Content uploaded by Austin Erickson
Author content
All content in this area was uploaded by Austin Erickson on Mar 15, 2021
Content may be subject to copyright.
Effects of Dark Mode Graphics on Visual Acuity and Fatigue with
Virtual Reality Head-Mounted Displays
Austin Erickson*Kangsoo Kim†Gerd Bruder‡Gregory F. Welch§
University of Central Florida
ABSTRACT
Current virtual reality (VR) head-mounted displays (HMDs) are
characterized by a low angular resolution that makes it difficult to
make out details, leading to reduced legibility of text and increased
visual fatigue. Light-on-dark graphics modes, so-called “dark mode”
graphics, are becoming more and more popular over a wide range
of display technologies, and have been correlated with increased
visual comfort and acuity, specifically when working in low-light
environments, which suggests that they might provide significant
advantages for VR HMDs.
In this paper, we present a human-subject study investigating the
correlations between the color mode and the ambient lighting with
respect to visual acuity and fatigue on VR HMDs. We compare two
color schemes, characterized by light letters on a dark background
(dark mode), or dark letters on a light background (light mode), and
show that the dark background in dark mode provides a significant
advantage in terms of reduced visual fatigue and increased visual
acuity in dim virtual environments on current HMDs. Based on our
results, we discuss guidelines for user interfaces and applications.
Index Terms:
Computer graphics—Graphics systems
and interfaces—Virtual reality; Human-centered computing—
Interaction paradigms—Virtual reality
1 INTRODUCTION
Researchers in computer graphics and human-computer interaction
have for a long time investigated the effects of color and light in user
interfaces among a wide range of display technologies, and analyzed
their effects and limitations for different types of tasks. Computer
displays generally strive to present information with a high signal-
to-noise ratio, in particular when presenting text to readers, which
emphasizes the importance of strong luminance differences (instead
of chromatic differences) between the foreground and background.
For instance, this means that text is usually presented as dark letters
on a light background (light mode) or as light letters on a dark
background (dark mode) [6, 23, 30, 32]. In particular, dark mode
graphical user interfaces are gaining popularity in recent years in that
they are characterized by a reversal of the dominant color choices
in traditional user interfaces. Normal and inverted color choices
were investigated over a wide range of display technologies and
environments, and were linked to effects of legibility, aesthetics,
energy savings, semantic effects, and emotions [6,23, 30,32].
The effects of these color modes depend on the display tech-
nology [6], and can affect the legibility of text and/or the visual
fatigue caused by viewing details on a computer display. In particu-
lar, virtual reality (VR) head-mounted displays (HMDs) differ from
traditional screens in that they are characterized by a low angular
*e-mail: ericksona@knights.ucf.edu
†e-mail: kangsoo.kim@ucf.edu
‡e-mail: bruder@ucf.edu
§e-mail: welch@ucf.edu
resolution that makes it difficult and strenuous to read small text. An
extenuating factor is that VR HMDs as well as video see-through
augmented reality (AR) HMDs are largely based on a flat-panel dis-
play design, which leverages a 2D array of individual light-emitting
elements laid out on a panel. The slight space between each pixel
does not emit light, which thus appears dark. The resulting impres-
sion of a dark visual grid is denoted the screen door effect, which
is characterized by noticeable horizontal, vertical, and/or diagonal
lines [10,12,40]. Recent advances in consumer HMDs have reduced
the noticeability of this effect by increasing the resolution, which
reduced the gaps between pixels to below humanly perceivable
thresholds. However, while reduced, the effect persists throughout
current consumer VR HMDs due to the short distance between the
users eyes and the display within the HMD. As shown in Figure 1,
the screen door is most noticeable as the contrast between the light
emitted from the pixels and the dark grids in between. These factors
make VR-HMD displays unique to other display types, which means
that the best practices for the display of text may be different for
VR-HMDs than for traditional flat-panel displays.
It is our hypothesis that text presented on an HMD using light/dark
foreground colors on a dark/light background, can have a significant
impact on users’ visual acuity and fatigue. In this paper, we discuss
and investigate these color schemes as well as the amount of cen-
tral/peripheral light on an HMD on the example of an Oculus S. We
asked participants in a human-subject study to read text in VR with
different color modes and to complete visual acuity tests with dif-
ferent lighting conditions. We assessed participants’ visual fatigue,
acuity, and preferences.
In particular, we investigated the following research questions:
•RQ1
: Are there subjective or objective benefits of dark mode
or light mode color schemes with respect to visual acuity and
fatigue?
•RQ2
: Do users subjectively prefer dark mode or light mode
color schemes?
•RQ3
: Do the subjective preferences match the objective bene-
fits and drawbacks on users’ visual acuity and fatigue?
•RQ4
: Can increased central or peripheral lighting improve
visual acuity and fatigue in the presence of different color
schemes?
The remainder of this paper is structured as follows: Section 2
presents an overview of related work. Section 3 describes the human-
subject study. The results are presented in Section 4 and discussed
in Section 5. Section 6 concludes the paper and discusses future
research.
2 RE LATED WORK
In this section, we present an overview of related work focusing on
visual acuity and fatigue in computer displays, text legibility and
color modes, as well as the screen door effect in VR.
2.1 Visual Acuity and Fatigue with Computer Displays
As the types and uses of displays diversify and become ubiquitous,
people spend more and more time with different display devices, and
various health issues related to the increasing use of displays have
stood out consequently. For example, human physiology, cognitive
performance, and brain function can be influenced by the lighting
from different displays [7, 16, 36, 43]. Among many health concerns
with displays, various ocular symptoms are dominant due to the
direct visual stimuli on the eyes from the displays, such as eye-
strain, tired eyes, and sensitivity to bright lights and eye discomfort,
which are referred to as computer vision syndrome (CVS) [3, 39].
To identify the cause of such symptoms and reduce them, various
recommendations have been made with regard to color and light for
the graphical content and the display configurations. Campbell and
Durden suggested that individual users should be able to adjust the
brightness of the computer devices to select the luminance and con-
trast depending upon the user’s context, such as time in use and the
lighting condition in the environment [8]. Recently, automatic color
adjustment systems using dark/night mode (or inverted imaging)
are actively used to protect the user’s eyes and increase the comfort
level, e.g., f.lux1and Apple’s night mode configuration2.
As immersive VR HMDs become more popular with the growing
public interest in VR [45], the perceptual issues and visual fatigue
continue and even extend while using such HMDs with near-eye
displays embedded [5, 21,41]. There have been some studies about
the effects of different displays on visual fatigue while also inves-
tigating the performance on visual acuity, using 3D displays and
VR headsets [25, 28]. Kooi and Toet studied the effects of binocular
image imperfection on visual fatigue in stereo vision systems, and
found that nearly all binocular image asymmetries seriously reduced
visual comfort [25]. Lambooij et al. also conducted a user study
to identify visual discomfort associated with 3D stereoscopic dis-
plays compared to 2D displays, and presented that the participants
with moderate binocular status experienced more visual discomfort
and showed performance decrease in a reading task [28]. Similarly
in AR, which becomes more popular and practical for daily use
cases [22], there were a few studies that investigated visual fatigue
and acuity influenced by the real background patterns and focal
distance [15, 34]. Our focus in this paper is on the effects of color
mode on the visual fatigue and acuity in VR. Although there was
a study that investigated the dark mode effect in AR [23], there is
still a large gap in the literature about the effects of color modes on
visual acuity and fatigue with immersive VR HMDs. Recent night
mode updates for VR HMDs, such as SteamVR or Oculus
3
, also
emphasize the timely importance of color mode research in VR.
2.2 Text Legibility in Virtual Reality
In our daily lives, we encounter text to read and write for absorbing
information and communicating with others, e.g., which includes
reading this present manuscript. Computer displays usually strive
to present information with a high signal-to-noise ratio, in particu-
lar when presenting text to readers, which emphasizes the benefits
of strong luminance differences instead of chromatic differences
between the foreground and background. Reading on electronic
computer displays often results in different reading outcomes com-
pared to reading on physical paper [17, 18].
In the early age of electronic display technology when cathode-
ray tube (CRT) monitors were prevalent, text was typically displayed
in light-on-dark color scheme interfaces, i.e., light text on a dark
1
f.lux, a cross-platform software that adjusts a display’s color temperature
accordingly (https://justgetflux.com/)
2
Apple, “How to Use Dark Mode on your Mac” (
https://support.
apple.com/en-us/HT208976).
3
Quest, Go, and Gear VR build 8.0 release notes (2019-08-19) - Night
mode added (https://support.oculus.com/release-notes/)
Figure 1: Illustration of the light mode (top) and dark mode (bottom)
color schemes on virtual reality head-mounted displays. The dark
grid indicates the screen door effect in VR HMDs, which affects the
signal-to-noise ratio of the foreground light/dark text on a dark/light
background.
background, because the text on the monitors was displayed by the
electron beam hitting the phosphorous material for luminescence
that is normally dark in its natural state. However, as the dark-
on-light color scheme, i.e., dark text on a light background, was
introduced in WYSIWYG editing systems to simulate ink on paper
in the real world, it has been dominant in many computer user inter-
faces. Presenting dark text on a light background is usually referred
to as positive contrast, which goes back to the signal processing
theory, where the peak-to-peak contrast (or Michelson contrast [35])
measures the ratio between the spread and the sum of two lumi-
nances. This ratio is defined as
c=Lb−Lt
Lb+Lt
with text luminance
Lt
and background luminance
Lb
, which is negative if
Lb<Lt
. While
both positive and negative contrast conditions can provide the same
theoretical peak-to-peak contrast ratio, a large body of literature
focused on identifying benefits of one of them over the other for
different display technologies and use cases.
Multiple studies have found that positive contrast has benefits
when the goal is to read text on computer screens [1, 9,44]. Trying
to understand this effect, Taptagaporn and Saito observed that partic-
ipants developed a smaller pupil diameter when they used a positive
contrast display compared to a negative contrast display [42]. A
small pupil diameter is known to increase the quality of the retinal
image with greater depth of field and less spherical aberration, and
it is largely affected by the amount of light reaching the observer’s
eyes. Buchner et al. investigated the display luminance in positive
and negative contrast modes, showing that it is usually higher in
positive contrast modes, e.g., when dark text is presented on a light
background [6], which can be traced back to the ratio of screen space
filled by (dark) letters or the (light) background.
While text is important for conveying information to users in a
virtual world, e.g., when mimicking the real world, there is a dearth
of research on the effects of color mode on visual acuity and fatigue
in immersive VR HMDs, specifically with respect to text-based
annotations. In VR/AR, the virtual content with text can spatially
float around the user’s environment, and the legibility of the content
can vary dynamically depending on the distance and the (real/virtual)
environmental conditions between the user and the content. For
example, Lages et al. presented adaptive AR workspaces using
virtual interfaces with text, which adaptively change its position
and orientation while the user was walking [27]. In such VR/AR
applications and interfaces for text annotation and visualization [29,
33], the color modes by which text are presented to users are an
important research direction for making such applications more
usable and effective.
2.3 Screen Door Effect and Lighting
As mentioned in the introduction, the screen door effect is caused by
a dark grid in between light-emitting pixels on a flat-panel display.
This results in the user observing a visual phenomenon similar
to looking through a fine screen, such as a mosquito net or pool
enclosure, which may reduce visual acuity (see Figure 1). There
has been some effort in the past aimed at reducing the screen door
effect. Sitter et al. describe one such technique, which involves
adding a diffractive film onto the outer layer of the display [40].
This film causes a diffraction to occur for each pixel on the display
that spreads the light of the pixel outward into the surrounding black
matrix. Their technique reduces the screen door effect while keeping
comparable brightness and color uniformity.
As the near-eye displays in VR HMDs are so close to the
users’ eyes, intense virtual lighting can easily cause a smaller pupil
size [46], which in turn may cause a sharper view of the content
presented on the display. While increasing the amount of light emit-
ted by pixels on a VR HMD may increase visual acuity due to a
reduced pupil size, it may also exacerbate the screen door effect and
make the dark regions between the pixels more noticeable, which
may reduce visual acuity. However, an alternative method that can
increase the amount of light reaching the user’s eyes would be to add
light in the periphery of the display, as proposed by Jones et al. [20]
and Lubos et al. [31]. Using a ring of light-emitting diodes (LEDs)
in the periphery of the HMD, these approaches can reduce the user’s
pupil size without causing additional issues with the screen door
effect. Jones et al. [20] further showed that such peripheral light can
significantly improve distance estimation in VR. Given the technical
challenges in modifying HMDs with such peripheral LEDs, and the
fact that modern VR HMDs already provide a reasonably large field
of view, a purely software-based solution is to leverage the perimeter
of the display, i.e., the outer frame of pixels on the HMD, to induce
perimeter/peripheral light, which could mitigate the negative effects
in the perception of virtual content, such as the aforementioned
visual fatigue and acuity.
3 EXPERIMENT
In this section we present a user study in which we evaluate the
impacts of color mode, perimeter lighting, and virtual lighting on
users’ visual acuity, visual fatigue, and preferences.
3.1 Participants
After initial pilot tests, we estimated the effect size of the expected
strong effects, and based on a power analysis, we made the decision
to recruit
18
participants, which proved sufficient to show significant
effects in our experiment. We recruited a total of
15
male and
3
female participants (ages between
19
and
35
,
M
=
24.5
,
SD
=
4.8
).
Eligible for participation in the experiment were only healthy people
who did not have any cognitive or motor impairments. All of our
participants had normal or corrected-to-normal vision. Seven wore
glasses and four wore contact lenses during the experiment. None
of the participants reported known visual or vestibular disorders,
such as color or night blindness, dyschromatopsia, or a displacement
of balance. The participants were student or non-student members
of the local university community, who responded to open calls
for participation, and received monetary compensation for their
participation. All participants had used a VR HMD before.
3.2 Material
In this section, we describe the material used for our experiment.
3.2.1 Physical Setup
Figure 2 shows a photo of a participant in the study. Participants
were seated in an office chair and were instructed to wear an Oculus
Rift S VR HMD. The HMD was tracked in position and orientation
using a the built in inside-out tracking, where position and orien-
tation updates are handled internally by the HMD through the use
of cameras placed on the device. The HMD has a resolution of
1280
×
1440 pixels per eye for a total resolution of 2560
×
1440 and
a refresh rate of 80 Hz. The virtual environment was rendered in
Unity 2018.2.21f1 on a host PC tethered to the HMD (Intel Core
i7-8700k @ 3.70 GHz, 32Gb Ram, NVIDIA GTX 1070Ti graphics
card, Windows 10 Pro).
Figure 2: Annotated photo showing a participant in the study with the
HMD seated at the table.
Figure 3: Illustrations of visual stimuli used in the experiment. Top-
left: light mode in bright environment. Top-right: dark mode in bright
environment. Bottom-left: light mode in dim environment. Bottom-
Right: dark mode in dim environment.
3.2.2 Virtual Environment
The visual stimuli consisted of a virtual room in which the participant
was placed near a wall facing a floating panel that contained text
relevant to the conditions being displayed to them. Figure 3 shows
the virtual content that we used for the tasks in the study.
The floating panel was designed to match realistic lighting condi-
tions such that a diffuse white background would appear white in
bright virtual lighting, but would appear gray in dim virtual lighting.
It also meant that a diffuse black background would remain black
independent of the amount of virtual light. The virtual lighting in
the experiment could be adjusted to bright or dim by varying the
intensity of a virtual point light located above the participant’s head
in the Unity scene. Bright in this case means that white pixels on
the display were drawn at an RGB value of
(1,1,1)
, and dim means
that environment lighting was reduced by 90% so white pixels on
the display appeared at an RGB value of (0.1,0.1,0.1).
It is important to note that the RGB values described above only
describe the colors specified to the unity engine and are not indicative
of the actual amount of light that was displayed from the Oculus S
HMD. The apparent contrast between black and white pixels on a
display will vary across different HMDs and displays depending on
the contrast ratio of the device as well as parameters associated with
the display hardware.
Separate from the amount of ambient light in the virtual envi-
ronment, we also implemented a perimeter lighting mode in which
bright light originated from a ring of the out-most pixels at the
perimeter of the HMD screen. This lighting mode was inspired by
previous work by Jones et al. [20], who found that light reaching a
VR user’s eyes from the far periphery can affect (improve) spatial
perception. The ring was scaled to take up
348
pixels or
13.6
percent
of the total width of the display and
338
pixels or
23.4
percent of
the total height of the display. These values were chosen based on
pilot testing which suggested that lesser values were not noticeable
to some users.
3.2.3 Task Stimuli
We implemented a visual acuity test that incorporated tumbling Lan-
dolt C characters, which could be oriented normally or at 90 degrees
incremental rotations, so that the opening on the ‘C’ character could
face up, down, left, or right [11]. These characters were randomly
generated each time a chart was displayed, and by using this tum-
bling ‘C’ format, all acuity tests were of comparable difficulty to
one another. The Landolt C characters were chosen in favor of a tra-
ditional Snellen variety acuity chart in order to avoid the possibility
of having differing degrees of difficulty between same sized letters
between participants. While a Snellen variety chart would have
been possible, it would have required careful design to ensure that
the charts used in each condition were similar in difficulty to each
other to avoid potentially biasing a subset of conditions. The acuity
chart was positioned at a distance of
1.52
meters (5 feet) away from
the user, and at the eye level of the user. This distance was chosen
because the letter sizes on our custom acuity chart were modeled
after an acuity chart that was designed to be read specifically from
this distance.
We further implemented a reading task, which consisted of four
paragraphs from the Pearson Test of English Read Aloud Practice
Questions, which were presented in the Liberation Sans font. This
task was chosen to allow a standard time for the user to be exposed
to the condition lighting and potentially induce visual fatigue prior
to reading the acuity chart. These paragraphs were displayed to the
participant during each condition at the same depth as the acuity
chart (1.52 meters) and at a consistent field of view between all
conditions as a means of evaluating the amount of eye strain induced
and the readability of text in each condition.
3.3 Methods
The study used a
2×2×2
full-factorial within subjects design in
which each participant experienced all eight of the different condi-
tions, and the conditions were counterbalanced among participants
through the use of a Latin square. The evaluated independent vari-
ables were:
•Color Mode
:light mode graphics consisting of black text on
a white background or dark mode graphics consisting of white
text on a black background.
•Virtual Lighting
:bright or dim ambient lighting in the virtual
environment.
•Perimeter Lighting:enabled or disabled perimeter lighting.
3.3.1 Procedure
To begin the study, participants were led into the laboratory and
were asked to read over a consent form describing what would take
place during the experiment. After giving consent, the participants
were asked to complete two questionnaires: one which gathered
demographic information, and an Ocular Surface Disease Index
Figure 4: Illustrations depicting the VR text-reading task participants
had to perform during the experiment. Top-Left: Light mode in bright
environment. Top-Right: Dark mode in bright environment. Bottom-
Left: Light mode in dim environment. Bottom-Right: Dark mode in
dim environment.
(OSDI) survey that gathered information about the current level of
comfort of the participant’s eyes [37].
For each of the eight conditions, participants were then asked
to don the HMD and observe a virtual panel consisting of four
short paragraphs of text which they would read non-verbally to
themselves (see Figure 4). After one minute of observation, the
paragraphs would disappear and be replaced with a visual acuity
chart consisting of Landolt C characters rotated at 90 degree incre-
ments [11] (see Figure 3). The participants would then read through
the chart until they reached the bottom or the characters were too
difficult for them to distinguish. We measured the number of errors
participants made when reading the letters, the maximum row that
participants could read without errors, as well as the time it took
them to complete the task. Following the completion of the acuity
test, the participant would take off the HMD and complete two short
questionnaires: a Short User Experience Questionnaire (UEQ-S)
that gathered information on the usability of the graphics interface
under the testing conditions [38], and a Convergence Insufficiency
Symptom Survey (CISS) questionnaire that gathered information
on any eye strain noted by the participant during the condition [4].
Immediately after completing the questionnaires, participants were
instructed to re-don the HMD and continue onto the next condition
with no additional time to rest.
After completion of all eight conditions, participants were asked
to don the HMD one final time to measure their subjective preference
of the different conditions as well as which conditions they found to
be easiest to read.
Specifically, we asked them to indicate their preference of color
mode (dark mode or light mode) on four questions:
1. Preference: Which condition do you prefer?
2. Comfort: Which condition was more comfortable?
3. Easy to read: Which condition was easier to read?
4. Performance
: Which condition do you think you performed
better, e.g., fast and accurate reading?
We further asked them to indicate their preference of perimeter
lighting being enabled or disabled.
3.4 Hypotheses
Inspired by the body of literature on vision modes, most notably
recent work by Kim et al. [23], we defined the following hypotheses.
•H1a (Virtual Lighting Affects Visual Acuity):
Participants
will make fewer errors on their visual acuity test with bright
virtual lighting than dim virtual lighting.
•H1b (Virtual Lighting Affects Eye Strain):
Participants will
experience less eye strain in dim virtual lighting conditions
than they will in bright virtual lighting conditions.
•H2a (Color Mode Affects Visual Acuity):
Participants will
make fewer errors on their visual acuity test in the dark mode
condition than in the light mode condition.
•H2b (Color Mode Affects Eye Strain):
Participants will ex-
perience less eye strain when experiencing the dark mode than
the light mode.
•H3a (Perimeter Lighting Affects Visual Acuity):
Partici-
pants will make fewer errors on their visual acuity test while
experiencing the perimeter lighting than they will without
perimeter lighting.
•H3b (Perimeter Lighting Affects Eye Strain):
Participants
will experience more eye strain in conditions with perimeter
lighting than they will in conditions without it.
•H4 (Subjective Preference of Color Mode):
Users will pre-
fer the dark mode over the light mode.
4 RE SULTS
In this section we present the analysis and results of our experiment.
We used parametric statistical tests to analyze the responses in line
with the ongoing discussion in the field of psychology indicating
that parametric statistics can be a valid and informative method for
the analysis of combined experimental questionnaire scales with in-
dividual ordinal data points measured by questionnaires or coded be-
haviors [24,26]. We analyzed the responses with repeated-measures
ANOVAs and Tukey multiple comparisons with Bonferroni correc-
tion at the 5% significance level. We confirmed the normality with
Shapiro-Wilk tests at the 5% level and QQ plots. Degrees of freedom
were corrected using Greenhouse-Geisser estimates of sphericity
when Mauchly’s test indicated that the assumption of sphericity had
been violated. We had to remove two questionnaire data sets from
the analysis due to incomplete responses by our participants. We
only report the significant effects.
4.1 Visual Acuity
Figure 5 shows the results for the visual acuity tests.
In line with Hypothesis
H1a
, we found that users made fewer
errors and could complete more rows without errors on the visual
acuity charts if
virtual lighting
was bright instead of dim, while it
took participants longer to complete the tasks. We found a significant
main effect of virtual lighting on the number of errors made by
participants on the visual acuity chart,
F(1,17) = 72.33
,
p<0.001
,
η2
p=0.81
, indicating that bright lighting resulted in fewer errors
than dim lighting. We also found a significant main effect of virtual
lighting on the maximum row without errors on the visual acuity
chart,
F(1,17) = 32.12
,
p<0.001
,
η2
p=0.65
, which indicates that
more rows could be completed with bright lighting than dim lighting.
Further, we found a significant main effect of virtual lighting on
the time spent on the visual acuity chart,
F(1,17) = 4.54
,
p = 0.048
,
η2
p=0.21
, indicating that it took participants longer to complete the
charts under bright lighting than under dim lighting.
Contrary to Hypothesis
H3a
, our results did not show any signif-
icant effects of the
perimeter lighting
on the participants’ errors,
maximum rows, or time when completing the visual acuity charts.
As we are not seeing significant benefits of perimeter lighting, we are
focusing on the results for color modes without perimeter lighting
in the following. We further found a significant interaction effect
between virtual lighting and color mode on the number of errors,
F(1,17) = 13.99
,
p = 0.002
,
η2
p=0.45
, and on the maximum row
without errors,
F(1,17) = 9.43
,
p = 0.007
,
η2
p=0.36
, so we present
the corresponding significant effects in the following. We found no
significant effects on time.
In line with Hypothesis
H2a
, without perimeter lighting, we
found significant effects of the
color mode
on the results. However,
interestingly, the results show the opposite effect depending on the
virtual lighting:
For bright environments, we found a significant effect of the
color mode on the number of errors participants made when read-
ing the visual acuity chart,
F(1,17) = 9.15
,
p = 0.008
,
η2
p=0.35
,
indicating fewer errors for the light mode over the dark mode. We
also found a non-significant trend between the color mode and the
maximum row that a participant could reach without making any
errors,
F(1,17) = 3.28
,
p = 0.088
,
η2
p=0.16
, suggesting that more
rows may be completed with the light mode than the dark mode.
For dim environments, we found a significant effect of the color
mode on the number of errors participants made when reading the
visual acuity chart,
F(1,17) = 4.91
,
p = 0.041
,
η2
p=0.22
, indicating
fewer errors for the dark mode over the light mode. We also found
a significant effect of the color mode on the maximum row that a
participant could reach without making any errors,
F(1,17) = 5.28
,
p = 0.035
,
η2
p=0.24
, indicating that more rows could be completed
with the dark mode than the light mode.
4.2 Visual Fatigue
Figure 6 shows the visual fatigue results for the CISS questionnaire.
In line with Hypothesis
H2b
, we found a significant main effect of
color mode
on the visual fatigue scores,
F(1,15) = 8.10
,
p = 0.014
,
η2
p=0.34
, indicating lower visual fatigue for the dark mode com-
pared to the light mode. Interestingly, this result is independent of
the virtual lighting and applies to both bright and dim environments
(see results for visual acuity).
Contrary to Hypothesis
H3b
, our results did not show any signifi-
cant effects of the
perimeter lighting
on the visual fatigue scores.
As we are not seeing significant benefits of perimeter lighting, we are
focusing on the results without perimeter lighting in the following.
In line with Hypothesis
H1b
, without perimeter lighting, we
found a significant effect of
virtual lighting
on the visual fatigue
scores,
F(1,15) = 5.17
,
p = 0.038
,
η2
p=0.26
, indicating higher
visual fatigue for the bright environment compared to the dim envi-
ronment.
4.3 Usability
Figure 7 shows the usability results for the UEQ-S questionnaire,
in which users rated various aspects of the condition using a seven
point scale [38]. We found no significant effects of virtual lighting,
perimeter lighting, or color mode on the usability results.
4.4 Subjective Preferences
Figure 8 shows the participants’ preferences of the dark mode for
all conditions in the experiment.
We performed a two-tailed binomial test analysis on the sub-
jective preference data with a test value of
0.5
and a confidence
interval of
95%
, where users responded to questions about their
preference of color mode between either dark mode or light mode
under each of the study conditions. Users were specifically asked
which color mode they preferred, which was more comfortable,
and which was easier to read. We found a non-significant trend
in the number of participants who preferred the dark mode when
trying to read in a dim virtual environment with perimeter lighting
turned on (
p = 0.096
,
Proport ion =0.722
) and turned off (
p = 0.096
,
Maximum Row #
7
6
5
4
3
2
1
0
Bright
Lighting
Dim
Perimeter
Lighting
No Perimeter
Lighting
Perimeter
Lighting
No Perimeter
Light Mode
Dark Mode
(a)
# Total Errors
16
14
12
10
8
6
4
2
0
Bright
Lighting
Dim
Perimeter
Lighting
No Perimeter
Lighting
Perimeter
Lighting
No Perimeter
Light Mode
Dark Mode
(b)
Bright
Lighting
Dim
Perimeter
Lighting
No Perimeter
Lighting
Perimeter
Lighting
No Perimeter
Completion Time (in sec)
45
40
35
30
25
20
15
10
5
0
Light Mode
Dark Mode
(c)
Figure 5: Results for the visual acuity tests: (a) maximum row on the acuity chart that could be completed without errors (between 0 and 9; higher
is better), (b) total number of errors on acuity chart (lower is better), and (c) completion time for the acuity chart (lower is better).
Bright
Lighting
Dim
Perimeter
Lighting
No Perimeter
Lighting
Perimeter
Lighting
No Perimeter
Light Mode
Dark Mode
Visual Fatigue Score
80
70
60
50
40
30
20
10
0
91
4
8
95
Figure 6: Results for the visual fatigue questionnaire (CISS): overall
fatigue scores (lower is better).
Overall Usability Score
3
2
1
0
-1
-2
-3
Bright
Lighting
Dim
Perimeter
Lighting
No Perimeter
Lighting
Perimeter
Lighting
No Perimeter
Light Mode
Dark Mode
Figure 7: Results for the usability questionnaire (UEQ-S): overall
usability scores (higher is better).
Proport ion =0.722
). We also found a non-significant trend in the
number of participants who preferred the dark mode as more visu-
ally comfortable in dim lighting conditions with perimeter lighting
turned off (
p = 0.096
,
Proport ion =0.722
). We also found that a sig-
nificant number of participants preferred having perimeter lighting
turned off as opposed to turned on (
p = 0.008
,
Proport ion =0.833
).
5 DISCUSSION
In this section, we discuss the main findings and their implications
for VR HMDs.
Preference of Dark Mode (%)
90
80
70
60
50
40
30
20
10
0
Bright
Lighting
Dim
Perimeter
Lighting
No Perimeter
Lighting
Perimeter
Lighting
No Perimeter
Preference Comfort Easy to read Performance
100
Figure 8: Subjective results in percent of participants who preferred
the dark mode in the different experimental conditions.
5.1 Dark Mode Improves Visual Acuity Only in Dim
Lighting Conditions
Our results shown in Section 4.1 indicate that the dark mode im-
proves the visual acuity of the user in dim lighting conditions on
VR HMDs, effectively making it easier for users to identify Lan-
dolt C characters or make out small visual details. Conversely, the
light mode improves the visual acuity of the user in bright virtual
environments.
This result stands in partial contrast to the results of prior work
by Kim et al., who investigated dark mode user interfaces in AR
optical see-through HMDs and found that the dark mode yielded
better visual acuity regardless of lighting conditions [23]. It stands
to reason that the difference in the display’s light model, in particular
the additive light model [14] of current optical see-through displays
as well as the screen door effect that is prevalent in current immersive
HMDs have a strong effect on the results.
In VR HMDs, the screen door effect is very prominent when
pixels on the display are illuminated with bright light, and is more
obscured from view when pixels are darker (see Figure 1). However,
even in the presence of an increased screen door effect in bright
virtual environments, we found a significant main effect that the
user’s visual acuity under bright lighting is significantly higher than
under dim lighting. This result matches previous work, which in-
dicates that increasing the amount of light reaching the user’s eyes
will reduce their pupil diameter, which in turn is known to increase
the quality of the retinal image with greater depth of field and less
spherical aberration [42]. It is interesting to see that the positive
effects of the increased light out-weighted the negative effects of the
increased screen door effect in VR HMDs.
As shown in Figure 1, with the dark mode, the screen door effect is
primarily hidden from view in the dark background but does appear
directly over the foreground letters. In contrast, in the light mode,
the screen door effect is clearly noticeable in the light background
but is more obscured in the foreground letters. One’s first intuition
may suggest that having the screen door effect over the letters and
not the background would make them more difficult to see, but our
results suggest that this only occurs in bright lighting conditions and
that the opposite occurs in dim lighting conditions.
It is further interesting to note that the aforementioned study
by Kim et al. incorporated a lighting-independent text mode (as
used in AR heads-up displays) for the visual acuity charts that were
displayed to the users, meaning that the light/dark RGB color values
on the chart were constant and were not affected by virtual lighting
in the AR environment [23]. In contrast, our study took place in VR
as opposed to AR, where the RGB color values of text in the virtual
environment were affected by changes in the amount of virtual light,
denoted as a lighting-dependent text mode. Because of this, when the
virtual lighting is bright, white text appears as RGB value
(1,1,1)
and black text as
(0,0,0)
. However, as the virtual lighting dims, the
white text darkens to a value between
(1,1,1)
and
(0,0,0)
, while
black text remains black
(0,0,0)
and unaffected by the amount of
virtual light. A decrease in virtual light thus reduces the visual
contrast between the light colors and dark colors and the signal-to-
noise ratio between the foreground text and its background, which
is known to reduce the visual acuity [19].
Our results indicate that it is advantageous to use the light mode
under bright virtual lighting, but when the contrast between letters
and their background is reduced due to dim virtual lighting, then
it is advantageous to switch to the dark mode. We believe that this
is partially due to a color bleeding effect that occurs when a light
colored letter is presented on a dark background, where the light
from the letter partially illuminates neighboring background pixels
and results in a letter that appears slightly larger [13]. It stands
to reason that the magnitude of this effect is affected by virtual
lighting paired with the nature of the letter identification task. In our
study, participants were asked to identify Landolt C characters on
the visual acuity chart, and if the magnitude of the color bleeding
effect was too significant (in the case of bright virtual lighting) then
it is possible that while the letters did appear slightly larger, the
opening on the ‘C’ is reduced to appear more as an ‘O,’ and thus the
direction of the opening is more difficult to distinguish. In the case
of dim lighting conditions, the characters still appear slightly larger,
but the magnitude of the color bleeding effect is not as strong as in
the bright lighting condition, resulting in an opening on the ‘C’ that
is easier to distinguish than for the light mode.
If this color bleeding effect is responsible for the results obtained
here, then it is possible that different results may be obtained from
a similar future study where the pixel density of the VR HMD is
increased. This increased pixel density may result in less of a color
bleeding effect around the perimeter of the letters, which means that
letters will appear slightly smaller on the high-density display and
thus be more difficult to read. However, there is a trade-off to this
reduced color bleeding effect as the openings in letters will be easier
to identify than when this effect is more apparent, which should
make letters with similar features such as ‘C,’ ‘E,’ and ‘O’ easier to
distinguish from one another.
5.2 Dark Mode Decreases Visual Fatigue
As shown in Section 4.2, our results show that the dark mode re-
sulted in significantly lower visual fatigue (CISS) scores than the
light mode, which suggests that the dark mode causes less eye strain
than the light mode. This result was also observed by Kim et al. for
AR optical see-through HMDs [23]. Further, in line with related
work in the field, we also found that increasing the amount of (vir-
tual) lighting caused more eye fatigue than our tested dim lighting
condition [2]. For the least amount of visual fatigue in VR HMDs,
our guideline is to dim the amount of virtual lighting and make use
of the dark mode when presenting text or other visual details.
5.3 Preference of Dark Mode over Light Mode
As shown in Section 4.4, the majority of participants responded with
a preference of the dark mode over the light mode, although this was
only a non-significant statistical trend and further research would be
required to come to a more general conclusion. Both color modes
offer benefits to the users’ visual acuity depending on the virtual
lighting of the scene. A slight shift in preference for the dark mode
might stem from perceived benefits due to reduced visual fatigue. It
is possible that the preferences would have become clearer in favor
of the dark mode after a longer VR exposure.
5.4 Perimeter Lighting Showed no Significant Effects
As shown in Section 4.4, our results indicate that the majority of
participants preferred the perimeter lighting to be turned off. We
also found no significant effects of perimeter lighting on visual
acuity, visual fatigue, or usability. We were surprised to not see
clear benefits of the perimeter lighting on the results as the relevant
literature suggested that a decrease in pupil size due to added light
should improve the retinal image due to greater depth of field and
reduced spherical aberration [42]. In theory, it should not matter
whether the light that is affecting the user’s pupil size originates in
the center or the periphery/perimeter of the display.
Some of our participants commented on the perimeter lighting,
e.g., stating that turning the perimeter light on felt like the rest of
the virtual environment was getting darker. Another mentioned that
they felt as though a dark gradient was placed over the center of the
screen when the perimeter lighting was on.
For future work in this direction, we suggest looking into far-
periphery lighting (instead of perimeter lighting) as used by Jones
et al. [20] or Lubos et al. [31], who added an LED strip around the
screen in the periphery of a VR HMD. We expect that an increased
amount of peripheral light might result in benefits for visual acuity
in VR, although we also see potential drawbacks due to increased
visual fatigue.
6 CONCLUSION AND FUTURE WORK
In this paper, we presented a human-subject study in which we in-
vestigated the effects of the color mode (dark mode or light mode),
perimeter lighting, and virtual lighting on visual acuity, visual fa-
tigue, usability, and preferences when reading text and completing
a visual acuity test in VR. Among other results, we showed unique
benefits of the dark mode under dim lighting conditions on visual
acuity and fatigue, of the light mode under bright lighting conditions
on visual acuity, as well as increased visual acuity and fatigue un-
der bright virtual lighting. Our results may serve as guidelines for
practitioners in the design and implementation of VR user interfaces
that require high legibility and/or reduced visual fatigue.
Its important to note that the results and conclusions obtained
from the described study above cannot be generalized to other head-
mounted displays due to differences in parameters associated with
the built-in display, such as contrast ratio and luminance. Because
our group did not have the equipment necessary to perform such
measurements at this time, future work should investigate other
head-mounted displays and take measurements of such parameters
so that their impact on user visual acuity and eye strain can be further
evaluated.
As the resolution of VR displays increases, future work should
investigate how the resolution and intentional color bleeding effects
as in the PlayStationVR may impact visual acuity and fatigue in
the presence of different color modes and lighting. Future studies
may also find further advantages and disadvantages in the use of
perimeter lighting to provide benefits to the user such as increased
visual acuity.
ACK NOWLEDGM ENT S
This material includes work supported in part by the Office of
Naval Research under Award Number N00014-17-1-2927 (Dr. Peter
Squire, Code 34) and the AdventHealth Endowed Chair in Health-
care Simulation (Prof. Welch). Any opinions, findings, and conclu-
sions or recommendations expressed in this material are those of the
author(s) and do not necessarily reflect the views of the supporting
institutions.
REFERENCES
[1]
D. Bauer and C. R. Cavonius. Ergonomic Aspects of Visual Display
Terminals, chap. Improving the legibility of visual display units through
contrast reversal. London: Taylor & Francis, 1990.
[2]
S. Benedetto, A. Carbone, V. Drai-Zerbib, M. Pedrotti, and T. Baccino.
Effects of luminance and illuminance on visual fatigue and arousal
during digital reading. Computers in Human Behavior, 41:112–119,
2014.
[3]
C. Blehm, S. Vishnu, A. Khattak, S. Mitra, and R. W. Yee. Computer
Vision Syndrome: A Review. Survey of Ophthalmology, 50(3):253–
262, 2005.
[4]
E. J. Borsting, M. W. Rouse, G. L. Mitchell, M. Scheiman, S. A.
Cotter, J. Cooper, M. T. Kulp, R. London, and C. Group. Validity and
Reliability of the Revised Convergence Insufficiency Symptom Survey
in Children. Optometry & Vision Science, 80(12):832–838, 2003.
[5]
G. Bruder, A. Pusch, and F. Steinicke. Analyzing effects of geomet-
ric rendering parameters on size and distance estimation in on-axis
stereographics. In Proceedings of the ACM Symposium on Applied
Perception, pp. 111–118, 2012.
[6]
A. Buchner, S. Mayr, and M. Brandt. The advantage of positive text-
background polarity is due to high display luminance. Ergonomics,
52(7):882–886, 2009.
[7]
C. Cajochen, S. Frey, D. Anders, J. Sp
¨
ati, M. Bues, A. Pross, R. Mager,
A. Wirz-Justice, and O. Stefani. Evening exposure to a light-emitting
diodes (LED)-backlit computer screen affects circadian physiology and
cognitive performance. Journal of Applied Physiology, 110(5):1432–
1438, 2011.
[8]
F. W. Campbell and K. Durden. The Visual Display Terminal Issue: A
Consideration of Its Physiological, Psychological and Clinical Back-
ground. Ophthalmic and Physiological Optics, 3(2):175–192, 1983.
[9]
A. H. S. Chan and P. S. K. Lee. Effect of display factors on Chinese
reading times, comprehension scores and preferences. Behaviour and
Information Technology, 24:81–91, 2005.
[10]
J.-m. Cho, Y.-d. Kim, S. H. Jung, H. Shin, and T. Kim. Screen Door
Effect Mitigation and Its Quantitative Evaluation in VR Display. In
SID Symposium Digest of Technical Papers, vol. 48-1, pp. 1154–1156.
Wiley Online Library, 2017.
[11]
S. Demirel, R. S. Anderson, S. C. Dakin, and L. N. Thibos. Detection
and resolution of vanishing optotype letters in central and peripheral
vision. Vision Research, 59:9–16, 2012.
[12]
P. R. Desai, P. N. Desai, K. D. Ajmera, and K. Mehta. A Review Paper
on Oculus Rift—A Virtual Reality Headset. arXiv, cs.HC(1408.1173),
2014.
[13]
B. V. Funt, M. S. Drew, and J. Ho. Color Constancy from Mutual
Reflection. International Journal of Computer Vision, 6(1):5–24, 1991.
[14]
J. Gabbard, J. Swan, J. Zedlitz, and W. W. Winchester. More than meets
the eye: An engineering study to empirically examine the blending of
real and virtual color spaces. In Proceeding of IEEE Virtual Reality
(VR), pp. 79–86, 2010.
[15]
J. L. Gabbard, D. G. Mehra, and J. E. Swan. Effects of ar dis-
play context switching and focal distance switching on human perfor-
mance. IEEE Transactions on Visualization and Computer Graphics,
25(6):2228–2241, 2019.
[16]
S. Higuchi, Y. Motohashi, Y. Liu, M. Ahara, and Y. Kaneko. Effects of
VDT tasks with a bright display at night on melatonin, core temperature,
heart rate, and sleepiness. Journal of Applied Physiology, 94(5):1773–
1776, 2015.
[17]
A. Holzinger, M. Baernthaler, W. Pammer, H. Katz, V. Bjelic-Radisic,
and M. Ziefle. Investigating paper vs. screen in real-life hospital
workflows: Performance contradicts perceived superiority of paper in
the user experience. International Journal of Human Computer Studies,
69(9):563–570, 2011.
[18]
J. Hou, J. Rashid, and K. M. Lee. Cognitive map or medium materi-
ality? Reading on paper and screen. Computers in Human Behavior,
67:84–94, 2017.
[19]
C. A. Johnson and E. J. Casson. Effects of luminance, contrast, and
blur on visual acuity. Optometry and vision science, 72:864–869, 1995.
[20]
J. A. Jones, J. E. Swan, II, G. Singh, and S. R. Ellis. Peripheral vi-
sual information and its effect on distance judgments in virtual and
augmented environments. In Proceedings of the ACM SIGGRAPH Sym-
posium on Applied Perception in Graphics and Visualization (APGV),
pp. 29–36, 2011.
[21]
F. Kellner, B. Bolte, G. Bruder, U. Rautenberg, F. Steinicke, M. Lappe,
and R. Koch. Geometric calibration of head-mounted displays and its
effects on distance estimation. IEEE Transactions on Visualization and
Computer Graphics, 18:589–596, 2012.
[22]
K. Kim, M. Billinghurst, G. Bruder, H. Duh, and G. F. Welch. Revis-
iting Trends in Augmented Reality Research: A Review of the 2nd
Decade of ISMAR (2008–2017). IEEE Transactions on Visualization
and Computer Graphics (TVCG), 24(11):2947–2962, 2018.
[23]
K. Kim, A. Erickson, A. Lambert, G. Bruder, and G. Welch. Effects
of dark mode on visual fatigue and acuity in optical see-through head-
mounted displays. In Proceedings of the ACM Symposium on Spatial
User Interaction (SUI), pp. 1–9, 2019.
[24]
T. R. Knapp. Treating ordinal scales as interval scales: an attempt to
resolve the controversy. Nursing research, 39(2):121–123, 1990.
[25]
F. L. Kooi and A. Toet. Visual comfort of binocular and 3D displays.
Displays, 25(2–3):99–108, 2004.
[26]
W. M. Kuzon Jr, M. G. Urbanchek, and S. McCabe. The seven deadly
sins of statistical analysis. Annals of plastic surgery, 37(3):265–272,
1996.
[27]
W. S. Lages and D. A. Bowman. Walking with adaptive augmented re-
ality workspaces. In Proceedings of the ACM International Conference
on Intelligent User Interfaces, pp. 356–366, 2019.
[28]
M. Lambooij, M. Fortuin, W. Ijsselsteijn, B. Evans, and I. Heynderickx.
Measuring visual fatigue and visual discomfort associated with 3-D
displays. Journal of the Society for Information Display, 18(11):931–
943, 2010.
[29]
J. Li, B. Bach, R. Sicat, J. Choi, M. Cordeil, H. Pfister, and W.-K. Jeong.
DXR: A Toolkit for Building Immersive Data Visualizations. IEEE
Transactions on Visualization and Computer Graphics, 25(1):715–725,
2018.
[30]
D. L
¨
offler, L. Giron, and J. Hurtienne. Night Mode, Dark Thoughts:
Background Color Influences the Perceived Sentiment of Chat Mes-
sages. In Proceeding of INTERACT, pp. 184–201, 2017.
[31]
P. Lubos, G. Bruder, O. Ariza, and F. Steinicke. Ambiculus: LED-based
Low-resolution Peripheral Display Extension for Immersive Head-
mounted Displays. In Proceedings of the Virtual Reality International
Conference (VRIC), pp. 13:1–13:4, 2016.
[32]
L. W. MacDonald. Using color effectively in computer graphics. IEEE
Computer Graphics and Applications, 19(4):20–35, 1999.
[33]
C. Menk and R. Koch. Interactive visualization technique for truthful
color reproduction in spatial augmented reality applications. Proceed-
ings of the IEEE International Symposium on Mixed and Augmented
Reality (ISMAR), pp. 157–164, 2011.
[34]
C. Merenda, M. Smith, J. Gabbard, G. Burnett, and D. Large. Effects of
real-world backgrounds on user interface color naming and matching
in automotive ar huds. In IEEE VR 2016 Workshop on Perceptual and
Cognitive Issues in AR (PERCAR), pp. 1–6, March 2016.
[35] A. A. Michelson. Studies in optics. University Press, Chicago, 1927.
[36]
K. Richards, N. Mahalanobis, K. Kim, R. Schubert, M. Lee, S. Daher,
N. Norouzi, J. Hochreiter, G. Bruder, and G. Welch. Analysis of
peripheral vision and vibrotactile feedback during proximal search
tasks with dynamic virtual entities in augmented reality. In ACM
Symposium on Spatial User Interaction, pp. 1–9, 2019.
[37]
R. M. Schiffman, M. D. Christianson, G. Jacobsen, J. D. Hirsch, and
B. L. Reis. Reliability and validity of the ocular surface disease index.
Archives of Ophthalmology, 118(5):615–621, 2000.
[38]
M. Schrepp, A. Hinderks, and J. Thomaschewski. Design and Evalua-
tion of a Short Version of the User Experience Questionnaire (UEQ-S).
International Journal of Interactive Multimedia and Artificial Intelli-
gence, 4(6):103, 2017.
[39]
A. L. Sheppard and J. S. Wolffsohn. Digital eye strain: prevalence, mea-
surement and amelioration. BMJ Open Ophthalmology, 3(1):e000146,
2018. doi: 10. 1136/bmjophth-2018-000146
[40]
B. Sitter, J. Yang, J. Thielen, N. Naismith, and J. Lonergan. Screen
Door Effect Reduction with Diffractive Film for Virtual Reality and
Augmented Reality Displays. In SID Symposium Digest of Technical
Papers, vol. 48-1, pp. 1150–1153. Wiley Online Library, 2017.
[41]
F. Steinicke, G. Bruder, S. Kuhl, P. Willemsen, M. Lappe, and K. Hin-
richs. Natural perspective projections for head-mounted displays. IEEE
Transactions on Visualization and Computer Graphics, 17:888–899,
2010.
[42]
S. Taptagaporn and S. Saito. How display polarity and lighting condi-
tions affect the pupil size of VDT operators. Ergonomics, 33:201–208,
1990.
[43]
G. Vandewalle, P. Maquet, and D. J. Dijk. Light as a modulator of
cognitive brain function. Trends in Cognitive Sciences, 13(10):429–438,
2009.
[44]
A.-H. Wang, J.-J. Fang, and C.-H. Chen. Effects of VDT leading-
display design on visual performance of users in handling static and
dynamic display information dual-tasks. International Journal of
Industrial Ergonomics, 32:93–104, 2003.
[45]
G. Welch, G. Bruder, P. Squire, and R. Schubert. Anticipating
Widespread Augmented Reality: Insights from the 2018 AR Visioning
Workshop. Technical report, University of Central Florida and Office
of Naval Research, August 2019.
[46]
B. Winn, D. Whitaker, D. B. Elliott, and N. J. Phillips. Factors affect-
ing light-adapted pupil size in normal human subjects. Investigative
Ophthalmology and Visual Science, 35(3):1132–1137, 1994.
