Professional Documents
Culture Documents
Full Text 01
Full Text 01
KRISTIN MICKOLS
MAX WIPPICH
Abstract
Online education has become more prominent during recent years and it is
therefore of interest to study and improve. Colour in presentations have shown
to increase the attention and retention among students and colour coding can
use colour to highlight important key words and topics in online presentations.
As part of this study experiments were conducted with 16 computer science
students. Some were shown colour coded presentation images and some were
shown the same images without colour and were then asked to answer ques-
tions about the content. An eye tracker was also used during the experiment to
gather quantitative data that was analysed with respect to eye-movements and
gaze fixations. The main conclusions drawn from this study are that colour
coding is positive as long as it is used sparsely and that too many different
colours may affect students negatively, which is in line with previous research.
4
Sammanfattning
Digital undervisning har under de senaste åren blivit allt mer framstående och
det är därför av intresse att undersöka möjligheter att förbättra denna. Använ-
dandet av färg i föreläsningar har bevisat en ökning av uppmärksamhet och bi-
behållande av information bland studenter. Färgkodning är ett sätt att använda
färg för att markera viktiga nyckelbegrepp och ämnen i digitala föreläsningar.
Som en del av studien utfördes experiment med 16 datateknikstudenter. Vissa
av dessa presenterades föreläsningsslides i svartvitt medan andra presentera-
des slides med färgkodning och fick sedan besvara frågor kring materialet.
För att tillföra kvantitativ data användes även en eye tracker under experimen-
tet vars data analyserades med hänsyn till ögonrörelse samt fixationspunkter.
Huvudslutsatserna är att färgkodning är gynnsamt såvida det används sparsamt
och att för många olika färger kan påverka studenter negativt, vilket överens-
stämmer med tidigare forskning.
Contents
1 Introduction 1
1.1 Research Question . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2 Background 3
2.1 Online Education, Attention and Retention . . . . . . . . . . . 3
2.2 Colour Coding and Readability of digital content . . . . . . . 4
2.3 Eye tracking . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.4 Previous research . . . . . . . . . . . . . . . . . . . . . . . . 6
3 Method 8
3.1 Experiment Overview . . . . . . . . . . . . . . . . . . . . . . 8
3.2 Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
3.3 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.4 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
4 Results 14
4.1 Survey Results . . . . . . . . . . . . . . . . . . . . . . . . . 14
4.2 Heat Maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
4.3 Eye Tracker Data . . . . . . . . . . . . . . . . . . . . . . . . 23
5 Discussion 24
5.1 Discussion of Results . . . . . . . . . . . . . . . . . . . . . . 24
5.2 Implications for theory and practice . . . . . . . . . . . . . . 27
5.3 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
5.4 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
6 Conclusions 29
Bibliography 30
5
6 CONTENTS
A Presentations 34
A.1 Group A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
A.2 Group B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
C Heat Maps 87
C.1 Heat maps, coloured images . . . . . . . . . . . . . . . . . . 87
C.2 Heat maps, black and white images . . . . . . . . . . . . . . . 96
Introduction
Online education where students and teachers meet through various forms of
communication over the internet has become progressively more prominent
in schools and universities. The rise of online education may be a result of
different factors, such as the development of digitalisation and accessibility to
these services. As of today, it is common for universities to use online learning
management platforms where the students can hand in assignments, as well as
see their schedule and be alerted of assignments and deadlines [1]. Online
services such as YouTube videos with educational purpose and websites that
offer interactive education have been more common [2], where students can
learn about new topics from home or further their understanding of concepts
discussed in class.
During 2020, the covid-19 pandemic forced a majority of schools [3] to
find online alternatives to the education, which have been done in primarily
two ways – live online lectures and pre-recorded lectures. It is common for
lecturers to record their live online lectures and make the recordings available
for students to watch in retrospect [4]. This has made it possible for students
to further study concepts and walk through a lecture at their own pace [5].
To ensure that the quality of online education is the same as on-site educa-
tion, some adaptations have to be done. Research has proven online lectures to
have lower attention and retention among students compared to traditional lec-
tures [6]. It is therefore important to study how the content of online education
can be improved to increase students’ ability to retain information. According
to previous research [7], a possible improvement to the layout of lectures could
be colour coding, which enables the lecturers to highlight certain key points
within the lecture material, with the expectations to improve students’ reten-
tion. This is what this study aims to investigate, how such changes, especially
1
2 CHAPTER 1. INTRODUCTION
1.2 Scope
This research investigates whether colour coding has an impact on students’
retention during online lectures where lecture slides are utilised. The research
is focused solely on lectures within the subject programming and computer
science and does not investigate results in other subjects. Furthermore, there
will be no analysis on what colours have the most impact on retention, although
the colours picked are based on previous research.
Chapter 2
Background
In this chapter, background regarding online education and its effect on reten-
tion compared to physical education will be presented in section 2.1, Online
Education, Attention and Retention. The next two sections 2.2, Colour Coding
and Readability of Digital Content & 2.3 Eye Tracking discuss the effects and
usages of colour coding and analysis of eye tracking data within educational
studies. Finally, section 2.4, Previous Research will present some previous
research within this field of study.
3
4 CHAPTER 2. BACKGROUND
are likely to drop their courses due to low retention rates, procrastination and
lack of motivation [11]. According to Onah et al. [9], massive open online
courses have a completion rate as low as 13%.
During physical lectures, lecturers often have the possibility to be more
interactive with their learning material. They have the possibility to highlight
certain topics or words of importance by either drawing on the chalkboard or
have more emphasis while talking about the topic. During online education
this interaction is lost if lecturers share their educational content through video
or online meetings. Additionally lecturers have less interactivity with their
students as they often cannot directly see how students react or what they are
doing during the lecture.
Previous research shows that it is easier to lose attention while attending
an online lecture, as opposed to a physical lecture as the environment in which
the student studies often is more prone to introduce distractions [6]. Other
research also shows that student engagement is lower during online classes
[12]. In these studies a traditional lecture format, with a lecturer presenting
with notes, was used. This suggests that teachers need to format online classes
with special care, in order to keep students’ focus on the material.
Although the development and usage of online education was already on-
going, online education has become prominent due to the covid-19 pandemic
drastically changing the circumstances revolving education. This forced rapid
adjustments and adaptations regarding education that had to move online. [13]
In this study this will be limited to the actual font being of a different colour
and omit the case where the background of a word is coloured. Colour coding
in this study will refer to the use of highlighting two or more words of similar
meaning to emphasise their connection and help the reader find the other word
when reading the first.
When creating digital content a few guidelines, supported by a previous
study conducted by Deubel et al. [16] in 2003, can be followed to increase
the readability of the material, while avoiding cognitive overload. Deubel
shows that contrast between foreground text and the background should be
high and can be even higher for highlighted text. They also show that back-
ground colours should be neutral, such as pastel or grey colours, to improve
readability. Additionally some colour combinations are advised against, such
as blue & orange, red & green and violet & yellow and highlights also should
not be used too often, as not to confuse the reader. Colour coding should also
be used with caution since colour coding may cause the reader to associate a
certain concept with one colour, which may cause confusion if this colour is
encountered again when used for a different concept [17].
easily could distinguish relevant topics from irrelevant. Molina et al. [26]
also conducted a study, researching how the design of multimedia materials
can influence in the learning efficiency within primary education. In these two
studies, eye tracking was used as objective data gathering technique.
Lai et al. [27] conducted a large study on how eye-tracking technology
has been applied to studies of learning, analysing 81 papers within the field,
concluding that eye tracking is valuable for educational researches to connect
cognitive processes and learning outcomes.
Chapter 3
Method
In this chapter the method of the study will be described. The first section
3.1 Experiment Overview will give a general overview of the experiments that
were conducted and how they were conducted, as well as giving justification to
these methods. Section 3.2 Process, describes the process of the experiment
in greater detail and expands on certain aspects of the experiment. Section 3.3
Materials, describes the materials, hardware and software used to conduct the
experiment as well as analysing the results. Finally section 3.4 Limitations,
will focus on predicting problems and errors that might affect the result.
8
CHAPTER 3. METHOD 9
them within the scope of the study. Topics discussed in the presentation were
data structures and functions in programming. Both groups were shown some
sections of the presentation with colour coding and some without, so that one
of the groups was shown the section with colour coding and the other was
shown the same section without colour coding, and vice versa. This ensures
that the data collected shows the difference in retention for the two types of
presentation, independent of the content of the material, similarly to the two
studies conducted by Ozcelik et al. [7] and Olurniola & Tayo [23]. Addition-
ally, since the test group is fairly small in comparison to the 52 students in the
study conducted by Ozcelik et al. [7], showing all participants presentation
images with and without colour coding may reduce the impact on the result
of one test group performing better on average, due to previous knowledge.
The presentation that participants were shown had four sections, the first two
sections describing two different, existing data structures with illustrations of
these and the description of their uses. The last two sections contained two
collections of functions with similar functionality, but nonsensical names, the
first such collection also had an illustration to exemplify their functionality.
The first type is motivated by the fact that they are similar to what an actual
lecture could contain, which causes them to be more realistic. However, as
they build upon previous knowledge and some answers to questions can be
deduced by intuitive logic they have the weakness of providing less exact re-
sults, as the participants may have a different understanding of the concepts
previous to the study. The other type overcomes these weaknesses, by not
building on previous concepts nor being logical, preventing previous experi-
ence to affect the outcome. Yet, as a consequence, these sections are slightly
less realistic.
Colours in these presentations were used according to the results presented
by Deubel et al. in 2003 [16], most importantly with a high contrast to the
background colour, to ensure that the effect of colour coding would not be
impacted by poor choice of colour. After watching each presentation, the par-
ticipants filled out a form with questions about the topics discussed in that
presentation. These questions and answers were later corrected and used as a
quantitative indication of the participants’ retention of the material. In com-
parison to the study conducted by Ozcelik et al., where participants were asked
if they remembered a certain concept [7], this study required full answers that
were corrected. This was to ensure that participants would not incorrectly
think that they remembered a concept and therefore would grant more correct
results. This is more in line with the study conducted by Olurniola & Tayo,
where participants were asked to recall a list of words [23]. After the main
10 CHAPTER 3. METHOD
section of the study the participants were asked a few more questions about
the presentations and questions, such as which sections they found more diffi-
cult and if the colour in the presentations helped them understand the content.
The answers to these questions were compiled to see if participants found par-
ticular sections harder or easier depending on if they were colour coded or not
and if participants’ experienced colour coded parts as easier. This was to see
if the participants, regardless of how well they performed in retaining infor-
mation experienced that the colour was helpful for them in understanding the
material. Even if the test does not give any conclusive evidence that students’
retention increases due to colour coding, the experience that colour helps may
still be of importance.
During the test the participants were equipped with an eye tracker that col-
lected data on the movement and position of participants’ gaze as was done
in the study conducted by Ozcelik et al. [7]. A script continuously collected
the position of the participants’ gaze during each of the presentation images.
In addition to the answers to questions given, this data could be analysed to
show differences in how participants viewed the material, depending on if they
were shown it with colour coding or without. This data could later be used to
generate heat maps and to analyse the data. Heat maps are used in this study
as they are a simple way of visualising the large amount of data an eye tracker
produces. In addition to heat maps the average speed of participants’ gaze is
analysed. Ozcelik et al. chose to analyse when participants gaze landed on
certain areas or words [7] using a scan path. However, due software limitation
this research instead focuses on generating heat maps and analysing the data
using other techniques.
3.2 Process
At the start of each experiment participants were told in a general manner
what the study was about and what were expected from them. They were told
to watch the presentation as if it was a regular lecture and that they would be
asked to answer questions on each part of the lecture. However, participants
were not told in advance that the study investigated the impact of colour cod-
ing in lectures, as to not introduce bias. Instead participants were only told
that the study investigated online lectures, using eye tracking technology. The
participants were then informed about the eye tracker which was calibrated to
their eyes. After this the participants were shown all lecture parts, with each
section followed by the corresponding questions. Finally the concluding ques-
tions were presented after which the participant was told about the purpose of
CHAPTER 3. METHOD 11
the study.
Participants in group A were presented with section 1 and 3 colour coded
while section 2 and 4 were without colour. Similarly participants in group
B were presented sections 2 and 4 colour coded and section 1 and 3 without
colour. Each image in the presentation was shown for the participants in dif-
ferent time spans, in proportion to the amount of text and content in the certain
slide in order for each participant to have the same time to learn and retain the
information on each image. To enforce this limitation the presentation was
pre-recorded. The presentation images, colour coded and without colour can
be found in Appendix A.
The questions were formulated to test the participants’ retention of both
important key words and their understanding of lecture content in general.
Both group A and B were given the same questions, but their answers were
divided, and analysed separately to compare results from colour coded and
non colour coded images. Participants were given approximately 10 questions
per section and had three minutes to fill out each form. All of these questions
can be found in Appendix B. The answers to each question was given a score
between 1 and 0, where 1 is correct and 0 is incorrect, some questions were
given a score of 0.5, meaning they had a partially correct answer. The only
exception to this was the last question of the third section of the test, where
instead each part of the question was graded 1 or 0 and the final score of this
question was the average of these parts. All the questions, answers and their
grading can be found in Appendix B. The grading was done in one sitting and
the same questions were always graded by the same person, so as to make the
grading equal for all answers.
After all experiments were performed the data gathered by the eye tracker
for each test group and image was compiled into one heat-map per image,
showing where on average each groups participants’ focus was directed. Ad-
ditionally the average gaze distance per time unit, e.g. the speed of the point
where the users gaze was directed, was calculated for both the coloured and
black and white images, respectively. This speed was then normalised depend-
ing on the number of total data points within coloured and black and white, so
as not to give a skewed result because one of the samples had more data. The
tests were performed on a screen with resolution 1920⇥1080, meaning that
a participant moving their gaze across the screen horizontally would result in
a traversed distance of 1920 pixels. The eye tracker can produce a different
amount of data points for different tests and different participants, as its read-
ing frequency is not constant and depends on external factors. The formula
used to get the normalised average is T ⇥ D/A, where T is the total gaze dis-
12 CHAPTER 3. METHOD
tance for colour respectively black and white, D is the number of data points
in the colour or black and white sample and A is the average number of data
points.
3.3 Materials
The experiment used an eye tracker of the model "Tobii eye tracker 4C", a re-
mote eye tracker using PCCR that is placed on the computer screen. It was
used to track the eye-movements of participants and collect the data which
could be compiled into heat maps and analysed in various ways after the ex-
periment. The eye tracker was attached to a laptop which the participants used
for watching the presentations as well as reading and filling in the questions.
Google Forms was used to present, answer and compile the answers for the
questionnaires.
A simple script was written in Python to allow the eye tracker to collect
individual data from each separate image in the presentation and another script
was written to analyse the data once all tests were concluded. For the extraction
of data from the eye tracker a program found on GitHub written by GitHub user
’commanderking’ was used by our script. This program, EyeTrackerWidget,
can be found on GitHub at github.com/commanderking/EyeTrackerWidget.
Other non-standard library packages that were used was ’pynput’, ’numpy’,
’pyplot’ which is a part of the ’matplotlib’ library and ’argparse’. ’pynput’
was used for reading input, this helped pause the main script until a certain
button was pressed. ’numpy’ and ’pyplot’ was used to create the heat maps
from the raw eye tracking data. Finally ’argparse’ was used to simplify the
code for passing command-line arguments to one script. Both these scripts
can be found in Appendix D.
3.4 Limitations
Limitations of the study that were identified prior to the experiments include
inaccuracy of the eye tracker, the number of participants as well as varying
difficulties of the different presentation sections. The eye tracker was tested
prior to the study and minor inaccuracies were noticed, especially some time
after calibration. This could affect the resulting heat maps generated after the
experiments and may cause slightly skewed and less accurate heat maps. How-
ever these inaccuracies should be consistent and similar for every user which
results in the gaze distance and speed remaining largely unaffected. The rel-
CHAPTER 3. METHOD 13
Results
In this chapter the results of the survey and the analysed data from the eye
tracker will be presented. The first section 4.1 Survey Results presents the
results gathered from the questionnaires given during the experiments. The
second section 4.2 Heat Maps presents select heat maps generated from the
eye-tracking data. The third and final section 4.3 Eye Tracker Data presents
the results of analysing the eye tracking data.
14
CHAPTER 4. RESULTS 15
Chart 1: Rope
0.6
0.4
0.2
1 2 3 4 5 6 7
Question
1
0.8
0.6
0.4
0.2
1 2 3 4 5 6 7 8
Question
16 CHAPTER 4. RESULTS
1
0.8
0.6
0.4
0.2
1 2 3 4 5 6 7 8 9 10
Question
1
0.8
0.6
0.4
0.2
1 2 3 4 5 6 7 8 9 10
Question
CHAPTER 4. RESULTS 17
Colour coded
Average correctness 1 Black and White
0.8
0.6
0.4
0.2
1 2 3 4
Section
Charts 6.1 & 6.2 presents whether the participants felt like the colours
helped or not, Group A and B respectively. In group A 62.5% found the colours
helpful and in group B the corresponding number was 87.5%. 37.5% of par-
ticipants in group A did not believe the colour made any difference, however
no one thought the colour was negative. In group B no one said that the colour
made no difference, yet one participant stated that the colour had a negative
impact on their memory of the images, this is equal to 12.5% of participants.
In total 75% of participants found the colours helpful, 18.75% did not think
they had any difference and 6.25% stated that it had a negative impact, pre-
sented in chart 6.3.
Yes it helped
87.5% No it was worse
0% It did not make any difference
12.5%
Yes it helped
75% No it was worse
It did not make any difference
18.75%
6.25%
Chart 7.2: Did you find any section more difficult to remember?
Colour coded
8 Black and White
7
6
5
4
3
2
1
Figure 4.1: The heat map over image 9 for participants that were shown this
image in black and white (Group B)
Figure 4.2: The heat map over image 9 for participants that were shown this
image in colour (Group A)
CHAPTER 4. RESULTS 21
Figure 4.3: The heat map over image 15 for participants that were shown this
image in black and white (Group A)
Figure 4.4: The heat map over image 15 for participants that were shown this
image in colour (Group B)
22 CHAPTER 4. RESULTS
Figure 4.5: The heat map over image 17 for participants that were shown this
image in black and white (Group B)
Figure 4.6: The heat map over image 17 for participants that were shown this
image in colour (Group A)
CHAPTER 4. RESULTS 23
Table 4.1: A table of the results found when analysing eye tracking data
Colour Black and white
Total Data Points 2570146 2149717
Total Gaze Distance Covered 2590132.4 pixels 2323417.2
Average Gaze Speed Normalised All Sections 326.3 pixels/s 350.0 pixels/s
Average Gaze Speed Normalised Section 1 301.2 pixels/s 342.9 pixels/s
Average Gaze Speed Normalised Section 2 376.8 pixels/s 396.8 pixels/s
Average Gaze Speed Normalised Section 3 366.8 pixels/s 324.2 pixels/s
Average Gaze Speed Normalised Section 4 231.0 pixels/s 198.8 pixels/s
Chapter 5
Discussion
This chapter will start by discussing the findings in the previous chapter, this
is followed by a discussion on what implications this study has on previous
theory. After this some discovered limitations of this study will be discussed.
Finally some future work will be proposed.
24
CHAPTER 5. DISCUSSION 25
"Det var för mycket text och för mycket klotter på varje slide.
Medans dem andra var mindre text att läsa men mycket mer förk-
larande på ett enkelt sätt." ["There was too much text and too
much clutter on each slide. While the others had less to read, but
far more simplified in an easy way"].
A similar problem might be the case for section 3 as participants were confused
by the nonsensical nature of the information and the colour in the illustration
might have further increased their confusion. As can be seen in the heat maps
in figure 4.5 & 4.6, participants who were shown the images in colour spent
less time on reading the text and more time on the illustration, while partic-
ipants who were shown the images in black and white did the opposite. The
group that were shown this image in black and white therefore might have
spent more time in trying to remember as much as possible, instead of try-
ing to figure out what the illustration meant. The group that was shown the
colour coded image instead tried to understand the illustration, as the colour
26 CHAPTER 5. DISCUSSION
might have attracted their gaze more, which caused them to lose time. This
is further strengthened by the fact that the average gaze speed for students in
the group shown coloured images was higher than the other group for this sec-
tion, as seen in Table 4.1, because that might indicate that the students spent
more time moving their gaze around, rather than reading and taking in infor-
mation. As the groups were only given 90 seconds to watch this image this
might have caused the group who were shown the coloured image to retain
less information and so perform worse in the questions.
Another interesting finding of this study is the fact that participants that
were shown black and white images moved their gaze more on average. Espe-
cially in the first two tests students who were shown coloured images moved
their gaze less, suggesting that they spent less time looking for information and
more time reading and trying to understand. These first two tests also resem-
bled a regular lecture the most, with multiple slides and information on existing
data structures. If students spend less time searching for the relevant informa-
tion they can use the time more efficiently to learn and retain the information.
This can also be seen in the heat maps in figure 4.1-4.6, where the heat maps
are noticeably more dense for the groups that were shown coloured images,
especially around coloured key words. This suggests that a teacher or lecturer
who knows what they want to draw students’ attention to may favourably use
colour to point the students’ attention to where they want. Despite the differ-
ence in methodology and the studies analysing different data generated by the
eye tracker, this aligns with the results of the study conducted by Ozcelik et al.
[7], who concluded that it is easier to find relevant information connected to
images when shown in colour.
The majority of the participants, as presented in chart 6.3, stated that colour
was helpful when remembering the content of the different presentation slides.
However, the results did not reflect this, since a significantly improved result
by colour coded slides were limited to the section "Bloom filter". One possi-
bility is that the participants were given the illusion that it helped, as a result
of a placebo effect. Regardless, this result is of importance since if learning
appears to be easier and less frustrating, it can be more motivating and en-
courages learning further about the topic, which improves the overall learning
experience. Additionally it could be of importance in online lectures that the
content of presentations is presented in an easy to understand way, that is ap-
preciated by the student, as this might cause students to be more motivated
to watch online lectures. In contrast, online lectures that are harder to follow
might cause students to lose their focus due to distractions that may be present
wherever they are studying.
CHAPTER 5. DISCUSSION 27
5.3 Limitations
Limitations to the study include the limited test group and the fact that par-
ticipants may have a greatly varying previous understanding of the concepts
presented in the experiment. Due to the test group being limited to a total of
16 students, which makes it unlikely to be able to draw any statistically con-
clusive conclusions, it was decided not to perform a statistical analysis of the
result. This, in comparison to the two similar studies conducted by Olurinola
& Tayo [23], who had 30 participants in their study, and Ozcelik et al. [7],
who had 52 participants in their study. On average, test group B scored higher
on every section in the questionnaire. This result may be derived from differ-
ent factors on an individual level, such as previous knowledge about similar
data structures or an overall better memory. However this could also be due
to the fact that group B were shown those images that were favoured by the
colour coding in colour and those that were negatively impacted by the colour
coding in black and white. This could make the content of the images easier
to understand and remember. Therefore, this could be a source of error and a
larger test group would give a more conclusive result.
28 CHAPTER 5. DISCUSSION
Conclusions
To conclude the results of this study are slightly mixed, however a few impor-
tant conclusions can be drawn. Primarily, colours seem to have a positive im-
pact on retention when used sparsely. Section 2, "Bloom Filter", illustrates this
as participants who were shown coloured images performed significantly bet-
ter, which suggests better retention of the presentation’s content. This is in line
with previous research performed by Ozcelik et al. [7] and Olurinola & Tayo
[23]. However using too many different colours may instead have an adverse
effect as it instead confuses the reader and reduces retention, which is in line
with research by Deubel et al. [16]. Additionally participants shown colour
coded images moved their gaze less in the presentations that were most similar
to a traditional lecture format, which suggests that more time is spent reading
and learning than searching for information in the images. Even though this
study does not show that colour affects learning and understanding of concepts
positively, better retention of highlighted key words may be positive for learn-
ing in a longer perspective. This study also indicates that teachers and lecturers
can use colours in moderation to draw students’ gaze to certain points in their
presentations and that students perceive lectures that are colour coded as easier
to understand, which may increase their focus.
29
Bibliography
30
BIBLIOGRAPHY 31
[7] Erol Ozcelik et al. “An eye-tracking study of how color coding affects
multimedia learning”. In: Computers & Education 53.2 (2009), pp. 445–
453. : 0360-1315. : 10 . 1016 / j . compedu . 2009 . 03 .
002. : https://www.sciencedirect.com/science/
article/pii/%20S0360131509000712.
[8] Anthony Picciano and Jeff Seaman. “K-12 online learning: A survey
of U.S. School district administrators”. In: Online Learning 11 (Jan.
2007). : 10.24059/olj.v11i3.1719.
[9] Daniel Onah, Jane Sinclair, and R Boyatt. “Dropout Rates of Massive
Open Online Courses: Behavioural Patterns”. In: July 2014. : 10.
13140/RG.2.1.2402.0009.
[10] Indira Dhull and Sakshi Arora. “Online Learning”. In: 3 (May 2019),
pp. 32–34.
[11] Pauline Muljana and Tian Luo. “Factors Contributing to Student Reten-
tion in Online Learning and Recommended Strategies for Improvement:
A Systematic Literature Review”. In: Journal of Information Technol-
ogy Education:Research 18 (Jan. 2019), pp. 19–57. : 10.28945/
4182.
[12] Scott A. Jensen. “In-Class Versus Online Video Lectures: Similar Learn-
ing Outcomes, but a Preference for In-Class”. In: Teaching of Psychol-
ogy 38.4 (2011), pp. 298–302. : 10.1177/0098628311421336.
: https://doi.org/10.1177/0098628311421336.
[13] Dimitrios Vlachopoulos. “COVID-19: Threat or opportunity for on-
line education?” In: Higher Learning Research Communication 10.1
(2020), pp. 16–19. : 10 . 18870 / hlrc . v10i1 . 1179. :
https://www.researchgate.net/publication/342127177_
COVID - 19 _ Threat _ or _ Opportunity _ for _ Online _
Education.
[14] Sonja Folker, Helge Ritter, and Lorenz Sichelschmidt. “Processing and
integrating multimodal material—the influence of color-coding”. In:
Proceedings of the annual meeting of the Cognitive Science Society.
Vol. 27. 27. 2005.
[15] Tanja Keller et al. “Information visualizations for knowledge acqui-
sition: The impact of dimensionality and color coding”. In: Comput-
ers in Human Behavior 22.1 (2006). Instructional Design for Effective
and Enjoyable Computer-Supported Learning, pp. 43–65. : 0747-
5632. : 10 . 1016 / j . chb . 2005 . 01 . 006. : https :
32 BIBLIOGRAPHY
[23] Oluwakemi Olurinola and Omoniyi Tayo. “Colour in Learning: Its Ef-
fect on the Retention Rate of Graduate Students.” In: Journal of Edu-
cation and Practice 6.14 (2015), pp. 1–5.
[24] Quyin Fan. “The Effects of Beacons, Comments, and Tasks on Program
Comprehension Process in Software Maintenance”. AAI3422807. PhD
thesis. USA, 2010. : 9781124226545.
[25] Marta KoÊ-Januchta et al. “Visualizers versus verbalizers: Effects of
cognitive style on learning with texts and pictures – An eye-tracking
study”. In: Computers in Human Behavior 68 (2017), pp. 170–179.
: 0747-5632. : https://doi.org/10.1016/j.chb.
2016.11.028. : https://www.sciencedirect.com/
science/article/pii/S0747563216307695.
[26] Ana Isabel Molina et al. “Evaluating multimedia learning materials in
primary education using eye tracking”. In: Computer Standards Inter-
faces 59 (2018), pp. 45–60. : 0920-5489. : https://doi.
org/10.1016/j.csi.2018.02.004. : https://www.
sciencedirect.com/science/article/pii/S0920548917303392.
[27] Meng-Lung Lai et al. “A review of using eye-tracking technology in
exploring learning from 2000 to 2012”. In: Educational Research Re-
view 10 (2013), pp. 90–115. : 1747-938X. : https://doi.
org / 10 . 1016 / j . edurev . 2013 . 10 . 001. : https :
/ / www . sciencedirect . com / science / article / pii /
S1747938X13000316.
Appendix A
Presentations
Each group was shown every other section in colour and the other in black
and white. Images 1-11 are part of the first section, 12-16 the second, 17 is
the third section and 18 is the fourth section.
A.1 Group A
Group A were shown sections 1 & 3 in colour.
34
APPENDIX A. PRESENTATIONS 35
A.2 Group B
Group B were shown sections 2 & 4 in colour.
53
54 APPENDIX B. QUESTIONS & SURVEY ANSWERS
giver
person
cause
8 where
number
Table B.28: Correct Answers: What argument names belong to which func-
tion
function argument name grading
correct row: 1 pt, one extra argument: 0.5
proclamate giver
pt
correct row: 1 pt, one extra argument: 0.5
rue number, person
pt
gerundive cause correct row: 1 pt
starfruit correct row: 1 pt
hereto where correct row: 1 pt
64 APPENDIX B. QUESTIONS & SURVEY ANSWERS
giver
person
cause
8 where
number
Table B.66: Correct Answers: What argument names belong to which func-
tion
function argument name grading
correct row: 1 pt, one extra argument: 0.5
proclamate giver
pt
correct row: 1 pt, one extra argument: 0.5
rue number, person
pt
gerundive cause correct row: 1 pt
starfruit correct row: 1 pt
hereto where correct row: 1 pt
APPENDIX B. QUESTIONS & SURVEY ANSWERS 79
Group B
Yes it helped
87.5% No it was worse
0% It did not make any difference
12.5%
6 Colour coded
Black and White
5
No
100% 0% Yes
Group B
No
87.5%
Yes
12.5%
No
100% 0% Yes
Group B
No
87.5%
Yes
12.5%
Appendix C
Heat Maps
87
88 APPENDIX C. HEAT MAPS
Fig. C.2.1 Heat map for participants shown image 1 in black and white.
APPENDIX C. HEAT MAPS 97
Fig. C.2.2 Heat map for participants shown image 2 in black and white.
Fig. C.2.3 Heat map for participants shown image 3 in black and white.
98 APPENDIX C. HEAT MAPS
Fig. C.2.4 Heat map for participants shown image 4 in black and white.
Fig. C.2.5 Heat map for participants shown image 5 in black and white.
APPENDIX C. HEAT MAPS 99
Fig. C.2.6 Heat map for participants shown image 6 in black and white.
Fig. C.2.7 Heat map for participants shown image 7 in black and white.
100 APPENDIX C. HEAT MAPS
Fig. C.2.8 Heat map for participants shown image 8 in black and white.
Fig. C.2.9 Heat map for participants shown image 9 in black and white.
APPENDIX C. HEAT MAPS 101
Fig. C.2.10 Heat map for participants shown image 10 in black and white.
Fig. C.2.11 Heat map for participants shown image 11 in black and white.
102 APPENDIX C. HEAT MAPS
Fig. C.2.12 Heat map for participants shown image 12 in black and white.
Fig. C.2.13 Heat map for participants shown image 13 in black and white.
APPENDIX C. HEAT MAPS 103
Fig. C.2.14 Heat map for participants shown image 14 in black and white.
Fig. C.2.15 Heat map for participants shown image 15 in black and white.
104 APPENDIX C. HEAT MAPS
Fig. C.2.16 Heat map for participants shown image 16 in black and white.
Fig. C.2.17 Heat map for participants shown image 17 in black and white.
APPENDIX C. HEAT MAPS 105
Fig. C.2.18 Heat map for participants shown image 18 in black and white.
Appendix D
Python Code
collect.py
import subprocess
import os
from pynput import keyboard
import time
106
APPENDIX D. PYTHON CODE 107
start_listener()
gather_eye_data(90, 18)
analyze.py
import time
import numpy as np
import matplotlib.pyplot as plt
import matplotlib
from matplotlib.colors import LinearSegmentedColormap
import math
import os
import argparse
# Constants:
DISTANCE_AVERAGE_STEP_SIZE = 10
NUM_PERSONS_PER_GROUP = 8
NUM_TESTS = 19
a_path = "A/P"
b_path = "B/P"
col_output_path = "Maps/Colour/FigCol"
bw_output_path = "Maps/BW/FigBW"
total_gaze_distance_col = 0
total_gaze_distance_bw = 0
total_col_data_points = 0
total_bw_data_points = 0
map_data_col = [0]*(1080//scale)
col_data_points = 0
map_data_bw = [0]*(1080//scale)
bw_data_points = 0
if i > NUM_PERSONS_PER_GROUP:
# Reading from group B
i = ((i - 1) % 8) + 1
110 APPENDIX D. PYTHON CODE
path = b_path
else:
# Reading from group A
path = a_path
if num < 10:
# This test was BW
is_col = True
current_test = 1
elif num < 17:
# This test was in Colour
is_col = False
current_test = 2
elif num == 17:
# This test was BW
is_col = True
current_test = 3
else:
# This test was in Colour
is_col = False
APPENDIX D. PYTHON CODE 111
current_test = 4
eye_data = []
eye_data.append([x, y])
map_y = x // scale - 1
map_x = y // scale - 1
if map_x >= 1080 // scale or map_y
>= 1920 // scale:
pass
else:
if is_col:
map_data_col[map_x][map_y]
+= 1
col_data_points += 1
else:
map_data_bw[map_x][map_y]
+= 1
bw_data_points += 1
data_points += 1
112 APPENDIX D. PYTHON CODE
gaze_distance = 0
avg_positions = []
eye_data_length = len(eye_data)
for i in range(0, eye_data_length -
step_size, step_size):
eye_pos_x = []
eye_pos_y = []
for j in range(0, step_size):
if eye_data_length <= i*step_size
+ j:
break
x = eye_data[i*step_size + j][0]
y = eye_data[i*step_size + j][1]
eye_pos_x.append(x)
eye_pos_y.append(y)
if len(eye_pos_x) > 0:
avg_x = sum(eye_pos_x)/len(
eye_pos_x)
avg_y = sum(eye_pos_y)/len(
eye_pos_y)
avg_positions.append([avg_x, avg_y
])
test_gaze_distances["".join([str(
current_test), "col"])][0] +=
gaze_distance
test_gaze_distances["".join([str(
current_test), "col"])][1] +=
col_data_points
else:
global total_gaze_distance_bw
global total_bw_data_points
total_gaze_distance_bw +=
gaze_distance
total_bw_data_points += bw_data_points
test_gaze_distances["".join([str(
current_test), "bw"])][0] +=
gaze_distance
test_gaze_distances["".join([str(
current_test), "bw"])][1] +=
bw_data_points
global should_plot
if should_plot:
plot(map_data_col, num, col_output_path)
print("".join(["(", str(col_data_points), " data
points)\n"]))
if __name__ == ’__main__’:
# Receive command line arguments.
parser = argparse.ArgumentParser("analyze")
parser.add_argument("-no-plot", dest="plot", help="
APPENDIX D. PYTHON CODE 115
should_plot = args.plot
print(should_plot)
# Results printout
print("Total Gaze Distance Travelled for colour:
{}".format(total_gaze_distance_col))
print("Total Gaze Distance Travelled for black-white
: {}".format(total_gaze_distance_bw))
print()
avg_gd_col = total_gaze_distance_col/(911*
NUM_PERSONS_PER_GROUP)
avg_gd_bw = total_gaze_distance_bw / (911*
NUM_PERSONS_PER_GROUP)
total_col_data_points)/2
print()
print("Average Gaze Distance Covered Normalised by
num data points, Colour: {}"
.format(average_data_points * avg_gd_col/
total_col_data_points))
print("Average Gaze Distance Covered Normalised by
num data points, BW: {}"
.format(average_data_points * avg_gd_bw/
total_bw_data_points))
avg_gd_col = test_gaze_distances[keyCol][0]/(
NUM_PERSONS_PER_GROUP*test_lengths[str(i)])
dp_col = test_gaze_distances[keyCol][1]
avg_gd_bw = test_gaze_distances[keyBW][0]/(
NUM_PERSONS_PER_GROUP*test_lengths[str(i)])
dp_bw = test_gaze_distances[keyBW][1]
avg_dp = (dp_col + dp_bw)/2
normalised_gd_col = avg_dp * avg_gd_col / dp_col
normalised_gd_bw = avg_dp * avg_gd_bw / dp_bw
www.kth.se