift.
OculusBestPractices.
[Internet].
2014
Available
http://static.oculusvr.com/sdk-downloads/documents/OculusBestPractices.pdf.

4

Chen YC, Dong X, Hagstrom J, Stoffregen TA. Control of a Virtual Ambulation Influences
Body Movement and Motion Sickness. BIO Web of Conferences The International
Conference SKILLS. 2011;1.

5

Kim YY, Kim HJ, Kim EN, Ko HD, Kim HT. Characteristic changes in the physiological
components of cybersickness. Journal on Applied Signal Processing. 2004 January;2004.

6

Nichols S, Haldane C, Wilson JR. Measurement of presence and its consequences in virtual
environments. International Journal of Human-Computer Studies. 2000 March;52(3):471491.

7

Samsung. 3D-TV Warning. [Internet]. 2011 [cited 2013 May 10]. Available from:
http://www.samsung.com/ca/pdf/3D-tv-warning_en.pdf.

8

Nintendo. Nintendo 3DS - Health and Safety Precautions. [Internet]. 2013 [cited 2013 May
11].
Available
from:
http://www.nintendo.com/consumer/info/en_na/docs.jsp?menu=3ds&submenu=ctr-dochealth-safety.

9

Kolasinski EM. Simulator Sickness in Virtual Environments. Final Technical repot.
Alexandria: Army Research Inst for the Behavioral and Social Sciences; 1995.

from:

10 So RHY. The Search For A Cybersickness Dose Value.; 1999.
11 Simon CW. The Use of Central-Composite Design in Human Factors Engineering
Experiments. Technical report. Urbana, Illinois: Aviation Research Labratory; 1970. AFOSR70-6.
12 Simon CW. New Research Paradigm for Applied Experimental Psychology: A System
Approach. Military Report. Westlake Village, California: Canyon Research Group Inc.; 1978.

154

CWS-04-77A.
13 Simon CW. Economical Multifactor Designs for Human Factors Engineering Experiments.
Military Report. Culver City, California: Equipment Engineering Division, Aerospace Group,
Hughes Aircraft Company; 1973. F73-326.
14 Jones MB, Kennedy RS, Stanney KM. Toward Systematic Control of Cybersickness.
Presence. 2004 October;13(5):589-600.
15 So RHY, Ho A, Lo WT. A Metric to Quantify Virtual Scene Movement for the Study of
Cybersickness: Definition, Implementation, and Verification. Presence. 2001
April;10(2):193-215.
16 Mon-Williams M, Wann JP, Rushton S. Design factors in stereoscopic virtual-reality
displays. Journal of the Society for Information Display. 1995 December;3(4):207-210.
17 Stanney KM, Kennedy RS, Drexler JM. Cybersickness is not simulator sickness. In:
Proceedings of the Human Factors and Ergonomics Society; 1997; Santa Monica.
18 Blea W, Bos JE, de Graaf B, Groen E, Wertheim AH. Motion Sickness: Only One Provocative
Conflict? Brain Research Bulletin. 1998 November;47(5):481–487.
19 Bos JE, Bles W, Groen EL. A Theory on Visually Induced Motion Sickness. Displays. 2008
March;29(2):47-57.
20 Smart LJJ, Otten EW, Stoffregen TA. It’s turtles all the way down: A comparative analysis of
visually induced motion sickness.; 2007.
21 Lampton DR, Knerr BW, Goldberg SL, Bliss JP, Moshell MJ. The Virtual Environment
Performance Assessment Battery (VEPAB): Development and Evaluation. Interim Report.
Alexandria: Army Research Inst. For the Behaviorial and Social Sciences; 1994.
22 Ehrlich SD. Motion sickness. [Internet]. 2012 [cited 2014 April 1]. Available from:
http://umm.edu/health/medical/altmed/condition/motion-sickness.
23 Stanney KM, Hale KS, Nahmens I, Kennedy RS. What to Expect from Immersive Virtual
Environment Exposure: Influences of Gender, Body Mass Index, and Past Experience.
Human Factors: The Journal of the Human Factors and Ergonomics Society. 2003
Fall;45(3):504-520.
24 Stanney KM, Kingdon KS, Kennedy RS. Dropouts and Aftereffects: Examining General
Accessibility to Virtual Environment Technology. Proceedings of the Human Factors and
Ergonomics Society Annual Meeting. 2002 September;46(26):2114-2118.

155

25 Kingdon KS, Stanney KM, Kennedy RS. Extreme responses to virtual environment exposure.;
2001; Santa Monica.
26 LaViola JJJ. A Discussion of Cybersickness in Virtual Environmnts. ACM SIGCHI Bulletin. 2000
January;32(1):47-56.
27 Virre E. Virtual Reality and the Vestibular Apparatus. IEEE Engineering in Medicine and
Biology Magazine. 1996 Mar/Apr;15(2):41 - 43, 69.
28 Jaekl PM, Jenkin MR, Harris LR. Perceiving a stable world during active rotational and
translational head movements. Experimental Brain Research. 2005 June;163(3):388-399.
29 Moss JD, Muth ER. Characteristics of Head-Mounted Displays and Their Effects on
Simulator Sickness. Human Factors: The Journal of the Human Factors and Ergonomics
Society. 2011 June;53(3):308-319.
30 Stoffregen TA, Faugloire E, Yoshida K, Flanagan MB, Merhi O. Motion Sickness and Postural
Sway in Console Video Games. Human Factors: The Journal of the Human Factors and
Ergonomics Society. 2008 April;50(2):322-331.
31 Villard SJ, Flanagan MB, Albanese GM, Stoffregen TA. Postural Instability and Motion
Sickness in a Virtual Moving Room. Human Factors: The Journal of the Human Factors and
Ergonomics Society. 2008 April;50(2):332-345.
32 Prothero JD, Draper MH, Furness TA, Parker DE, J. WM. The use of an independent visual
background to reduce Simulator side-effects. Aviation, Space, and Environmental Medicine.
1997;70:277-283.
33 Kennedy RS, Stanney KM. Postural instability induced by virtual reality exposure:
Development of a certification protocol. International Journal of Human-Computer
Interaction. 1996;8(1):25-47.
34 Kennedy RS, Lane NE, Berbaum KS, Lilienthal MG. Simulator Sickness Questionnaire: An
Enhanced Method for Quantifying Simulator Sickness. The International Journal of Aviation
Psychology. 1993;3(3):203-220.
35 Bouchard S, Robillard G, Renaud P. Revising the factor structure of the Simulator Sickness
Questionnaire. In: Wiederhold BK, Bouchard S, Riva G, editors. Annual Review of
CyberTherapy and Telemedicine Volumne 5. San Diego: Interactive Media Institute; 2007.
36 Young SD, Adelstein BD, Ellis SR. Demand Characteristics of a Questionnaire Used to Assess
Motion Sickness in a Virtual Environment.; 2006; Alexandria.
37 Ling Y, Nefs HT, Brinkman WP, Qu C, Heynderickx I. The Relationship between Individual
Characteristics and Experienced Presence. Computers in Human Behavior. 2013;29:1519–

156

1530.
38 Cobb SVG. Measurement of postural stability before and after immersion in a virtual
environment. Applied Ergonomics. 1999 February;30(1):47-57.
39 Ehrlich JA. Simulator sickness and HMD configurations.; 1997; Pittsburgh.
40 Chang CH, Pan WW, Tseng LY, Stoffregen TA. Postural Activity and Motion Sickness During
Video Game Play in Children and Adults. Experimental Brain Research. 2011 January 1-11.
41 Roberts WK, Gallimore JJ. A Physiological Model of Cybersickness during Virtual
Environment Interaction. Proceedings of the Human Factors and Ergonomics Society
Annual Meeting. 2005 September;49(26):2230-2234.
42 Kiryu T, Uchiyama E, Jimbo M, Iijima A. Time-Varying Factors Model with Different TimeScales for Studying Cybersickness. Lecture Notes in Computer Science. 2007b;4563:262269.
43 Watanabe H, Ujike H. The Activity of ISO/Study Group on “Image Safety” and Three
Biological Effect. Second International Symposium on Universal Communication. 2008 210214.
44 Oyamada H, Iijima A, Tanaka A, Ukai K, Toda H, Sugita N, Yoshizawa M, Bando T. A pilot
study on pupillary and cardiovascular changes induced by stereoscopic video movies.
Journal of NeuroEngineering and Rehabilitation. 2007;4(37).
45 Harvey C, Howarth PA. The Effect of Display Size on Visually-Induced Motion Sickness
(VIMS) and Skin Temperature.; 2007; Hong Kong.
46 Ji J, So RHY, Lor F, Cheung TFR, Howrth P, Stanney K. A Search for Possible Neural Pathways
Leading to Visually Induced Motion Sickness. Vision. 2005;17(2):131-134.
47 Chang E, Hwang I, Jeon H, Chun Y, Kim HT, Park C. Effects of Rest Frames on Cybersickness
and Oscillatory Brain Activity. In: International Winter Workshop on Brain-Computer
Interface; 2013; High 1, Sabuk-eup.
48 Renkewitz H, Alexander T. Perceptual Issues of Augmented and Virtual Environments.
Wachtberg : FGAN-FKIE; 2007.
49 van Emmerik ML, de Vries SC, Bos JE. Internal and external fields of view affect
cybersickness. Displays. 2011 October;32(4):169-174.
50 Draper MH, Viirre ES, Furness TA, Gawron VJ. Effects of Image Scale and System Time Delay
on Simulator Sickness within Head-Coupled Virtual Environments. Human Factors: The

157

Journal of the Human Factors and Ergonomics Society. 2001 Spring;43(1):129-146.
51 Peli E. The visual effects of head-mounted display (HMD) are not distinguishable from
those of desktop computer display. Vision Research. 1998 June;38(13):2053-2066.
52 Howarth PA, Costello PJ. The Occurrence of Virtual Simulation Sickness Symptoms when An
HMD was used As a Personal Viewing System. Display. 1997 December;18(2):107-116.
53 Chen W, Yuen SL, So RHY. A Progress Report on the Quest to Establish a Cybersickness Dose
Value. Proceedings of the Human Factors and Ergonomics Society Annual Meeting. 2002
September;46(26):2119-2123.
54 Rebenitsch L, Owen C. Individual Variation in Susceptibility to Cybersickness. In: User
Interface Software and Technology Symposium; 2014; Waikiki, HI. Available from:
http://dx.doi.org/10.1145/2642918.2647394.
55 Arns LL, Cerne MM. The Relationship Between Age and Incidence of Cybersickness Among
Immersive Environment Users. In: Proceedings of the IEEE Virtual Reality 2005; 2005; Bonn,
Germany. p. 267-268.
56 Park GD, Allen RW, Fiorentino D, Rosenthal TJ, Cook ML. Simulator Sickness Scores
According to Symptom Susceptibility, Age, and Gender for an Older Driver Assessment
Study. In: Proceedings of the Human Factors and Ergonomics Society Annual Meeting;
2006; San Francisco, USA. p. 2702-2706.
57 Reason J. Motion Sickness—Some Theoretical and Practical Considerations. Applied
Ergonomics. 1978;9(3):163–167.
58 Ujike H, Yokoi T, Saida S. Effects of Display Size on Vection and Visually Induced Motion
Sickness.; 2005; Takamatsu.
59 Jaeger BK, Mourant RR. Comparison of simulator sickness using static and dynamic walking
simulators. In: Proceedings of the Human Factors and Ergonomics Society; 2001; Santa
Monica.
60 Stanney KM, Kennedy RS, Drexler JM, Harm DL. Motion sickness and proprioceptive
aftereffects following virtual environment exposure. Applied Ergonomics. 1999
February;30(1):27-38.
61 Hakkinen J, Vuori T, Paakka M. Postural stability and sickness symptoms after HMD use.
International Conference on Systems, Man and Cybernetics. 2002 October 147-152.
62 Harm DL, Taylor LC, Bloomberg JJ. Adaptive Changes in Sensorimotor Coordination and
Motion Sickness Following Repeated Exposures to Virtual Environments. Technical report.

158

League City: NASA Human Research Program Investigators' Meeting; 2007.
63 Clemes SA, Howarth PA. The Menstrual Cycle and Susceptibility to Virtual Simulation
Sickness. Journal of Biological Rhythms. 2005 February;20(1):71-82.
64 Burg A. Lateral Visual Field as related to Age and Sex. Journal oj Applied Psychology.
1968;52(1):10-15.
65 Graeber DA, Stanney KM. Gender Differences in Visual Induced Motion Sickness.
Proceedings of the Human Factors and Ergonomics Society 46th Annual Meeting. 2002
2109-2113.
66 Klosterhalfen S, Kellermann S. Gender and Race as Determinants of Nausea Induced by
Circular Vection. Gender Medicine. 2006;3(3):236-242.
67 Klosterhalfen S, Kellermann S, Pan F, Stockhorst U, Hall G, Enck P. Effects of Ethnicity and
Gender on Motion Sickness. Aviation Space Environment Medicine. 2005;76:1051-1057.
68 Stern RM, Hu S, LeBlanc R, Koch KL. Chinese Hyper-susceptibility to Vection-induced
Motion Sickness. Aviation Space and Environmental Medicine. 1993;64(9):827-830.
69 Stanney KM, Kennedy RS. The Psychometrics of Cybersickness. Communications of the
ACM. 1997 Aug. ;40(8):66-68.
70 Hill KJ, Howarth PA. Habituation to the side effects of immersion in a virtual environment.
Displays. 2000 March;21(1):25-30.
71 Toet A, de Vries SC, van Emmerik ML, Bos JE. Cybersickness and desktop simulations: field
of view effects and user experience. Proceedings of the SPIE Enhanced and Synthetic
Vision. 2008;6957:69570-69611.
72 Howarth PA, Hodder SG. Characteristics of habituation to motion in a virtual environment.
Displays. 2008 March;29(2):117-123.
73 Bailenson JN, Yee N. A Longitudinal Study of Task Performance, Head Movements,
Subjective Report, Simulator Sickness, and Transformed Social Interaction in Collaborative
Virtual Environments. PRESENCE: Teleoperators and Virtual Environments. 2006;15(6):699716.
74 Regan C. An Investigation into Nausea and Other Side-Effects of Head-Coupled Immersive
Virtual Reality. Virtual reality. 1995;1(1):17-31.
75 Kharbanda P, Ganjoo RK, Chadha DS. A Strange Case of Headache in a Fighter Aircrew: a
Case Study. Indian J Aerospace Med. 2008;52:20-25.

159

76 Barbosa JS, Lorenzatto W, Tulius Silva M. Helmet-Related, External Compression Headache
Among Police Officers in Rio de Janeiro. Medscape General Medicine. 2004;6(2):45.
77 So RHY, Lo WT, Ho ATK. Effects of Navigation Speed on Motion Sickness Caused by an
Immersive Virtual Environment. Human Factors: The Journal of the Human Factors and
Ergonomics Society. 2001b Fall;43(3):452-461.
78 So RHY, Lo WT. Cybersickness: An Experimental Study to Isolate the Effects of Rotational
Scene Oscillations. In: Proceedings of IEEE Virtual Reality '99 Conference; 1999; Houston.
79 Stanney KM, Kingdon KS, Graeber D, Kennedy RS. Human Performance in Immersive Virtual
Environments: Effects of Exposure Duration, User Control, and Scene Complexity. Human
Performance. 2002;15(4):339-366.
80 Lo WT, So RHY. Cybersickness in the presence of scene rotational movements in different
axis. Applied Ergonomics. 2001 February;32(1):1-14.
81 Stanney KM, Lanham DS, Kennedy RS, Breaux R. Virtual environment exposure drop-out
thresholds. In: Proceedings of the Human Factors and Ergonomics Society; 1999; Santa
Monica.
82 Keshavarz B, Hecht H, Zschutschke L. Intra-visual conflict in visually induced motion
sickness. Displays. 2011 October;32(4):181–188.
83 Keshavarz B, Hecht H. Axis Rotation and Visually Induced Motion Sickness: The Role of
Combined Roll, Pitch, and Yaw Motion. Aviation, Space, and Environmental Medicine. 2011
November:1023-1029.
84 Keshavarz B, Hecht H. Stereoscopic Viewing Enhances Visually Induced Motion Sickness but
Sound Does Not. Presence. 2012 Spring:213-228.
85 Yang S, Sheedy JE. Effects of Vergence and Accommodative Responses on Viewer’s Comfort
in Viewing 3D Stimuli. In: Stereoscopic Displays and Applications XXII; 2011; San Francisco,
USA.
86 Parsons TD, Larson P, Kratz K, Thiebaux M, Bluestein B, Buckwalter JG, Rizzo AA. Sex
Differences in Mental Rotation and Spatial Rotation in a Virtual Environment.
Neuropsychologia. 2004;42:555–562.
87 Barret GV, Thornton CL. Relationship between Perceptual Style and Simulator Sickness.
Journal of Applied Psychology. 1986;52(4):304-308.
88 Rash CE. A 25-year Retrospective Review of Visual Complaints and Illusions Associated with
a Monocular Helmet-Mounted Display. Displays. 2008;29:70-80.

160

89 Hale KS, Stanney KM. Effects of low stereo acuity on performance, presence and sickness
within a virtual environment. Applied Ergonomics. 2006 May;37(3):329–339.
90 Dong X, Stoffregen TA. Postural Activity and Motion Sickness among Drivers and Passengers
in a Console Video Game. Proceedings of the Human Factors and Ergonomics Society
Annual Meeting. 2010 September;54(18):1340-1344.
91 Dong X, Yoshida K, Stoffregen TA. Control of a Virtual Vehicle Influences Postural Activity
and Motion Sickness. Journal of Experimental Psychology: Applied. 2011 June;17(2):128138.
92 Bos JE, Ledegang WD, Lubeck AJA, Stins JF. Cinerama sickness and postural instability.
Ergonomics. 2013;56(9):1430-1436.
93 Solimini AG, Mannocci A, Domitilla DT, La Torre G. A survey of visually induced symptoms
and associated factors in spectators of three dimensional stereoscopic movies. BMC Public
Health. 2012 September;12(1):1-11.
94 Bos JE, de Vries SC, van Emmerik ML, Groen EL. The Effect of Internal and External Fields of
View on Visually Induced Motion Sickness. Applied Ergonomics. 2010 July;41(4):516-521.
95 Taha Z, Hartomo, Jen YH, Gadzila RAR, Chai APT. The Effect of Body Weight and Height on
Incidence of Cyber sickness Among Immersive Environment Malaysian Users. In: 17th
World Congress on Ergonomics; 2009; Beijing, China.
96 Howarth PA. Oculomotor changes within virtual environments. Applied Ergonomics.
1999;1(59-67):30.
97 Karpicka E, Howarth PA. Heterophoria Adaptation during the Viewing of 3D Stereoscopic
Stimuli. Ophthalmic and Physiological Optics. 2013 September;33(5):604–610.
98 Mon-Williams M, Wann JP. Binocular virtual reality displays: when problems do and don't
occur. Human Factors. 1998 March;40(1).
99 Kolasinski EM, Gilson RD. Simulator Sickness and Related Findings in a Virtual Environment.
In: Proceedings of the Human Factors and Ergonomics Society; 1998; Santa Monica.
100 Kinsella A. The Effect of 0.2 Hz and 1.0 Hz Frequency and 100 ms and 20-100 ms Amplitude
of Latency on Simulatory Sickness in a Head Mounted Display. Thesis. Tiger Prints; 2014.
101 Dizio P, Lackner JR. Circumventing Side Effects of Immersive Virtual Environments. In:
International Conference on Human-Computer Interaction; 1997; San Francisco.
102 Chen W, Plancoulaine A, Ferey N, Touraine D, Nelson J, Bourdot P. 6DoF Navigation in
Virtual Worlds: Comparison of Joystick-based and Head-controlled Paradigm. In:

161

Proceedings of the 19th ACM Symposium on Virtual Reality Software and Technology;
2013; Singapore.
103 Howarth PA, Finch M. The nauseogenicity of two methods of navigating within a virtual
environment. Applied Ergonomics. 1999 February;30(1):39-45.
104 McGee MK. Assessing Negative Side Effects in Virtual Environments. Masters Thesis.
Blacksburg, Virginia: Virginia Polytechnic Institute and State University; 1998.
105 Seay AF, Krum DM, Hodges L, Ribarsky W. Simulator Sickness and Presence in a High FieldOf-View Virtual Environment. CHI '02 extended abstracts on Human factors in computing
systems. 2002:784-785.
106 Duh HBL, Lin JJW, Kenyon RV, Parker DE, Furness TA. Effects of Field of View on Balance in
an Immersive Environment. In: Virtual Reality Proceedings; 2001; Yokohama.
107 Vinson NG, Lapointe JF, Parush A, Roberts S. Cybersickness Induced by Desktop Virtual
Reality. In: Proceedings of the 2012 Graphics Interace Conference; 2012; Toronto, Canada.
108 Sharples S, Cobb S, Moody A, Wilson JR. Virtual reality induced symptoms and effects
(VRISE): Comparison of head mounted display (HMD), desktop and projection display
systems. Displays. 2008 march;29(2):58–69.
109 Liu CL, Uang ST. Effects of Presence on Causing Cybersickness in the Elderly within a 3D
Virtual Store. In: Human Computer Interaction International: Users and Applications; 2011;
Orlando, USA.
110 Money KE. Motion Sickness. Physiological Reviews. 1970;50(1).
111 McClure JA, Fregly AR. Effect of Environmental Temperature on Sweat onset During Motion
Sickness. Technical Report. Pensacola, Forida: Bureau of Medicine and Surgery; 1971 Naval
Aerospace Medical Research Laboratory. AD0740801.
112 Kim H, Park J, Noh K, Gardner CJ, Deok Kong S, Kim J, Jin S. An X–Y Addressable Matrix
Odor-Releasing System Using an On–Off Switchable Device. Angewandte Chemie
International. 2011;50(30):6903 –6907.
113 So RHY, Lo WT, Ho ATK. Effects of Navigation Speed on Motion Sickness Caused by an
Immersive Virtual Environment. Human Factors: The Journal of the Human Factors and
Ergonomics Society. 2001b Fall;43(3):452-461.
114 Lawther A, Griffin MJ. Prediction of the Incidence of Motion Sickness from the Magnitude,
Frequency, and Duration of Vertical Oscillation. The Journal of the Acoustical Society of
America. 1987;82(3):957-966.

162

115 O'Hanlon JF, McCauley ME. Motion Sickness Incidence as a Function of the Frequency and
Acceleration of Vertical Sinusoidal Motion. Technical Report. Springfeild, Virginia: National
Tecnical Information Service; 1973 Canyon Research Group Inc Goleta CA Human Factors
Research. AD0768215.
116 McCauley ME, Royal JW, Wylie CD, O'Hanlon JF, Mackie RR. Motion Sickness Incidence:
Exploratory Studies of Habituation, Pitch and Roll, and the Refinment of a Mathmatical
Model. Technical report. Goleta, California: Human Factors Research Inc.; 1976 Office of
Naval Research. AD-A024709.
117 Flanagan MB, May JG, Dobie TG. The Role of Vection, Eye Movements, and Postural
Instability in the Etiology of Motion Sickness. Journal of Vestibular Research. 2004;14:335346.
118 So RHY, Lo WT. Cybersickness with Virtual Reality Training Applications: a claustrophobia
phenomenon with head-mounted displays? In: Proceeding of the First World Congress on
Ergonomics For Global Quality and Productivity; 1998; Hong Kong.
119 Chen W,CJZ, So RHY. Visually Induced Motion Sickness: Effects of Translational Visual
Motion along Different Axes. Contemporary Ergonomics 2011. 2011.
120 Mourant RR, Thattacheny TR. Simulator Sickness in a Virtual Environments Driving
Simulator. Proceedings of the Human Factors and Ergonomics Society Annual Meeting.
2000 July;44(5):534-537.
121 Duh HBL, Parker DE, Philips JO, Furness TA. "Conflicting" Motion Cues to the Visual and
Vestibular Self- Motion Systems Around 0.06 Hz Evoke Simulator Sickness. Human Factors:
The Journal of the Human Factors and Ergonomics Society Spring. 2004;46(1):142-153.
122 Chow EHC, Ji JTT, So RHY, Cheung RTF. The Effects of Visual Stimulus Oscillation Frequency
on Postural Disturbance in Roll and in Fore-and-Aft Direction. In: 165-169; 2007; Hong
Kong. p. 165-169.
123 Chen D, So R, Kwok K, Cheung R. Visually Induced Motion Sickness after Watching Scenes
Oscillating at Different Frequencies and Amplitudes. In: McCabe PT, editor. Ergonomics &
Human Factors; 2012 July; Blackpool, UK. p. 253-260.
124 Bonato F, Bubka A, Palmisano S. Combined Pitch and Roll and Cybersickness in a Virtual
Environment. Aviation, Space, and Environmental Medicine. 2009 November;80(11):941945.
125 Kiryu T, Nomura E, Iijima A, Bando T. Time-frequency structure of image motion vectors
under vection-induced cybersickness. In: Human Factors Issues in Human-Computer
Interaction; 2005; Las Vegas.

163

126 Hettinger LJ, Berbaum KS, Kennedy WP. Vection and Simulator Sickness. Military
Psychology. 1990;2(3):171-181.
127 Webb NA, Griffin MJ. Eye Movement, Vection, and Motion Sickness with Foveal and
Peripheral Vision. Aviation, Space, and Envionmental Medicine. 2003;74:6225-625.
128 Ji TY, So RHY, Cheung RTF. Isolating the Effects of Vection and Optokinetic Nystagmus on
Optokinetic Rotation-Induced Motion Sickness. Human Factors. 2009;51(5):pp.739-751.
129 Golding JF, Doolan K, Acharya A, Tribak M, Gresty MA. Cognitive cues and visually induced
motion sickness. Aviation, space, and environmental medicine. 2012 May:477-482.
130 Stanney KM, Hash P. Locus of User-Initiated Control in Virtual Environments: Influences on
Cybersickness. Presence. 1998 October;7(5):447-459.
131 Kalloniatis M, Luu C. Temporal Resolution. In: Kolb H, Nelson R, Fernandez E, Jones B,
editors. Webvision: The Organization of the Retina and Visual System. Vol Part VIII.
University of Utah Health Sciences Center; 2007.
132 Bowmaker JK, Dartnall HJ. Visual pigments of rods and cones in a human retina. The Journal
of Physiology. 1980;1(298):501-511.
133 So RHY, Yuen SL. Comparing Symptoms of Visually Induced Motion Sickness Among Viewers
of Four Similar Virtual Environments with Different Color. Proceedings of the 2nd
international conference on Virtual reality. 2007 386-391.
134 Bonato F, Bubka A, Alfieri L. Display Color Affects Motion Sickness Symptoms in an
Optokinetic Drum. Aviation and Environmental Medicine. 2004;75(4):306-311.
135 Matthews ML. The influence of colour on CRT reading performance and subjective comfort
under operational conditions. Applied Ergonomics. 1987;18(4):323 328.
136 Mon-Williams M, Plooy A, Burgess-Limerix R, Wann J. Gaze angle: a possible mechanism of
visual stress in virtual reality headsets. Ergonomics. 1998 ;41(3):280-285.
137 Diels C, Ukai K, Howarth PA. Visually Induced Motion Sickness with Radial Displays: Effects
of Gaze Angle and Fixation. Aviation, Space, and Environmental Medicine. 2007
July;78(7):659-665.
138 Duh HBL, Abi-Rache H, Parker DE, Furness TA. Effects on balance distubance of
manipulating depth of an independent visual background in a stereographic display. In:
Proceedings of the Human Factors and Ergonomics Society; 2001; Santa Monica.
139 Duh HBL, Parker DE, Furness TA. An "independent visual background" reduced balance
disturbance envoked by visual scene motion: implication for alleviating simulator sickness.;

164

2001.
140 Merhi O, Faugloire E, Flanagan M, Stoffregen TA. Motion Sickness, Console Video Games,
and Head-Mounted Displays. Human Factors: The Journal of the Human Factors and
Ergonomics Society. 2007 October;49(5):920-934.
141 Golding JF. Motion Sickness Susceptibility Questionnaire Revised and Its Relationship to
Other Forms of Sickness. Brain Research Bulletin. 1998 November;47(8):507–516.
142 Rebenitsch L, Owen C. Cybersickness Statistical Measures and Display Effects. In
Preparation. 2015.
143 Rebenitsch L, Owen C, Coburn S. World and Object Designs for Virtual Environments. In:
Meaningful Play; 2014; East Lansing, Michigan.
144 Golding JF. Predicting individual differences in motion sickness susceptibility by
questionnaire. Personality and Individual Differences. 2006;41:237–248.
145 Stanney KM, Kennedy RS. Aftereffects From Virtual Environment Exposure: How Long do
They Last? Proceedings of the Human Factors and Ergonomics Society Annual Meeting.
1998 October;42(21):1476-1480.
146 Keshavarz B, Hecht K. Validating an Efficient Method to Quantify Motion Sickness. Human
Factors. 2011 August;53(4):415-426.

165