Frontiers in Education Technology Vol. 5, No.2, 2022 www.scholink.org/ojs/index.php/fet ISSN 2576-1846 (Print) ISSN 2576-1854 (Online) 67 Original Paper Characteristics of Eye Movements during Musical Expression in Early Childhood by Tonality: Through Quantitative Analysis by Eye Tracking Mina Sano1 1 Department of Education and Care of Early Childhood, Faculty of Education and Care of Early Childhood, Tokoha University, Shizuoka-City, Japan Received: June 1, 2022 Accepted: June 23, 2022 Online Published: July 1, 2022 doi:10.22158/fet.v5n2p67 URL: http://dx.doi.org/10.22158/fet.v5n2p67 Abstract Early childhood children frequently use body movements while singing songs accompanied to the music. When trying to express the recognition of musical elements by body movement, the child in early childhood creates his own musical expression by thinking and judging while looking at the surrounding children and the accompaniment teacher. This study aims to quantitatively analyze changes in eye movements during musical expression. 3-year-old, 4-year-old, and 5-year-old children at two nursery schools in 2020 and two kindergartens in 2021 (n=118) participated in eye tracking during singing a song using an eye tracker (Tobii3). Quantitative analysis by three-way ANOVA was mainly conducted on the calculated data. As a result, the increase in data such as number of saccade occurrences and size, and the moving average velocity of saccade, showed that saccades during musical expression in early childhood tended to be larger in major key than in minor key. From the calculated data on saccade, which is the eye movement during musical expression, it was predicted that effective feature quantities of eye movement during musical expression for machine learning could be derived in the same way as feature quantities depended on the results of quantitative analysis of body movement during musical expression. Keywords saccade during musical expression in early childhood, ANOVA, tonality, nursery rhyme, the moving average velocity of saccade, feature quantity of movement www.scholink.org/ojs/index.php/fet Frontiers in Education Technology Vol. 5, No. 2, 2022 68 Published by SCHOLINK INC. 1. Introduction The author has used a 3D motion capture system (MVN) to analyze the movement in musical expression of early childhood children and clarify the close relationship between the recognition of musical elements and body movements in early childhood (Sano, 2018). In particular, as the practical process of the step- by-step activity deepened the recognition of the musical elements of the children, the increase in the moving average acceleration of the hands and pelvis by the analysis of movement showed characteristic changes with age and differences among child facilities. The author used motion capture technology to extract the developmental characteristics of musical expressions in early childhood, and devised a method for evaluating the developmental process of musical expression in early childhood applying machine learning technique (Fernandez-Delgado et al., 2014; Sano, 2019; 2020). In the research report that used 3D motion capture in the field of education, some discussions were presented on the results of specific analysis of movement in traditional Japanese dance and sawing, mainly for adults, and the development of learning support methods (Ando et al., 2012; Sato et al., 2010). In addition, research on music and movement showed the reaction to adult sounds (Burger, 2013), viewing experiments and video analysis on the relationship between the movement and expression of the performer (Dahl & Friberg, 2007; Thompson & Luck, 2012). However, these researchers did not capture the musical expressions that occurred in the process of continuous musical practice for young children. Regarding machine learning methods, it is also used in educational fields such as behavior recognition and individual recognition methods in daily life (Kodama et al., 2015; Takada et al., 2012), and learning support of movement such as upturning of elementary school students (Matsumoto et al., 2014). But, there have been no research reports that have used machine learning for musical expression in early childhood. The author has conducted machine learning using feature quantities of movement that have caused characteristic changes in the practical process of musical expression in early childhood. As a result, a certain level of classification accuracy was obtained by several classifiers, and derived a discriminant model for the musical expression in early childhood (Sano, 2019; 2020). The author focused on the eye movements of early childhood children during musical expression because of the need for additional feature quantities of movement based on the results of these studies. The participant children in the activity of musical expression often watched around other children or a teacher’s accompaniment. The children participating in the activity of musical expression often watched around other children or a teacher’s accompaniment in my investigation by the motion capture technique. The author thought the eye movement during musical expression affected the body movement in musical expression by the participant child. Previous studies on eye movements have shown quantitative analysis of eye movements and eye movements related to character recognition, and how eye movements affect judgment by cognitive function (Higuchi, 2019; Watanabe et al., 2019; Kusunoki et al., 2017; Mpofu, 2016; Seong-un Kim, 2016; Moreno-Estevaa et al., 2018; Lin et al., 2018；Reumont, 2020; Valtakari et al., 2021). An analysis www.scholink.org/ojs/index.php/fet Frontiers in Education Technology Vol. 5, No. 2, 2022 69 Published by SCHOLINK INC. of the line of sight of a craftsman during forging work using a camera has also be conducted (Kitajima et al., 2017). For research using eye trackers in the field of music, eye movements during rhythm hearing and reading (Plöchl et al., 2017; Lörch et al., 2017; Fusase, 2017), gaze behavior of musical instrument players (Bishopa et al., 2019; Vandemoortele et al., 2018), eye-hand synchronization during instrumental performance (Marandola, 2017), differences in reading and listening to melody between music experts and non-music experts (Drai-Zerbib & Baccino, 2018; Puutinen, 2018) etc. are presented, but the attention was mainly paid to the participant’s line of sight looking at the display. Studies such as eye contact on the possibility of enhancing music conducting skills through a virtual reality learning environment have also been shown (Orman et al., 2017; Orman, 2016). There is a study (Burger et al., 2017) that attempted to analyze the response to music by combining eye tracking and motion capture, but the subject of the study was adults (Fink et al., 2019). The author considered using a glasses-type as a wearable system of eye tracker to capture changes in eye movements during musical expression in early childhood children. The eye movements captured by the eye tracker are mainly divided into saccade (rapid eye movement) and fixation. The movement of the eyeball can be quantified by the moving velocity (angular velocity) and the moving distance (magnitude of the angle). To that end, the author thought to focus on saccades during musical expression and find effective feature quantities of eye movement. 2. Purpose of This Study This study aims to quantitatively analyze the eye movements of early childhood children during musical expression using an eye tracker, and to find effective feature quantities of eye movement in the same method the feature quantities of movement depended on the results of analysis of movement in musical expression. In this paper, the author focuses on changes related to saccade when a child singing a song, and extract characteristic differences depended on tonality, songs, and age. 3. Method In order to quantitatively capture the saccade that occurs when an early childhood child is singing, the following research method is used. 3.1 Eye Tracking of Children in Early Childhood When Singing a Song Children in early childhood are singing songs and watching the accompaniment of the teachers and the children around them. In my past studies, it was confirmed by video analysis when observing the behavior of children in early childhood, but it was not possible to quantitatively capture the actual aspect of eye movements. Therefore, in this study, the author tried to quantify the eye movements of children in early childhood during singing by using an eye tracker (Tobii Glass 3). Tobii Glass 3 is a glasses-type eye tracker. Each participant child wears it in the same way as glasses, is calibrated while being fixed with a www.scholink.org/ojs/index.php/fet Frontiers in Education Technology Vol. 5, No. 2, 2022 70 Published by SCHOLINK INC. strap, starts singing with a signal, and eye-tracks until the end of the singing. It is a 1/50 second time frame at 50Hz, and audio and video are recorded at the same time. 3.2 The Participant Children of Eye Tracking, Inspection Contents and Inspection Schedule Table 1 below shows the number of children who participated in eye tracking in 2020 and 2021 and the data acquisition dates for each song in major and minor. Table 1. Songs by Children Participating in Eye Tracking and The Date Acquiring Data in 2020 and 2021 U nursery school (n=28) M nursery school (n=30) Y kindergarten (n=30) N kindergarten (n=30) Major key “Kaerunouta” Lyrics: Toshiaki Okamoto, Composer: German Folk Song August 19, 2020 14: 30-16: 00 June 21, 2021 9: 30-11: 00 July 6, 2021 9: 30-11: 00 "Umi" (G major,) Lyrics: Ryuha Hayashi, Composer: Takeshi Inoue August 18, 2020 9: 30-11: 00 “Musundehiraite” Lyrics: Unknown, Composition: Rousseau May 31, 2021 9: 30-11: 00 “Tewotatakimasho” Lyrics: Junichi Kobayashi, Composer: Spanish Folk Song June 7, 2021 9: 30-11: 00 June 29, 2021 9: 30-11: 00 “Shiawasenaratewotatakou” Lyrics: Kimura Rihito Composer: American folk song June 8, 2021 9: 30-11: 00 minor key “Ureshiihinamatsuri” Lyrics: Hachirou Satou, Composer: Koyo Kawamura August 18, 2020 9: 30-11: 00 “Darumasan” Nursery rhyme in Japan June 7, 2021 9: 30-11: 00 “Hotarukoi” Nursery rhyme in Japan June 8, 2021 9: 30-11: 00 “Teruterubouzu” Lyrics: Asahara Kagamimura Composer: Shinpei Nakayama August 19, 2020 14: 30-16: 00 June 29, 2021 9: 30-11: 00 “Genkotsuyamanotanuki” Lyrics: Yoshiko Kayama Composer: Akihiro Komori June 21, 2021 9: 30-11: 00 In 2020, it was difficult to adjust multiple survey schedules due to COVID-19. In Osaka prefecture, at U nursery school from 9:30 to 11:00 on August 18 and 14:30 to 16:00 on August 19 at M nursery school, Only one measurement was taken. Since the situation was similar in 2021, the author decided to conduct only a short-term inspection schedule, and eye tracking when singing 5 songs each at 2 kindergartens in Shizuoka prefecture according to the schedule shown in Table 1. Data at Y kindergarten was acquired in www.scholink.org/ojs/index.php/fet Frontiers in Education Technology Vol. 5, No. 2, 2022 71 Published by SCHOLINK INC. 9:30-11:00 on May 31, June 7, and June 21st. Data at N kindergarten was acquired in 9:30-11:00 on June 8, June 29, and July 6th. In 2020, the number of subjects was 8 for 3-year-old children, 10 for 4-year-old children, and 10 for 5- year-old children at U nursery school. At M nursery school, there were 9 of 3-year-olds, 9 of 4-year-olds, and 12 of 5-year-olds. In 2021, the participant number of children was 10 each for 3-year-old children, 4-year-old children, and 5-year-old children in Y kindergarten and N kindergarten. These investigations have been approved by the research ethics committee to which the author belongs, as well as permission by the person in charge of the kindergarden that cooperates in the research, the parents of the participant children, and the submission of consent forms. Here, out of the songs experienced by the participants in the four child facilities during their everyday life, one song was sung in major key and another one in minor key. The songs are as follows. In 2020, at U nursery school, the song in major key is “Umi” (G major, lyrics: Hayashi Yanaginami, composition: Takeshi Inoue), and the song in minor key is “Ureshii Hina Matsuri” (a minor, lyrics: Sato Hachiro, composition: Mitsuyo Kawamura). The participant children sung, while in M nursery school, the song in major key was “Kaeru nouta” (C major, lyrics: Toshiaki Okamoto, composition: German folk song), and the song in minor key was “Teruterubozu” (a minor, Written by Kagamimura Asahara, composed by Shinpei Nakayama) was sung. In 2021, in Y kindergarten, the songs in major key are “Musundehiraitte” (lyrics: unknown, composition: Rousseau), “Tewotatakimashou” (lyrics: Junichi Kobayashi, composition: Spanish folk song), “Kaerunouta” (lyrics: Toshiaki Okamoto, composition: German folk song) is sung, and “Darumasan” (Nursery rhyme=warabeuta) and “GenkotsuyamanoTanuki” (lyrics: Miko Kayama, composition: Akihiro Komori) are sung in minor key. In N kindergarten, the songs as “Shiawasenaratewotatakou” (lyrics: translation: Kimura Rihito, composition: American folk song), “Tewotatakimashou” (lyrics: Junichi Kobayashi, composition: Spanish folk song), “Kaerunouta” (lyrics: Toshiaki Okamoto, composition: German folk song) are sung in major key, and the songs are “Hotarukoi” (warabeuta) and “Teruterubozu” (lyrics: Kagamimura Asahara, composition: Shinpei Nakayama) are sung in minor key. The measurement time per person was 20 seconds for songs in major key and 20 seconds for songs in minor key. 3.3 Quantitative Analysis of Data Acquired by Eye Tracking In this study, in order to capture eye movements, the author conducted a saccade-specific quantitative analysis from fixation to next fixation. Therefore, the author quantitatively analyzed the number of occurrences of saccade, the moving average velocity (angular velocity), the average value of the size of movement (total angle), the average of moving distance (total angle), the angular velocity of the moving average of the first saccade (rapid eye movement), and the magnitude (angle) of the first saccade. It was inspected whether these data differed by early childhood facility, age, and major key/minor key. www.scholink.org/ojs/index.php/fet Frontiers in Education Technology Vol. 5, No. 2, 2022 72 Published by SCHOLINK INC. 4. Result In this paper, among the calculated data, an example of individual calculation data of eye movement analyzed by analysis software (Tobii Pro analyzer) and a part of the result of quantitative analysis on the data items presented by methodology are shown. Regarding the occurrence of saccade, the analysis results of “Tewotatakimashou” (major key) “Kaeru no Uta” (major key), “Darumasan” (minor key), and “Teruterubouzu” (minor key), an individual example is shown by age such as 3-year-old, 4-year-old, and 5-year-old. Firstly, an example of individual calculation data for eye movement are shown for a song in major key by age. 4.1 Movement of the Line of Sight When Singing “Tewotatakimashou” (major key) 4.1.1 Analysis Results of a 3-year-old Child When Singing “Tewotatakimashou” The following Figure 1 shows the order of eye movements of a 3-year-old child when singing “Tewotatakimashou” and Figure 2 shows the heat map. Figure 1. Gaze Plot When A 3-Year-Old Child Sings “Tewotatakimashou” www.scholink.org/ojs/index.php/fet Frontiers in Education Technology Vol. 5, No. 2, 2022 73 Published by SCHOLINK INC. Figure 2. Heat Map When A 3-Year-Old Child Sings “Tewotatakimashou” 4.1.2 Analysis Results When a 4-year-old Child Sings “Tewotatakimashou” The following Figure 3 shows the order of eye movements of a 4-year-old child when singing “Tewotatakimashou” and Figure 4 shows the heat map. Figure 3. Gaze Plot When A 4-Year-Old Child Sings “Tewotatakimashou” www.scholink.org/ojs/index.php/fet Frontiers in Education Technology Vol. 5, No. 2, 2022 74 Published by SCHOLINK INC. Figure 4. Heat Map When A 4-Year-Old Child Sings “Tewotatakimashou” 4.1.3 Analysis Results When a 5-year-old Child Sings “Tewotatakimashou” The following Figure 5 shows the order of eye movements of a 5-year-old child when singing “Tewotatakimashou” and Figure 6 shows the heat map. Figure 5. Gaze Plot When A 5-Year-Old Child Sings "Tewotatakimashou" www.scholink.org/ojs/index.php/fet Frontiers in Education Technology Vol. 5, No. 2, 2022 75 Published by SCHOLINK INC. Figure 6. Heat Map When A 5-Year-Old Child Sings “Tewotatakimashou” The above figures regarding an individual case are examples of visualization of the movement of the line of sight. The gaze plot shows the order of movement of the line of sight, and the size of circle indicates the relative length of time that the line of sight stayed there and the participant child was gazing at that part. The heat map shows that the time spent watching in the order of red, yellow, and yellow-green was relatively long. All of the participant children performed their own singing and accompanying movements while paying close attention to the teachers and other children nearby. Regarding songs in major key with body movements such as “Tewotatakimashou” and “Kaeru no Uta”, the 3-year-old saccade tended to alternate between the teacher on the right side and the other children on the left side, and finally the other children each other. However, 4-year-olds and 5-year-olds tended to look at the center of the place where the participant child could see the teacher and other children at the same time. Regarding songs in minor key such as “Darumasan”, participant children tended to sing with facing forward without moving their eyes too much. Regarding “Teruterubozu” in minor key, the eyes of the participating children were relatively moving because singing was accompanied by their own spontaneous movements with watching around other children at the same time. Next, in order to clarify whether there is a statistically significant difference or not in characteristic differences between major key and minor key from these eye tracking results, the author conducted a three-way ANOVA (non-repeated tonality factors: 2 levels, non-repeated early childhood facility factors: 4 levels, non-repeated age factors: 3 levels). www.scholink.org/ojs/index.php/fet Frontiers in Education Technology Vol. 5, No. 2, 2022 76 Published by SCHOLINK INC. 4.2 Characteristic Differences in Saccade between Major Key and Minor Key Regarding the data acquired by eye tracking, the author focused the change of saccade as calculated data. The following results were obtained as a result of a three-way ANOVA (non-repeated factors as 2 levels in major key/minor key, non-repeated factors as 4 levels in U nursery school, M nursery school, Y kindergarten, N kindergarten, non-repeated factors as 3 levels in 3-year-old, 4-year-old, 5-year-old). 4.2.1 The Number of Occurrences of Saccade Table 2 below shows the average number of saccade occurrences by major key/minor key and child facility. Table 2. The Number of Occurrences of Saccade Major/minor key Child facility Age Average SD N 3-year-old 19.875 9.26495 8 U nursery 4-year-old 18 10.50926 10 school 5-year-old 14.9 8.33267 10 3-year-old 14.7778 7.01388 9 M nursery 4-year-old 11 8.80656 10 Major key school 5-year-old 16 6.45262 12 3-year-old 26.4483 14.49036 29 Y kindergarten 4-year-old 26.1 18.75311 30 5-year-old 30.3333 16.75654 30 3-year-old 16.8966 15.13592 29 N kindergarten 4-year-old 18.7 14.72308 30 5-year-old 15.1071 12.32985 28 3-year-old 21.125 5.66789 8 U nursery 4-year-old 18.1 9.46866 10 school 5-year-old 24.2 12.14542 10 3-year-old 4.8889 1.83333 9 M nursery 4-year-old 6.5556 4.50309 9 minor key school 5-year-old 8.5833 3.75278 12 3-year-old 12 4.45814 17 Y kindergarten 4-year-old 12.1111 4.25495 18 5-year-old 15.35 6.45857 20 3-year-old 19 11.49534 15 N kindergarten 4-year-old 21.5 13.54719 20 5-year-old 13.3 10.60834 20 As a result of three-way ANOVA for these acquired data, a main effect/interaction between subjects of the test showed a statistically significant difference (major/minor factor (F(1, 379)=9.953, p<.005, child facility factor (F(3, 379)=9.922, p<.005), major/minor factor * child facility factor (F(3, 379) =11.938, p<.005)). Therefore, a simple main effect test and a multiple comparison test by Bonferroni’s method www.scholink.org/ojs/index.php/fet Frontiers in Education Technology Vol. 5, No. 2, 2022 77 Published by SCHOLINK INC. were conducted. Concerning major/minor factor/ child facility factor* major/minor factor * age factor, the simple main effect was statistically significant for 3-year-old children (F(1, 379)=15.036, p <.005), 4-year-old (F(1, 379)=14.796, p<.005), and 5-year-old (F(1, 379)=18.106, p<.005) in Y kindergarten. As a result of multiple comparison, in Y kindergarten, major key was larger than minor key. Concerning child facility factor/child facility factor* major/minor factor * age factor, the simple main effect was statistically significant for major key (4-year-old: (F(3, 379)=4.486, p<.005), 5-year-old: (F(3, 379)=9.537, p<.005)). As a result of multiple comparison, 3-year-old and 5-year-old showed a statistically significance in Y kindergarten, larger than 3 other facilities regarding major key. 4-year-old in Y kindergarten was significantly larger than M nursery school in major key. Regarding minor key, 3- year-old children in U nursery school and N kindergarten were larger than M nursery school, 4-year-old children in N nursery school were larger than M nursery school, and 5-year-old children in U nursery school were larger than M nursery school. A statistically significant difference was not observed by age, but it was found that the average number of occurrences was significantly larger in major than in minor. 4.2.2 The Moving Average Velocity (angular velocity) of Saccade (degrees/second) Table 3. The Moving Average Velocity of Saccade Major/ minor key Child facility Age Average SD N 3-year-old 225.815 66.5225 8 U nursery 4-year-old 224.392 37.61292 10 school 5-year-old 254.645 63.00146 10 3-year-old 238.0011 45.50136 9 M nursery 4-year-old 218.273 47.78794 10 Major key school 5-year-old 253.1667 63.9497 12 3-year-old 191.97 97.63757 29 Y kindergarten 4-year-old 221.4853 89.85153 30 5-year-old 349.7953 531.74638 30 3-year-old 187.5766 33.10906 29 N kindergarten 4-year-old 216.9787 55.67087 30 5-year-old 205.26 41.1883 28 3-year-old 231.2688 47.18051 8 U nursery 4-year-old 246.329 66.13405 10 school 5-year-old 226.766 37.44762 10 3-year-old 186.3311 44.71741 9 M nursery 4-year-old 220.0822 56.74914 9 minor key school 5-year-old 259.2358 61.12917 12 3-year-old 215.8435 40.8591 17 Y kindergarten 4-year-old 213.5211 54.991 18 5-year-old 221.8145 40.48186 20 www.scholink.org/ojs/index.php/fet Frontiers in Education Technology Vol. 5, No. 2, 2022 78 Published by SCHOLINK INC. 3-year-old 208.7093 33.18712 15 N kindergarten 4-year-old 195.1075 39.5047 20 5-year-old 198.2135 34.7814 20 Similarly, for the calculation data of the moving average velocity of saccade, a three-way ANOVA was conducted based on the major/minor factor, child facility factor, and the age factor. As a result of the test of the effect between the subjects, a statistically significant difference was not observed, but for songs in major key, 5-year-olds were larger than 3-year-olds and 4-year-olds in Y kindergarten. 4.2.3 Average Size of Saccade Similarly for the calculation data of the average value of the size of saccade, a three-way ANOVA was conducted based on major/minor factors, child facility factors, and age factors. As a result of the test of the effect between subjects, the main effect/interaction showed a statistically significant difference (major/minor factor: F (1, 379)=8.499, p<.005, child facility factor: F(3, 379) =12.407, p<.005, major/minor factors * child facility factor: F (3, 379)=12.308, p<.005). Therefore, a simple main effect test and a multiple comparison test by Bonferroni’s method were conducted. Concerning the major/minor factor/major/minor * child facility * age factor, a simple main effect was statistically significant in Y kindergarten (3-year-old child: F (1, 379)=15.076, p<.005, 4-year-old child: F (3, 379)=18.671, p<.005, 5-year-old child: F (3, 379)=24.166, p<.005). As a result of multiple comparisons, 3-year-old, 4-year-old, and 5-year-old in Y kindergarten showed larger in major key than minor key. Concerning the child facility factor/major/minor factor * child facility factor * age factor, a simple main effect was statistically significant in major (3-year-old child: F(3, 379)=8.518, p<.005, 4-year-old: F(3, 379)=10.357, p<.005, 5-year-old child: F(3, 379)=12.585, p<.005). As a result of multiple comparisons, 3-year-old, 4-year-old, and 5-year-old children in Y kindergarten were significantly larger than other three facilities in major key. 4.2.4 The Average Value of the Total Moving Distance of Saccade Similarly, for the calculation data of the total moving distance of saccade, a three-way ANOVA was conducted based on the factors of major/minor factor, child facility factor, and age factor. www.scholink.org/ojs/index.php/fet Frontiers in Education Technology Vol. 5, No. 2, 2022 79 Published by SCHOLINK INC. Table 4. Average Total Moving Distance of Saccade (Degrees) Major/ minor key Child facility Age Average SD N 3-year-old 177.3925 87.82232 8 U nursery 4-year-old 166.643 106.03598 10 school 5-year-old 126.014 84.17718 10 3-year-old 148.9222 99.34946 9 M nursery 4-year-old 117.874 114.52879 10 Major key school 5-year-old 165.0575 96.21823 12 3-year-old 190.5079 132.80593 29 Y kindergarten 4-year-old 184.7997 141.44637 30 5-year-old 204.1963 103.82075 30 3-year-old 114.8897 127.50648 29 N kindergarten 4-year-old 138.708 121.25421 30 5-year-old 111.4611 105.18776 28 3-year-old 204.5575 83.86732 8 U nursery 4-year-old 175.55 106.52629 10 school 5-year-old 235.001 148.61351 10 3-year-old 40.8122 30.54065 9 M nursery 4-year-old 55.8367 48.28457 9 minor key school 5-year-old 96.085 56.75702 12 3-year-old 103.5582 54.15056 17 Y kindergarten 4-year-old 92.6878 40.46515 18 5-year-old 123.2695 59.87629 20 3-year-old 139.6707 102.94324 15 N kindergarten 4-year-old 143.3985 105.13491 20 5-year-old 97.488 89.47006 20 As a result of the test of the effect between subjects, the main effect/interaction was statistically significant (child facility factor: F (3, 379)=6.539, p<.005, major/minor factor * child facility factor: F (3, 379)=8.23, p<.005). Therefore, a simple main effect test and a multiple comparison test by Bonferroni’s method were conducted. Concerning the major/minor factor/child facility factor * major/minor factor * age factor, a simple main effect was statistically significant in Y kindergarten (4-year-old: F (1, 379)=8.718, p<.005)) regarding major key. As a result of multiple comparisons, 5-year-old in minor key was larger than major key in U nursery school, 3-year-old in major was larger than minor key in M nursery school. 3-year-old, 4-year- old, and 5-year-old in major key was larger than minor key in Y kindergarten. Concerning the child facility factor/major/minor factor * child facility factor * age factors, a simple main effect was statistically significant in minor key (5-year-old: F (3, 379)=4.438, p<.005). As a result of multiple comparison, in major key, 3-year-old and 5-year-old in Y kindergarten were larger than N www.scholink.org/ojs/index.php/fet Frontiers in Education Technology Vol. 5, No. 2, 2022 80 Published by SCHOLINK INC. kindergarten. Regarding minor key, 3-year-old in U nursery school was significantly larger than M nursery school and 5-year-old in U nursery school was significantly than other three facilities. Figure 7 shows the total moving distance of saccade in M nursery school, and Figure 8 shows the total moving distance of saccade in Y kindergarten. Figure 7. The Total Moving Distance of Saccade M nursery School Figure 8. The Total Moving Distance of Saccade Y Kindergarten As shown in Figure 7 and Figure 8, the total moving distance of saccade tended to be larger in major key than in minor key, which was remarkable in M nursery school and Y kindergarten. 4.2.5 The Size of the First Saccade Similarly, for the calculation data regarding the size of the first saccade, a three-way ANOVA was conducted based on major / minor factor, child facility factor, and age factor. www.scholink.org/ojs/index.php/fet Frontiers in Education Technology Vol. 5, No. 2, 2022 81 Published by SCHOLINK INC. Table 5. First Saccade Size (Degrees) Major/ minor key Child facility Age Average SD N 3-year-old 10.24 5.26771 8 U nursery 4-year-old 6.64 3.65792 10 school 5-year-old 6.061 2.17636 10 3-year-old 12.4633 9.17476 9 M nursery 4-year-old 11.579 8.34354 10 Major key school 5-year-old 13.4 7.1227 12 3-year-old 76.2845 129.89091 29 Y kindergarten 4-year-old 71.2517 105.14894 30 5-year-old 64.698 93.78386 30 3-year-old 6.7159 5.66665 29 N kindergarten 4-year-old 7.0097 4.80204 30 5-year-old 6.7146 5.00763 28 3-year-old 8.2925 6.50261 8 U nursery 4-year-old 8.67 5.047 10 school 5-year-old 12.444 8.62215 10 3-year-old 4.9422 3.86937 9 M nursery 4-year-old 8.84 5.01415 9 minor key school 5-year-old 10.0992 7.26828 12 3-year-old 8.5282 4.7698 17 Y kindergarten 4-year-old 7.1517 3.98349 18 5-year-old 8.9725 5.78889 20 3-year-old 7.3647 7.30578 15 N kindergarten 4-year-old 6.7825 4.90401 20 5-year-old 8.723 9.07599 20 As a result of the test of the effect between subjects, the main effect/interaction was statistically significant (child facility factor: F (3, 379)=10.239, p<.005, major/minor factor * child facility factor: F (3, 379)=10.222, p<.005). Therefore, a simple main effect test and a multiple comparison test by Bonferroni’s method were conducted. Concerning the major/minor factor/major/minor factor * child facility factor * age factor, the simple main effect was statistically significant in major key (3-year-old child (F (3, 379)=17.614, p<.005), 4-year-old child (F (3, 379)=16.548, p<.005) and 5-year-old children (F (3, 379)=13.34, p<.005). As a result of multiple comparison, in songs of major key, Y kindergartens for 3-year-olds, 4-year-olds, and 5-year- olds were larger than U kindergarten, M kindergarten, and N kindergarten. Concerning the child facility factor/major/minor factor * child facility factor * age factor, the simple main effect was statistically significant in major key (3-year-old (F(3, 379)=9.796, p<.005), 4-year-old (F(3, 379)=9.015, p<.005), 5-year-old (F(1, 394)=7.196, p<.005)) in Y kindergarten. As a result of www.scholink.org/ojs/index.php/fet Frontiers in Education Technology Vol. 5, No. 2, 2022 82 Published by SCHOLINK INC. multiple comparison, 3-year-old, 4-year-old, and 5-year-old children in Y kindergarten showed larger magnitude in major key than other three facilities. Figure 9. First Saccade Size in M Nursery School Figure 10. First Saccade Size in Y Kindergarten As shown in Figure 9 and Figure 10, the size of the saccade that occurred at the first time tended to be larger in major key than in minor key, which was remarkable in M nursery school and Y kindergarten. 4.3 The Characteristics of Saccade by Songs Next, in order to inspect whether a statistically significant difference was observed or not in the calculated data regarding saccade for all 10 specific songs, a two-way ANOVA was carried out (non-repeated song factor as 10 levels and non-repeated age factor as 3 levels). 4.3.1 The Number of Occurrences of Saccade by Songs Table 6 shows the average data regarding number of occurrences of saccade by song. www.scholink.org/ojs/index.php/fet Frontiers in Education Technology Vol. 5, No. 2, 2022 83 Published by SCHOLINK INC. Table 6. Number of Occurrences of Saccade by Song Song Age Average SD N 3-year-old 15.4138 8.65001 29 Kaerunouta 4-year-old 16.4 10.54252 30 5-year-old 17.7097 10.31243 31 3-year-old 19.875 9.26495 8 Umi 4-year-old 18 10.50926 10 5-year-old 14.9 8.33267 10 3-year-old 22.1111 14.47795 9 Musundehiraite 4-year-old 23.4 21.22472 10 5-year-old 29.7 17.21143 10 3-year-old 31.5789 18.7626 19 Tewotatakimashou 4-year-old 29.1 21.06356 20 5-year-old 29.1053 19.24374 19 3-year-old 14.4 9.41866 10 Shiawasenara 4-year-old 16.2 12.85647 10 Tewotatakou 5-year-old 12.6 10.65833 10 3-year-old 21.125 5.66789 8 Ureshii 4-year-old 18.1 9.46866 10 Hinamasturi 5-year-old 24.2 12.14542 10 3-year-old 9.75 3.05894 8 Darumasan 4-year-old 11.4444 5.17472 9 5-year-old 11.5 4.74342 10 3-year-old 16.125 10.66955 8 Hotarukoi 4-year-old 16 10.79094 10 5-year-old 10.3 6.11101 10 3-year-old 12.7222 11.88576 18 Teruterubouzu 4-year-old 17.3158 14.86253 19 5-year-old 12.0909 10.00433 22 3-year-old 14 4.69042 9 Genkostuyamano 4-year-old 12.7778 3.27024 9 Tanuki 5-year-old 19.2 5.71159 10 As a result of the test of the effect between subjects, the main effect was statistically significant with the music factor (F (9, 375)=9.64, p<.005). As a result of multiple comparison, the 3-year-old child’s “Tewotatakimashou” was significantly larger than “Kaerunouta”, “Shiawasenaratewotatakou”, “Darumasan”, “Teruterubouzu”, and “Genkotsuyamanotanuki”. 4-year-old child’s “Tewotatakimashou” was significantly larger than “Kaerunouta” and “Darumasan”. Regarding 5-year-old, “Musundehiraite” was significantly larger than “Hotarukoi” and “Teruterubouzu”, “Tewotatakimashou” was significantly larger than “Darumasan”, “Hotarukoi”, and “Teruterubouzu”. www.scholink.org/ojs/index.php/fet Frontiers in Education Technology Vol. 5, No. 2, 2022 84 Published by SCHOLINK INC. Figure 11. Age-Specific Changes in The Number of Saccades That Occur Depending on The Song As shown in Figure 11, even among the songs in major key, the number of occurrences in “Tewotatakimashou” and “Musundehiraite” were remarkable. 4.3.2 The Moving Average Velocity of Saccade In order to inspect whether a statistically significant difference was observed or not in the calculated data of the moving average velocity of saccade, a two-way ANOVA was conducted by the song factor (10 levels) and the age factor (3 levels). Table 7. The Moving Average Velocity of Saccade (Degrees/Second) Song Age Average SD N 3-year-old 198.4259 47.81486 29 Kaerunouta 4-year-old 215.969 51.89155 30 5-year-old 222.711 54.63452 31 3-year-old 225.815 66.5225 8 Umi 4-year-old 224.392 37.61292 10 5-year-old 254.645 63.00146 10 3-year-old 202.4444 175.4381 9 Musundehiraite 4-year-old 222.4 140.10567 10 5-year-old 643.1 875.51286 10 3-year-old 190.6111 27.01139 19 Tewotatakimashou 4-year-old 204.255 53.5489 20 5-year-old 202.4784 22.30516 19 3-year-old 195.09 30.01467 10 Shiawasenara 4-year-old 219.518 64.88319 10 www.scholink.org/ojs/index.php/fet Frontiers in Education Technology Vol. 5, No. 2, 2022 85 Published by SCHOLINK INC. Tewotatakou 5-year-old 209.701 48.17227 10 3-year-old 231.2688 47.18051 8 Ureshii 4-year-old 246.329 66.13405 10 Hinamasturi 5-year-old 226.766 37.44762 10 3-year-old 211.405 36.22365 8 Darumasan 4-year-old 197.7244 23.98627 9 5-year-old 216.874 53.97441 10 3-year-old 212.2813 24.66576 8 Hotarukoi 4-year-old 210.799 36.65742 10 5-year-old 195.801 41.2738 10 3-year-old 192.9467 44.82111 18 Teruterubouzu 4-year-old 198.6789 50.6791 19 5-year-old 232.595 56.64364 22 3-year-old 219.7889 46.40676 9 Genkostuyamano 4-year-old 229.3178 72.72708 9 Tanuki 5-year-old 226.755 22.18492 10 As a result of the test of the effect between subjects, the main effect/interaction was statistically significant (song factor: F (9, 375)=3.054, p<.005, song factor * age factor: F (18, 375)=2.803, p<.005). Therefore, a simple main effect test and a multiple comparison test by Bonferroni’s method were conducted. Concerning the song factor/song factor * age factor, a simple main effect was statistically significant in 5-year-old (F(9, 375)=8.641, p<.005). As a result of multiple comparison, 5-year-old children showed significantly larger data in “Musundehiraite” than in other nine songs. Concerning the age factor/song factor * age factor, age factor, a simple main effect was statistically significant in “Musundehiraite” (F(2, 375)=28.012, p<.005). As a result of multiple comparison, 5-year- old children showed significantly larger data in “Musundehiraite” than 3-year-old and 4-year-old children. 4.3.3 Average Size Data of Saccade In order to inspect whether a statistically significant difference was observed or not in the calculated data of the size of Saccade, a two-way ANOVA was conducted by the song factor (10 levels) and the age factor (3 levels). www.scholink.org/ojs/index.php/fet Frontiers in Education Technology Vol. 5, No. 2, 2022 86 Published by SCHOLINK INC. Table 8. Average Size Data of Saccade (Degrees) Song Age Average SD N 3-year-old 7.5052 2.85719 29 Kaerunouta 4-year-old 8.6037 2.96961 30 5-year-old 8.4171 3.48662 31 3-year-old 9.1463 2.63067 8 Umi 4-year-old 8.805 2.01257 10 5-year-old 7.839 3.1224 10 3-year-old 400.9067 173.44916 9 Musundehiraite 4-year-old 401.422 120.26263 10 5-year-old 442.11 181.74859 10 3-year-old 6.6184 1.22373 19 Tewotatakimashou 4-year-old 6.748 2.26008 20 5-year-old 7.3774 1.79441 19 3-year-old 6.266 1.81167 10 Shiawasenara 4-year-old 6.916 1.92473 10 Tewotatakou 5-year-old 6.937 2.29484 10 3-year-old 9.4875 1.90246 8 Ureshii 4-year-old 9.721 3.46217 10 Hinamasturi 5-year-old 9.339 2.53537 10 3-year-old 8.88 2.02932 8 Darumasan 4-year-old 8.1267 3.01173 9 5-year-old 7.616 2.90651 10 3-year-old 6.71 2.10036 8 Hotarukoi 4-year-old 6.932 2.33268 10 5-year-old 7.118 3.12936 10 3-year-old 7.3222 2.98825 18 Teruterubouzu 4-year-old 6.7268 2.44988 19 5-year-old 9.1723 4.32873 22 3-year-old 8.3289 2.94255 9 Genkostuyamano 4-year-old 7.53 2.19371 9 Tanuki 5-year-old 8.293 2.59907 10 As a result of the test of the effect between subjects, the main effect was statistically significant in the song factor (F (9, 375)=276.291, p<.005). According to multiple comparisons, 3-year-old, 4-year-old and 5-year-old showed significantly larger data in “Musundehiraite” than other nine songs. 4.3.4 The Total Moving Distance of Saccade In order to inspect whether a statistically significant difference was observed or not in the calculated data of the total moving distance of saccade, a two-way ANOVA was conducted by the song factor (10 levels) and the age factor (3 levels). www.scholink.org/ojs/index.php/fet Frontiers in Education Technology Vol. 5, No. 2, 2022 87 Published by SCHOLINK INC. Table 9. The Average of Total Moving Distance of Saccade Song Age Average SD N 3-year-old 126.8666 102.49737 29 Kaerunouta 4-year-old 141.7307 107.32831 30 5-year-old 147.6052 92.5503 31 3-year-old 177.3925 87.82232 8 Umi 4-year-old 166.643 106.03598 10 5-year-old 126.014 84.17718 10 3-year-old 150.9378 152.44822 9 Musundehiraite 4-year-old 135.103 112.99221 10 5-year-old 168.456 79.01437 10 3-year-old 218.5353 152.32141 19 Tewotatakimashou 4-year-old 217.206 178.77244 20 5-year-old 213.4763 140.28595 19 3-year-old 100.709 96.94952 10 Shiawasenara 4-year-old 123.663 116.54486 10 Tewotatakou 5-year-old 91.112 83.91147 10 3-year-old 204.5575 83.86732 8 Ureshii 4-year-old 175.55 106.52629 10 Hinamatsuri 5-year-old 235.001 148.61351 10 3-year-old 86.0938 32.70565 8 Darumasan 4-year-old 88.7289 46.00924 9 5-year-old 92.945 56.501 10 3-year-old 113.3513 98.03137 8 Hotarukoi 4-year-old 114.315 93.43521 10 5-year-old 77.22 53.50204 10 3-year-old 97.6444 96.73537 18 Teruterubouzu 4-year-old 117.2289 104.78197 19 5-year-old 105.9355 86.21995 22 3-year-old 119.0822 65.97339 9 Genkostuyamano 4-year-old 96.6467 36.43381 9 tanuki 5-year-old 153.594 48.29937 10 As a result of the test of the effect between subjects, the main effect was statistically significant in the song factor (F (9, 375)=6.67, p<.005). As a result of multiple comparison, in 3-year-old child children’s data regarding the average of total moving distance of saccade, “Tewotatakimashou” was larger than “Teruterubouzu”. 5-year-old children’s data of “Ureshiihinamatsuri” was larger than “Hotarukoi”. 4.3.5 The First Saccade In order to inspect whether a statistically significant difference was observed or not in the calculation data of the size of the saccade that occurred the first time, a two-way ANOVA with the song factor (10 levels) and the age factor (3 levels) was conducted. Table 10 shows the first size of saccade (degrees). www.scholink.org/ojs/index.php/fet Frontiers in Education Technology Vol. 5, No. 2, 2022 88 Published by SCHOLINK INC. Table 10. The First Size of Saccade (Degrees) Song Age Average SD N 3-year-old 7.7986 6.61434 29 Kaerunouta 4-year-old 8.8443 6.67036 30 5-year-old 9.0384 5.91631 31 3-year-old 10.24 5.26771 8 Umi 4-year-old 6.64 3.65792 10 5-year-old 6.061 2.17636 10 3-year-old 232.0822 139.33696 9 Musundehiraite 4-year-old 199.29 90.89385 10 5-year-old 181.769 74.0519 10 3-year-old 6.2274 3.9628 19 Tewotatakimashou 4-year-old 6.876 4.68213 20 5-year-old 7.1658 5.81135 19 3-year-old 8.596 8.30099 10 Shiawasenara 4-year-old 5.776 3.10534 10 Tewotatakou 5-year-old 5.572 2.68285 10 3-year-old 8.2925 6.50261 8 Ureshii 4-year-old 8.67 5.047 10 Hinamasturi 5-year-old 12.444 8.62215 10 3-year-old 9.7813 4.37451 8 Darumasan 4-year-old 8.3511 4.33887 9 5-year-old 8.852 6.24701 10 3-year-old 6.0675 4.87504 8 Hotarukoi 4-year-old 6.014 4.76115 10 5-year-old 10.701 11.69227 10 3-year-old 6.5317 6.71919 18 Teruterubouzu 4-year-old 8.1616 5.00062 19 5-year-old 8.5745 6.54268 22 3-year-old 7.4144 5.0786 9 Genkostuyamano 4-year-old 5.9522 3.41426 9 Tanuki 5-year-old 9.093 5.62926 10 As a result of the test of the effect between subjects, the main effect was statistically significant in the song factor (F (9, 375)=148.385, p<.005). As a result of multiple comparison, 3-year-old’s data of “Musundehiraite” was significantly larger than the other 9 songs in 4-year-olds and 5-year-olds. 5. Discussion In this article, the author inspected whether a characteristic difference was observed in eye movements during singing by early childhood children depending on whether the song is in major key or minor key. Firstly, regarding the acquired data by eye tracking, the author specified in saccade and conducted a three-way ANOVA based on major key/minor key (2 levels), child facility (4 levels), and age (3 levels). As a result of the analysis, it was found that saccades occurred more frequently in major key than in www.scholink.org/ojs/index.php/fet Frontiers in Education Technology Vol. 5, No. 2, 2022 89 Published by SCHOLINK INC. minor key. Regarding the moving average velocity of saccade, the data of 5-year-old children at Y kindergarten was larger than 3-year-old and 4-year-old in major key. Regarding the size of saccade, the data in major key was significantly larger than the data in minor key regarding 3-year-old, 4-year-old, and 5-year-old children in Y kindergarten. The total moving distance of saccade tended to be larger in major key than in minor key, which was more remarkable in Y kindergarten. On the other hand, for 5- year-old children in U nursery school, the total moving distance of saccade was larger in minor key than in major key, but the data of 3-year-old children in M nursery school in major key was larger than the data in minor key. The size of the first saccade that occurred was remarkable in Y kindergarten. The size of the first saccade occurred in Y kindergarten and M nursery school were clearly characterized by being larger in major key than in minor key. Secondly, in order to inspect whether a statistically significant difference was observed in the calculated data regarding saccade for all 10 specific songs, a two-way ANOVA was carried out (non-repeated song factor (10 levels), non-repeated age factor (3 levels)). As a result, regarding the number of occurrences of saccade, the number of occurrences of “Tewotatakimashou” and “Musundehiraite” were remarkable in the songs of major key. As for the average moving velocity of saccade, “Musundehiraite” was remarkable for 5-year-old children, and “Musundehiraite” was also large for the size of saccade. Regarding the total moving distance of saccade, the data of “Tewotatakimashou” for 3-year-old children were large. It was found that the size of saccade occurred in “Musundehiraite” for the first time was large in all ages of 3-year-old children, 4-year-old children, and 5-year-old children. In this way, as a result of quantitative analysis by song, the number of occurrences of saccade was large in “Tewotatakimashou” for 3-year-old children. The moving average velocity and size of saccade in “Musundehiraite” for 5-year-old children were remarkable. Regarding the size of the saccade occurred for the first time, “Musundehiraite” was large for all ages of 3-year-old children, 4-year-old children, and 5-year-old children. Those results showed that participant children involved in a spontaneous body movement accompanied to singing a nursery rhyme from the beginning of singing “Musundehiraite” during every-day life in early childhood facility. In the singing, the participant children, looking at the teacher, other children nearby, and their own hands, put their hands forward and shake them up and down accompanied to the beat. “Tewotatakimashou” is also a Japanese lyrics that encourages the movement while singing “Te-wo-ta-ta-ki-ma-sho”, and the children sung to link with the movement, it was naturally that the moving average velocity and size of saccade especially increased in 5-year-old children. The change in the numerical value of eye movement verified that as the recognition of musical elements progresses, the early childhood child consciously thinks to express his or her recognition of musical elements by the spontaneous body movement accompanied to singing. www.scholink.org/ojs/index.php/fet Frontiers in Education Technology Vol. 5, No. 2, 2022 90 Published by SCHOLINK INC. 6. Conclusion As a result of those above quantitatively analysis, it was found that the saccade in the musical expression in early childhood tended to be larger in major key than in minor key, which was remarkable in Y kindergarten, depended on the increase in the acquired data for two years. These results of this study made it possible to clarify the characteristics extracted the above quantitative analysis based on the result of analysis in 2020 depending more added data by eye-tracking. According to a result of quantitative analysis in 2020, some differences were found out depended on the two childcare forms between play- centered childcare form and a childcare form of following the Montessori method. In the results of eye- tracking in 2020, participant children involved in Montessori method tended to express the recognition of regularity and tonality regarding musical elements although participant children taking a play-centered childcare form tended to express the recognition of Japanese lyrics (Sano, 2021). By increasing more added data in 2021 regarding the number of songs and participant children, those results showed more clearly that the children tended to show characteristic change of eye movement depending on tonality and songs. From the calculated data on saccade, which was the eye movement during musical expression of early childhood children, it was verified that effective feature quantities of eye movement for machine learning could be derived in the same way as feature quantities of movement based on the results of quantitative analysis during musical expression in early childhood. In order to extract another effective feature quantity from body movement included in musical expression, simultaneous analysis of musical expression in early childhood using motion capture with eye tracker is the next research subject. Acknowledgement This work was supported by JSPS KAKENHI Grant Number 21K02369. References Ando, A., & Sumikawa, Y. (2012). Development and functional evaluation of teaching materials for observing sawing motion using motion capture and virtual space. Journal of the Japan Society for Educational Technology, 36(2),103-110. https://doi.org/10.15077/jjet.KJ00008274459 Bishopa, L., Cancino-Chacón, C., & Goebl, W. (2019). Eye gaze as a means of giving and seeking information during musical interaction. Consciousness and Cognition, 68, 73-96. https://doi.org/10.1016/j.concog.2019.01.002 Burger, B. (2013). Move the way you feel: Effects of musical features, perceived emotions, and personality on music-induced movement. Department of Music, University of Jyväskylä. Burger, B., Puupponen, A., & Tommi, J. (2017). Synchronizing eye tracking and optical motion capture:How to bring them together. Conference on Music & Eye-Tracking, 2017, Max Planck Institute for Empirical Aesthetics, 11. https://doi.org/10.16910/jemr.11.2.5 https://doi.org/10.16910/jemr.11.2.5 www.scholink.org/ojs/index.php/fet Frontiers in Education Technology Vol. 5, No. 2, 2022 91 Published by SCHOLINK INC. Dahl, S., & Friberg, A. (2007). Visual perception of expressiveness in musicians’ body movements. Music Perception, 24, 433-454. https://doi.org/10.1525/mp.2007.24.5.433 Drai-Zerbib, V., & Baccino, T. (2018). Cross-modal music integration in expert memory: Evidence from eye-movements. Journal of Eye Movement Research, 11(2), 4. https://doi.org/10.16910/jemr.11.2.4 Fernandez-Delgado, M., Cernadas, E., Barro, S., & Amorim, D. (2014). Do we need hundreds of classifiers to solve real world classification problems? Journal of Machine Learning Research, 15(90), 3133-3181. Fink, K., Lange, E., & Groner, R. (2019). The application of eye-tracking in music research. Journal of Eye Movement Research, 11(2), 1-4. https://doi.org/10.16910/jemr.11.2.1 Fusase, E. (2017). In “Childcare Expression Technology” Music I Consideration of specialized skills- line of sight when playing the piano: Through the difference in behavioral proficiency, Nagoya Yanagishiro Junior College Lab Bulletin of Research, 39, 313-331. Higuchi, D., Okumura, U., & Kobayashi, T. (2019). Development of Hiragana Character Sound Knowledge for Toddlers: Examination Using Eye Tracker (Basics of Human Communication), IEICE technical report, 118(437), 83-88. Kitajima, H., Nakatani, Y., Ogura, A., Hamada, H., & Nanami, N. (2017). Eye Movement Analysis of Japanese Sewing Scissors Craftsman. International Conference on Applied Human Factors and Ergonomics, AHFE 2017: Advances in Physical Ergonomics and Human Factors, 479-490. https://doi.org/10.1007/978-3-319-60825-9_50 Kodama, U., Oba, N., & Ishi, N. (2015). Examination of behavior recognition method in general living environment using machine learning. IEICE Technical Report, 114(502), 73-78. Kusunoki, K., Ozawa, W., & Kanemori, Y. (2017). Basic research on reading difficulty of children with foreign roots: Through visual function evaluation and reading ability measurement. Proceedings of the Japan Digital Textbook Society, 6(0), 7-8. Lin, D., Chen, G., Liu, Y., Liu, J., Pan, J., & Mo, L. (2018). Tracking the Eye Movement of Four Years Old Children Learning Chinese Words. Journal of Psycholinguistic Research, 47(1), 79-93. https://doi.org/10.1007/s10936-017-9515-x Lörch, L., Fehringer, B., & Münzer, S. (2017). Reading music. How tonality and notation influence music reading experts’ eye movements, pupil dilation and performance in a pattern matching task. Conference on Music & Eye-Tracking, 2017, Max Planck Institute for Empirical Aesthetics, 17. Marandola, F. (2017). Eye-Hand synchronisation and interpersonal interaction in xylophone performance: A comparison between African and Western percussionists. Conference on Music & Eye-Tracking, 2017, Max Planck Institute for Empirical Aesthetics, 18. Matumoto, A., Mikami, H., Kawamura, H., & Kojima, A. (2014). Examination of form classification method for supporting motor learning of elementary school students who performed motion analysis https://doi.org/10.16910/jemr.11.2.4 https://doi.org/10.16910/jemr.11.2.1 https://doi.org/10.1007/978-3-319-60825-9_50 https://doi.org/10.1007/s10936-017-9515-x www.scholink.org/ojs/index.php/fet Frontiers in Education Technology Vol. 5, No. 2, 2022 92 Published by SCHOLINK INC. using image processing and machine learning techniques, Video Information Media Society Technical Report, 38, 9-12. Moreno-Estevaa, E., White, S., Woodc, J., & Blackc, A. (2019). Application of mathematical and machine learning techniques to analyze eye tracking data enabling better understanding of children’s visual cognitive behaviours. Frontline Learning Research, 6(3), 72-84. https://doi.org/10.14786/flr.v6i3.365 Mpofu, B. (2016). University Students Use of Computers and Mobile Devices for Learning and their Reading Speed on Different Platforms. Universal Journal of Educational Research, 4(4), 926-932. https://doi.org/10.13189/ujer.2016.040430 Orman, K. (2016). Effect of Virtual Reality Exposure and Aural Stimuli on Eye Contact, Directional Focus, and Focus of Attention of Novice Wind Band Conductors. International Journal of Music Education, 34(3), 263-270. https://doi.org/10.1177/0255761415619058 Orman, K., Price, H., & Russell, R. (2017). Feasibility of Using an Augmented Immersive Virtual Reality Learning Environment to Enhance Music Conducting Skills. Journal of Music Teacher Education, 27(1), 24-35. https://doi.org/10.1177/1057083717697962 Plöchl, M., & Obleser, J. (2017). Do auditory rhythms influence eye movement statistics? Conference on Music & Eye-Tracking, 2017, Max Planck Institute for Empirical Aesthetics, 19. Puurtinen, M. (2018). Learning on the job: Rethinks and realizations about eye tracking in music-reading studies. Frontline Learning Research, 6(3), 148-161. https://doi.org/10.14786/flr.v6i3.380 Remount, F., & Budke, A. (2020). Strategies for Successful Learning with Geographical Comics: An Eye-Tracking Study with Young Learners. Education Sciences, 10, 293. https://doi.org/10.3390/educsci10100293 Sano, M. (2018). Development of a quantitative methodology to analyze the growth of recognition of musical elements in early childhood from a viewpoint of change of body movement. Asia-Pacific Journal of Research in Early Childhood Education, 12(1), 61-80. https://doi.org/10.17206/apjrece.2018.12.1.61 Sano, M. (2019). Predicting developmental degrees of music expression in early childhood by machine learning classifiers with 3D motion captured body movement data. Journal of Educational Research and Reviews (International), 7(7), 155-168. https://doi.org/10.33495/jerr_v7i7.19.128 Sano, M. (2020). Verification of a classification prediction method for the development of musical expression in early childhood using a machine learning method based on 3D motion capture data. Advances in Social Sciences Research Journal (International), 7(9), 338-358. https://doi.org/10.14738/assrj.79.9028 Sano, M. (2021). Quantitative Analysis of Eye Movements during Singing of Early Childhood Children. Research Journal of Education, 7(3), 125-140. https://doi.org/10.32861/rje.73.125.140 https://doi.org/10.1177/0255761415619058 https://doi.org/10.1177/1057083717697962 https://doi.org/10.14786/flr.v6i3.380 https://doi.org/10.3390/educsci10100293 https://doi.org/10.17206/apjrece.2018.12.1.61 https://doi.org/10.33495/jerr_v7i7.19.128 https://doi.org/10.14738/assrj.79.9028 https://doi.org/10.32861/rje.73.125.140 www.scholink.org/ojs/index.php/fet Frontiers in Education Technology Vol. 5, No. 2, 2022 93 Published by SCHOLINK INC. Sato, K., Kaiga, T., & Watabe, S. (2010). Utilization of motion capture to support the mastery of dance. Journal of the Japan Society for Educational Technology, 34, 133-136. Seong-un Kim, Sung-man Lim, Eun-ae Kim, & Il-ho Yang. (2016). An Analysis of Eye Movement and Cognitive Load about the Editorial Design in Elementary Science Textbook. Universal Journal of Educational Research, 4(3), 570-575. https://doi.org/10.13189/ujer.2016.040314 Takada, K., Kitasuka, T., & Aritugi, M. (2012). Examination of personal identification method by gait using markerless motion capture device. Research Report Entertainment Computing (EC), 2012, 1- 7. Thompson, M., & Luck, G. (2012). Exploring relationships between pianists’ body movements, their expressive intentions, and structural elements of the music. Musicae Scientiae, 16, 19-40. https://doi.org/10.1177/1029864911423457 Valtakari, N., Hooge1, I., Viktorsson, C., Nyström, P., Falck-Ytter, T., & Hessels, R. (2021). Eye tracking in human interaction: Possibilities and limitations. Behavior Research Methods, 53, 1592-1608. https://doi.org/10.3758/s13428-020-01517-x Vandemoortele, S., Feyaerts, K., Reybrouck, M., De Bièvre, G., Brône, G., & De Baets, T. (2018). Gazing at the partner in musical trios: A mobile eye-tracking study. Journal of Eye Movement Research, 11(2), 6. https://doi.org/10.16910/jemr.11.2.6 Watabe, M., Matumura, A., & Uda, N. (2019). An attempt to extract children’s tastes from gaze movements and facial expressions in storytelling of picture books. Proceedings of the 81st National Convention, 2019(1), 563-564. https://doi.org/10.3758/s13428-020-01517-x https://doi.org/10.16910/jemr.11.2.6