Comparison Experiment of Learning State Between Visual Programming Language and Text Programming Language

Abstract

Recently, visual programming languages such as Scratch have been popular among novice programmers. Afterward, they employ text-based programming languages such as C and Java. Nevertheless, there are significant barriers between visual and text-based languages. Thus, it is important to establish a seamless transition from visual to text-based languages. In this study, we clarify the difference in the learning process between visual language and text-based language by measuring brain waves. Specifically, experiments will be conducted to solve problems with various difficulty levels for learning visual and text-based languages. The brain waves will be measured, and the values of beta/alpha will be evaluated. Results show that the values of beta/alpha when solving difficult problems increased in the text-based language, but not in the visual language. This suggests that beginners may be thinking differently in the learning process of visual and text-based languages.

Full text

IEEE Copyright Notice
© 2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for
all other uses, in any current or future media, including reprinting/republishing this material for
advertising or promotional purposes, creating new collective works, for resale or redistribution to
servers or lists, or reuse of any copyrighted component of this work in other works.

XXX-X-XXXX-XXXX-X/XX/$XX.00 ©20XX IEEE
Comparison Experiment of Learning State
between Visual Programming Language
and Text Programming Language
Katsuyuki Umezawa
Department of Information Science
Shonan Institute of Technology
Kanagawa, Japan
umezawa@info.shonan-it.ac.jp
Yutaka Ishii
Faculty of Education
Chiba University
Chiba, Japan
yishii@chiba-u.jp
Makoto Nakazawa
Department of Industrial Information Science
Junior College of Aizu
Fukushima, Japan
nakazawa@jc.u-aizu.ac.jp
Michiko Nakano
Faculty of Education and Integrated
Arts and Sciences
Waseda University
Tokyo, Japan
nakanom@waseda.jp
Manabu Kobayashi
Center for Data Science
Waseda University
Tokyo, Japan
mkoba@waseda.jp
Shigeichi Hirasawa
Research Institute for Science and
Engineering
Waseda University
Tokyo, Japan
hira@waseda.jp
Abstract—Recently, visual programming languages such as
Scratch have been popular among novice programmers.
Afterward, they employ text-based programming languages
such as C and Java. Nevertheless, there are significant barriers
between visual and text-based languages. Thus, it is important
to establish a seamless transition from visual to text-based
languages. In this study, we clarify the difference in the
learning process between visual language and text-based
language by measuring brain waves. Specifically, experiments
will be conducted to solve problems with various difficulty
levels for learning visual and text-based languages. The brain
waves will be measured, and the values of β/α will be evaluated.
Results show that the values of β/α when solving difficult
problems increased in the text-based language, but not in the
visual language. This suggests that beginners may be thinking
differently in the learning process of visual and text-based
languages.
Keywords—programming, learning state, visual language,
text language, learning analytics
I. INTRODUCTION
Recently, a visual programming language (hereinafter
visual-type language) is being used as an introduction to
programming. Then, the learner migrates to a text-type
programming language (hereinafter text-type language) such
as C or Java programming language. However, no seamless
transition method has been established.
The purpose of this study is to establish a methodology
for the transition from visual-type language to text-type
language. Once this research is established, programming
language beginners will start learning visual-type languages
and effortlessly and spontaneously transition to learning text-
type languages. This is a critical study that will help
programming instructors handle difficulties that they will
encounter in the future.
In our previous study [14], brain waves in the learning of
visual-type language (Scratch) and text-type language (C
language) were measured. Consequently, we questioned
whether the brains that humans use to learn visual-type and
text-type languages are different.
In this study, we clarified the difference in the learning
process of visual-type and text-type language by measuring
brain waves. Specifically, experiments were conducted to
solve problems of various difficulty levels for learning
visual-type and text-type languages, brain waves were
measured concurrently, and β/α values evaluated.
Section 2 provides an outline of related works; Section 3
describes the participants of the experiment, the tasks used in
the experiment, and the experimental method; Section 4
shows the experimental results, and the discussion is given in
Section 5, and Section 6 presents the conclusion of the work.
II. RELATED WORK
A. Visual-type language and text-type language
Visual-type languages fall into two major categories:
block-type imperative languages and flow-type functional
languages.
Mason et al. conducted hundreds of experiments to
program simple problems designed to be similar in the two
categories of block-type and flow-type languages. They
conducted an empirical study to evaluate the relative benefits
of the two categories [1].
Mladenović et al. surveyed student misunderstandings on
loops, one of the basic concepts of programming, in 207
elementary school students. The students learned three
programming languages: a block-type language (Scratch)
and a text-type language (Logo and Python). They observed
that block-type languages minimized misunderstandings
about loops. This difference became more obvious as tasks
became more complex, such as nested loops. The legitimacy
of employing a visual language for programming beginners
is argued in this study since it does not generate syntax
mistakes [2]. However, there is no mention of bridging the
gap between the two languages in this study.
Daskalov et al. proposed an environment for beginners to
use a hybrid-type language of text-type and visual-type
languages. It is a hybrid-type environment of flow-type
visual language and text-type language instead of block-type

language and claims to be suitable for training novice
programmers [3].
Weintrop compared text-type, visual-type, and hybrid-
type languages. In conclusion, while hybrid-type languages
showed characteristics of both text-type and visual-type
languages, they also demonstrated that hybrid-type
languages outperformed block-type and text-type languages
in certain dimensions [4].
Tóth et al. highlighted the existence of a gap between
visual-type and text-type languages. They observed the
migration from a visual-type language (MIT App Inventor 2)
to a text-type language (Android Studio) using Java Bridge
Code Generator as a mediator of knowledge transfer. They
claimed that the gap between visual-type and text-type
languages was bridged by the Java Bridge Code Generator
[5].
B. Browsing history/editing history system
Aramoto et al. proposed a web-based browsing history
visualization system that collects the browsing history of
PDF contents [6]. In addition, they developed an editing
history visualization system that collects not only browsing
history but also programming editing history [7]. Nakano et
al. [8] applied this system to English learning, whereas Goto
et al. [9] applied it to programming learning and
experimentally evaluated the effect. In addition, sufficient
studies have been conducted to improve developers’ coding
skills by analyzing the editing process of programming [10]
[11].
C. Application of brain waves to learning
Giannitrapani estimated the learner's learning state by
measuring the α and β waves obtained by performing a
discrete Fourier transform on the brain waves. Here he
observed that low beta waves increased during intellectual
work [12]. Uwano et al. also discovered that the ratio of α
waves to β waves can effectively estimate the learner’s
learning state [13]. Yoshida et al. also showed that the
learner's learning state can be estimated by measuring the
ratio of α waves to β waves [14]. In our previous experiment,
we used a typing software that can change the difficulty level
of the learning material and observed that β/α increased
during the execution of difficult tasks, and confirmed that
low-β/low-α affected the difficulty level [15] [16].
To better understand the characteristics of learners,
several studies have been undertaken to monitor brain waves
during programming learning. Crk et al. used
electroencephalograph (EEG) to directly measure
programmer expertise. They proposed a basic approach for
investigating the role of expertise in understanding
programming languages [17]. Lee et al. also used EEG to
observe the difference between programming beginners and
experts. They deduced from EEG data that programming
experts have excellent abilities in understanding programs
[18].
III. E
XPERIMENT
A. Experiment participants
We hosted a “Matsudai Science Course” for high school
students in a neighborhood, mostly Matsudai High School
students from Niigata Prefecture, and an experiment was
conducted in the science course [19]. Seven students
participated in the experiment using a visual-type language
(Scratch), whereas nine students participated using a text-
type language (C language). The text-type language
experiment was conducted using two sets of easy and
difficult problems.
B. Task
Figs. 1 and 2 show the tasks used in the experiment of
visual-type language (Scratch), whereas Figs. 3–6 show the
problems of the text-type language (C language).
Fig. 1. Easy Question of Visual Language (Scratch)
Fig. 2. Difficult Question of Visual Language (Scratch)
Fig. 3. Easy Question 1 of Text-based Language (C Language)
Fig. 4. Difficult Question 1 of Text-based Language (C Language)
Fig. 5. Easy Question 2 of Text-based Language (C Language)
Fig. 6. Difficult Question 2 of Text-based Language (C Language)
C. How to measure brain waves
The EEG was measured using a NeuroSky EEG control
MindWave® Mobile headset. The log collection application
collects brain wave logs via TCP/IP communication with
ThinkGear Connector after connecting the headset and
connector via Bluetooth. The ThinkGear Connector is a
driver that provides a communication function with the
Let's create a program that meets the following conditions.
When you press the up arrow key, “Saru-kun” moves upward on the stage.
Conversely, when you press the down arrow key, “Saru-kun” moves downward
on the stage. In both cases, the amount of movement is equivalent to 10 in
coordinate values.
Let's add
the following functions to the program in question above.
When you click the flag, “Cat-chan” keeps moving in the left-right direction
(horizontal direction).
“Cat-chan” bounces back when it reaches the left and right "edges".
“Cat-chan” moves as if walking.
When “Saru-kun” hits “Cat-chan”, he says “I was killed”
When
“Saru-kun”
can touch
“Banana”
safely, the banana is hidden.
(After showing an example that repeats 5 times)
Modify the “for” statement in the example and create a program to repeat it 10
times and execute it.
There is a program that displays “1st repeat”, “2nd repeat” ... “10th repeat”.
Create and execute a program that modifies this so that it is displayed as
“10th repeat”, “20th repeat” ... “100th repeat”.
(After showing an example of finding the sum of up to 10)
Create and execute a program that calculates the sum of 1 to 100 and displays
the calculation result.
T
he expression i% 2 == 1 means "the remainder of dividing i by 2 is e qual to
1." In other words, "i is an odd number". Use this fact to create and execute a
program that calculates and displays the sum of odd-numbered values from 1
to 100.

MindWave Mobile headset provided by NeuroSky Inc. In
addition, the types of brain waves that can be acquired are
the eight types shown in Table I, and each value is a 4-byte
floating-point number without a unit [20].
As shown in Table I, the EEG used for this measurement
measures two types of brain waves: high frequency and low
frequency for α and β wave, respectively. In particular, when
considering β/α, which is the ratio of α wave to β wave, four
types of combinations of βl/αl, βh/αh, βl/αh, and βh/αl can be
considered. In addition, the average ratio (βl + βh) / (αl + αh)
of low frequency and high frequency (hereinafter βl+h/αl+h)
was added, and we focused on all five types of β/α.
TABLE I. THE KIND OF BRAIN WAVES WHICH CAN BE ACQUIRED
Kind Frequency (Hz)
δ wave
θ wave
low α wave (αl)
high α wave (αh)
low β wave (βl)
high β wave (βh)
low γ wave
mid γ wave
0.5-2.75
3.5-6.75
7.5-9.25
10-11.75
13-16.75
18-29.75
31-39.75
41-49.75
IV. EXPERIMENTAL RESULT
In the experiment, the brain waves at 1-s intervals were
initially measured while each participant was solving the task.
Then, various β/α per second, that is, five types of βl/αl, βh/αh,
βl/αh, βh/αl, and βl+h/αl+h were calculated. The average of
various β/α values while solving the task was calculated. The
calculated average values are shown in the Tables V to X in
the Appendix.
Tables II, III, and IV show the ratios of various β/α
values when solving easy and difficult tasks. For example, an
numerical value in a cell in Table II show the results of
division between the numerical values in the cells at the
same location in Tables V and VI in the Appendix. The value
of the gray shaded cell shows 1.00 or more. The numerical
values marked with * represent the significant numerical
values (p-value ≥ 0.05) at the significance level of 5% in the
t-test (test of the difference between the average values).
Moreover, it showed that this average value was considered
different. In particular, the gray shaded areas marked with *
indicate that the average β/α values were statistically higher
when solving difficult tasks. However, the part marked with
* without gray shading indicates that the average value of β/α
was statistically lower when solving a difficult task. The
specific values of the p-value are shown in Tables XI to XIII
of the Appendix.
TABLE II.
R
ATIO OF
“D
IFFICULT
”
TO
“E
ASY
”
IN
V
ISUAL
L
ANGUAGE
ID
β
l
/
α
l
β
h
/
α
h
β
l
/
α
h
β
h
/
α
l
β
l+h
/α
l+h
ma002
1.051
1.041
1.069
1.040
1.047
ma004
0.814
1.047
1.007
0.877
0.976
ma006
0.932
0.739*
1.020
0.759*
0.909
ma008
0.822
0.886
1.016
0.703*
0.869*
ma011
0.852
0.771*
0.983
0.651*
0.805*
ma013
0.832
0.875
1.081
0.691*
0.879
ma016
0.861
0.943
1.021
0.791
0.924
TABLE III.
R
ATIO OF
“D
IFFICULT
”
TO
“E
ASY
”
IN
T
EXT
-
BASED
L
ANGUAGE
1
ID
β
l
/
α
l
β
h
/
α
h
β
l
/
α
h
β
h
/
α
l
β
l+h
/α
l+h
ma002
1.938*
0.895
1.049
1.558
1.264*
ma005
0.519*
1.382
1.020
0.693
1.153
ma006
1.740
0.859
1.171
1.171
1.046
ma007
0.850
1.029
0.992
1.074
1.026
ma011
1.264
1.592*
1.092
1.945*
1.466*
ma013
1.272
1.560*
1.195
2.119*
1.505*
ma015
0.805
1.119
1.153
0.892
1.021
ma021
1.150
1.361*
1.120
1.344
1.224*
ma024
1.468
1.637
1.611
1.384
1.108
ma026
0.971
1.113
1.072
0.983
1.053
TABLE IV.
R
ATIO OF
“D
IFFICULT
”
TO
“E
ASY
”
IN
T
EXT
-
BASED
L
ANGUAGE
2
ID
β
l
/
α
l
β
h
/
α
h
β
l
/
α
h
β
h
/
α
l
β
l+h
/α
l+h
ma001
0.972 0.885 0.600 1.292 0.961
ma003
0.683 1.072 0.875 1.025 0.969
ma004
0.916 1.769* 1.113 1.468 1.354*
ma007
1.293 1.067 1.217 1.154 1.320*
ma009
1.111 1.051 0.922 1.077 0.966
ma014
1.146 1.394* 1.199 1.214 1.225*
ma016
0.830 0.965 0.836 0.927 0.916
ma022
1.115 1.164 1.194* 1.195 1.233*
ma023
1.653* 1.403* 1.106 1.990* 1.302*
V. CONSIDERATION
Comparing the experimental result of the visual-type
language (Table II), with the experimental results of the text-
type language (Tables III and IV) showed that the visual-type
language had less gray shading, whereas the text-type had
many gray shades. This implied that when learning a text-
type language, the values of various β/α were higher when
solving difficult tasks, but this was not the case with the
visual-type language.
The difficulty level of the visual-type language task may
have been set inappropriately, but it is also likely that the
visual- and text-type language use the brain differently, or
think in different ways.
VI. CONCLUSION
We conducted experiments to solve tasks with various
difficulty levels for learning visual-type and text-type
languages, and measured the brain waves concurrently. By
analyzing the β/α value indicating the difficulty level of the
tasks according to the experiments, it was confirmed that the
β/α value increases with the difficulty level for the text-type
language. However, the β/α values for visual-type language
experiments did not increase with the difficulty level. This
suggests that in the learning process of visual-type language
and text-type language, different thinking processes were
performed.
In the future, we will conduct more extensive and
rigorous experiments to further our study. In addition,
educational content (intermediate language) that bridges the
gap between the two language types will be developed.
Finally, we aim to minimize the proportion of students who
are frustrated in learning text-type languages, improve
learning efficiency, and enhance comprehension.

ABOUT RESEARCH ETHICS
The Research Ethics Committee of Shonan Institute of
Technology has approved these experiments. We also have
received consent to participate in this experiment from
participants and their parents.
ACKNOWLEDGMENT
Part of this research was carried out as part of the
research project “Research on e-learning for next-generation”
of the Waseda Research Institute for Science and
Engineering, Waseda University. Part of this work was
supported by JSPS KAKENHI Grant Numbers JP20K03082,
JP19H01721, JP19K04914, and JP17K01101, and Special
Account 1010000175806 of the NTT Comprehensive
Agreement on Collaborative Research with the Waseda
University Research Institute for Science and Engineering.
Research leading to this paper was partially supported by the
grant as a research working group ICT and Education of
JASMIN.
REFERENCES
[1] D. Mason and K. Dave, “Block-based versus flow-based
programming for naive programmers,” 2017 IEEE Blocks and
Beyond Workshop (B&B), 2017, pp. 25-28, doi:
10.1109/BLOCKS.2017.8120405.
[2] M. Mladenoviá, I. Boljat, Ž. Žanko “Comparing loops
misconceptions in block-based and text-based programming
languages at the K-12 level,” Education and Information
Technologies 23(4), pp. 1483-1500, 2018.
[3] R. Daskalov, G. Pashev, S. Gaftandzhieva, “Hybrid Visual
Programming Language Environment for Programming Training,”
TEM Journal. Volume 10, Issue 2, pp. 981-986, May 2021.
[4] D. Weintrop, U. Wilensky, “Between a Block and a Typeface:
Designing and Evaluating Hybrid Programming Environments,” IDC
'17: Proceedings of the 2017 Conference on Interaction Design and
Children, pp. 183-192, June 2017.
[5] T. Tóth, G. Lovászová, “Mediation of Knowledge Transfer in the
Transition from Visual to Textual Programming,” Informatics in
Education, DOI 10.15388/infedu. 2021.
[6] M. Aramoto, D. Koizumi, T. Suko, S. Hirasawa, “The e-learning
materials production supporting system based on the existing PDF
file,” 76th National Convention of Information Processing Society of
Japan, Vol. 4, pp. 359-360. 2014.
[7] M. Aramoto, M. Kobayashi, M. Nakazawa, M. Nakano, M. Goto, S.
Hirasawa, “Learning Analytics via Visualization System of Edit
Record - System Configuration and Implementation,” 78th National
Convention of Information Processing Society of Japan, Vol. 4, pp.
527-528. 2016.
[8] M. Nakano, M. Aramoto, S. Yoshida, K. Koutou, “Learning
Analytics via Visualization System of Edit Record - Application to
English Writing Task: Error Gravity and Error Correction Time,”
78th National Convention of Information Processing Society of Japan,
Vol. 4, pp. 531-532. 2016.
[9] M. Goto, K. Mikawa, G. Kumoi, M. Kobayashi, M. Aramoto, S.
Hirasawa, “Learning Analytics via Visualization System of Edit
Record - Analytics Model Based on Edit Record and Evaluation
Score Data for C-Programming Courses,” 78th National Convention
of Information Processing Society of Japan, Vol. 4, pp. 533-534. 2016.
[10] P. Ardimento, M. Cimitile, M. L. Bernardi, F. M. Maggi, “Evaluating
Coding Behavior in Software Development Processes,” A Process
Mining Approach. In: 2019 IEEE/ACM International Conference on
Software and System Processes (ICSSP), pp. 84-93. 2019.
[11] P. Ardimento, M. L. Bernardi, M. Cimitile, G. D. Ruvo, “Mining
Developer's Behavior from Web-Based IDE Logs,” 2019 IEEE 28th
International Conference on Enabling Technologies: Infrastructure for
Collaborative Enterprises (WETICE), pp. 277-282. 2019.
[12] D. Giannitrapani, “The role of 13-hz activity in mentation,” The EEG
of Mental Activities, pp. 149-152. 1988.
[13] H. Uwano, K. Ishida, Y. Matsuda, S. Fukushima, N. Nakamichi, M.
Ohira, K. Matsumoto, and Y. Okada, “Evaluation of Software
Usability Using Electroencephalogram - Comparison of Frequency
Component between Different Software Versions,” Journal of Human
Interface Society, vol. 10(2), pp. 233-242. 2008.
[14] K. Yoshida, Y. Sakamoto, I. Miyaji, K. Yamada, “Analysis
comparison of brain waves at the learning status by simple
electroencephalography,” Proceedings, Knowledge-Based Intelligent
Information and Engineering Systems (KES'2012), pp. 1817-1826.
2012.
[15] K. Umezawa, T. Ishida, T. Saito, M. Nakazawa, S. Hirasawa,
“Collection and analysis of the browsing history, editing history, and
biological information for high school students,” National Conference
of JASMIN 2016 Autumn, Japan Society for Management
Information, D2-1. pp. 1-6. 2016.
[16] K. Umezawa, T. Ishida, T. Saito, M. Nakazawa, S. Hirasawa, “A
judgment method of difficulty of task for a learner using simple
electroencephalograph,” Information Processing Society of Japan
(IPSJ) SIG Technical Report (CE-137), pp. 1-6. 2016.
[17] I. Crk, T. Kluthe, A. Stefik, “Understanding Programming Expertise:
An Empirical Study of Phasic Brain Wave Changes,” ACM
Transactions on Computer-Human Interaction, pp. 1-29. 2015.
[18] S. Lee, A. Matteson, D. Hooshyar, S. Kim, J. Jung, G. Nam, H. Lim,
“Comparing Programming Language Comprehension between
Novice and Expert Programmers Using EEG Analysis,” IEEE 16th
International Conference on Bioinformatics and Bioengineering
(BIBE), pp. 350-355. 2016.
[19] K. Umezawa, T. Ishida, M. Kobayashi, S. Hirasawa, “Effectiveness
Evaluation of Practical Use of the Electronic Teaching Materials for
University Education,” National Conference of JASMIN 2013
Autumn, Japan Society for Management Information, pp. 45-48. 2013.
[20] ThinkGear Serial Stream Guide, http://developer.neurosky.com/docs/
doku.php?id=thinkgear_communications_protocol. Last accessed 12
November 2020.
APPENDIX
A. Experimental result
1) Average value of brain waves during the experiment
Tables V to X show the average values of various β/α
(βl/αl，βh/αh，βl/αh，βh/αl，βl+h/αl+h) when solving problems
with different difficulty levels for visual and text languages.
TABLE V. AVERAGE OF
Β
/
Α
WHEN SOLVING AN EASY QUESTION IN
VISUAL LANGUAGE
ID
β
l
/
α
l
β
h
/
α
h
β
l
/
α
h
β
h
/
α
l
β
l+h
/α
l+h
ma002
1.181
1.117
1.218
1.082
0.835
ma004
1.546
3.623
1.576
3.290
1.863
ma006
1.351
2.179
1.410
1.982
1.224
ma008
1.825
2.242
1.569
2.482
1.436
ma011
1.322
1.694
1.283
1.952
1.167
ma013
1.411
1.319
1.189
1.398
0.971
ma016
1.611
1.404
1.455
1.687
1.100
TABLE VI. AVERAGE OF Β/Α WHEN SOLVING A DIffiCULT QUESTION
IN VISUAL LANGUAGE
ID
β
l
/
α
l
β
h
/
α
h
β
l
/
α
h
β
h
/
α
l
β
l+h
/α
l+h
ma002 1.242
1.163
1.302
1.126
0.874
ma004 1.258
3.794
1.587
2.887
1.819
ma006 1.259
1.610
1.437
1.504
1.113
ma008 1.500
1.985
1.594
1.746
1.249
ma011 1.127
1.306
1.260
1.271
0.939
ma013 1.174
1.154
1.286
0.966
0.854
ma016 1.386
1.325
1.486
1.334
1.016

TABLE VII.
A
VERAGE OF
Β
/
Α
WHEN
S
OLVING AN
E
ASY
Q
UESTION
1
IN
T
EXT
-
BASED
L
ANGUAGE
ID
β
l
/
α
l
β
h
/
α
h
β
l
/
α
h
β
h
/
α
l
β
l+h
/α
l+h
ma002
1.078
1.243
1.300
1.062
0.881
ma005
1.406
1.738
1.096
2.346
0.993
ma006
1.169
2.389
1.185
2.420
1.328
ma007
1.307
1.346
1.449
1.015
0.915
ma011
1.057
1.484
1.222
1.065
0.922
ma013
0.989
1.083
1.146
0.856
0.748
ma015
1.536
1.347
1.347
1.147
0.952
ma021
1.130
0.994
1.301
0.939
0.824
ma024
0.887
0.923
0.988
0.896
0.724
ma026
1.233
1.326
1.226
1.408
0.953
TABLE VIII.
A
VERAGE OF
Β
/
Α
WHEN
S
OLVING A
D
IffiCULT
Q
UESTION
1
IN
T
EXT
-
BASED
L
ANGUAGE
ID
β
l
/
α
l
β
h
/
α
h
β
l
/
α
h
β
h
/
α
l
β
l+h
/α
l+h
ma002
2.089
1.113
1.363
1.655
1.113
ma005
0.730
2.401
1.118
1.626
1.144
ma006
2.034
2.053
1.387
2.835
1.390
ma007
1.111
1.386
1.437
1.091
0.939
ma011
1.336
2.362
1.335
2.070
1.352
ma013
1.258
1.689
1.369
1.814
1.125
ma015
1.237
1.507
1.553
1.024
0.971
ma021
1.299
1.353
1.457
1.262
1.008
ma024
1.302
1.510
1.593
1.241
0.803
ma026
1.198
1.476
1.314
1.383
1.003
TABLE IX.
A
VERAGE OF Β
/
Α WHEN
S
OLVING AN
E
ASY
Q
UESTION
2
IN
T
EXT
-
BASED
L
ANGUAGE
ID
β
l
/
α
l
β
h
/
α
h
β
l
/
α
h
β
h
/
α
l
β
l+h
/α
l+h
ma001
1.509
1.688
2.126
1.258
1.169
ma003
1.521
1.285
1.364
1.111
0.900
ma004
1.279
1.600
1.295
1.443
0.970
ma007
0.950
1.240
1.134
1.008
0.738
ma009
1.683
0.943
1.620
1.089
0.939
ma014
1.192
1.326
1.137
1.259
0.909
ma016
2.164
1.290
1.677
1.818
1.157
ma022
1.108
1.248
0.957
1.158
0.770
ma023
0.766
1.522
1.101
1.083
0.894
TABLE X.
A
VERAGE OF
Β
/
Α
WHEN
S
OLVING A
D
IffiCULT
Q
UESTION
2
IN
T
EXT
-
BASED
L
ANGUAGE
ID
β
l
/
α
l
β
h
/
α
h
β
l
/
α
h
β
h
/
α
l
β
l+h
/α
l+h
ma001
1.467
1.495
1.276
1.626
1.124
ma003
1.039
1.377
1.194
1.139
0.872
ma004
1.171
2.831
1.441
2.120
1.313
ma007
1.228
1.323
1.380
1.164
0.974
ma009
1.869
0.991
1.494
1.174
0.907
ma014
1.366
1.848
1.363
1.529
1.113
ma016
1.795
1.245
1.401
1.686
1.060
ma022
1.236
1.453
1.143
1.384
0.949
ma023
1.266
2.135
1.218
2.155
1.165
2) t-test result
Table XI shows the results (p-values) of the t-test (test of
the difference between the average values) on the raw data
before averaging in Tables V and VI. Similarly, the results of
the t-tests in Tables VII and VIII are shown in Table XII, and
the results of the t-tests in Tables IX and X are shown in
Table XIII. The * mark in the table indicates that it became
significant at the significance level of 5%, that is, it was
judged that there was a difference in the average value.
TABLE XI. T-TEST RESULT (P-VALUE) IN VISUAL LANGUAGE
ID
β
l
/
α
l
β
h
/
α
h
β
l
/
α
h
β
h
/
α
l
β
l+h
/α
l+h
ma002
0.634883
0.545313
0.322132
0.705520
0.339147
ma004
0.125522
0.611459
0.931893
0.227844
0.724510
ma006
0.598656
0.001288*
0.818535
0.038909*
0.121765
ma008
0.092467
0.156451
0.813730
0.013392*
0.005916*
ma011
0.096757
0.000440*
0.817327
0.015376*
0.000055*
ma013
0.440751
0.175280
0.421831
0.031084*
0.109491
ma016
0.185285
0.470217
0.802335
0.104487
0.121738
TABLE XII. T-TEST RESULT (P-VALUE) IN TEXT-BASED LANGUAGE 1
ID
β
l
/
α
l
β
h
/
α
h
β
l
/
α
h
β
h
/
α
l
β
l+h
/α
l+h
ma002
0.005392*
0.458485
0.784792
0.059061
0.038411*
ma005
0.023048*
0.184138
0.923266
0.373443
0.365842
ma006
0.105114
0.339583
0.199367
0.568820
0.728297
ma007
0.361896
0.817342
0.946617
0.579817
0.730354
ma011
0.177854
0.001264*
0.355504
0.000023*
0.000057*
ma013
0.260772
0.011038*
0.255414
0.006023*
0.003613*
ma015
0.457094
0.545623
0.346812
0.529667
0.848399
ma021
0.521512
0.021955*
0.374611
0.120032
0.034931*
ma024
0.212846
0.127854
0.058032
0.470085
0.380936
ma026
0.849875
0.399918
0.558195
0.929348
0.581759
TABLE XIII. T-TEST RESULT (P-VALUE) IN TEXT-BASED LANGUAGE 2
ID
β
l
/
α
l
β
h
/
α
h
β
l
/
α
h
β
h
/
α
l
β
l+h
/α
l+h
ma001
0.896547
0.623422
0.151539
0.176466
0.755684
ma003
0.222387
0.669967
0.328523
0.892585
0.767879
ma004
0.757246
0.000218*
0.519064
0.053904
0.001915*
ma007
0.084254
0.829427
0.418251
0.369240
0.001495*
ma009
0.569347
0.711265
0.687064
0.650284
0.703209
ma014
0.591098
0.000428*
0.080738
0.180132
0.010632*
ma016
0.329402
0.751488
0.177898
0.707091
0.279207
ma022
0.741026
0.217580
0.044875*
0.453410
0.004701*
ma023
0.002532*
0.000871*
0.314160
0.000007*
0.001484*