Read Writ (2017) 30:1447–1472

DOI 10.1007/s11145-017-9732-6

Eye-movement patterns and reader characteristics

of students with good and poor performance
when reading scientific text with diagrams

Yu-Cin Jian1

Published online: 3 March 2017

 Springer Science+Business Media Dordrecht 2017

Abstract This study investigated the cognitive processes and reader characteristics
of sixth graders who had good and poor performance when reading scientific text
with diagrams. We first measured the reading ability and reading self-efficacy of
sixth-grade participants, and then recorded their eye movements while they were
reading an illustrated scientific text and scored their answers to content-related
questions. Finally, the participants evaluated the difficulty of the article, the
attractiveness of the content and diagram, and their learning performance. The
participants were then classified into groups based on how many correct responses
they gave to questions related to reading. The results showed that readers with good
performance had better character recognition ability and reading self-efficacy, were
more attracted to the diagrams, and had higher self-evaluated learning levels than
the readers with poor performance did. Eye-movement data indicated that readers
with good performance spent significantly more reading time on the whole article,
the text section, and the diagram section than the readers with poor performance did.
Interestingly, readers with good performance had significantly longer mean fixation
duration on the diagrams than readers with poor performance did; further, readers
with good performance made more saccades between the text and the diagrams.
Additionally, sequential analysis of eye movements showed that readers with good
performance preferred to observe the diagram rather than the text after reading the
title, but this tendency was not present in readers with poor performance. In sum,
using eye-tracking technology and several reading tests and questionnaires, we
found that various cognitive aspects (reading strategy, diagram utilization) and
affective aspects (reading self-efficacy, article likeness, diagram attraction, and self-
evaluation of learning) affected sixth graders’ reading performance in this study.

& Yu-Cin Jian

[email protected]
1
Department of Educational Psychology and Counseling, National Taiwan Normal University,
No. 162, HePing East Road, Section 1, Taipei, Taiwan

123
1448 Y.-C. Jian

Keywords Science reading
Cognitive process
Eye movements
Reader

characteristics
Sequential analysis

Introduction

Primary school students often read illustrated texts in textbooks and in their daily
lives, especially in the scientific domain. To successfully acquire knowledge while
reading illustrated texts, the reader needs to decode, organize, and integrate the
multiple representations of verbal and graphical information (Mayer, 2005; Schnotz
& Bannert, 2003), which is a complicated cognitive activity; however, not all
students are capable of learning successfully during reading.
Diagrams play an important role in scientific articles (Ainsworth, 1999; Cook,
2006; Ferk, Vrtacnik, Blejec, & Gril, 2003; Slough & McTigue, 2010).
Diagram literacy means that the reader has the ability to decode and interpret the
information that is presented in the diagram (McTigue & Flowers, 2011). McTigue
and Flowers posited that children have poor diagram literacy because primary
schools usually do not teach students how to either decode the information (e.g., the
shapes and spatial positions of the components, the part-and-whole relationship,
arrow meanings) that is presented in diagrams in scientific articles, or to use
diagram reading strategies (e.g., referring to relevant information in text and
diagram sections). In addition, young children have limited experience in reading
scientific content, and diagram literacy is not necessarily improved if scientific
knowledge is increased. Rather, it seems to be a cognitive behavior that develops
slowly (Kress & van Leeuwen, 1996; Moore & Scevak, 1997).
Previous studies have used multiple methodologies to investigate text and
diagram reading in young students. Each methodology elucidated a distinct feature
of reading. Some studies used think-aloud protocols, in which the participants self-
reported their thoughts as they were reading (Moore & Scevak, 1997; Norman,
2012). Others used outcome measures, in which the participants were asked to
complete tests after reading scientific text with or without diagrams (McTigue,
2009; Rusted & Coltheart, 1979; Segers et al., 2008; Small, Lovett, & Scher, 1993).
Recently, eye tracking has been used to investigate the cognitive processes of young
readers while they read illustrated science texts. This method involves recording eye
movements during reading to investigate individual cognitive processes (Hannus &
Hyönä, 1999; Jian, 2016; Mason, Pluchino, Tornatora, & Ariasi, 2013a; Mason,
Tornatora, & Pluchino, 2013b; Mason, Tornatora, & Pluchino, 2015). However, the
findings obtained in these studies are inconsistent and still controversial. Several
eye-tracking studies have reported that better learning performance is associated
with greater integrative processing of textual and graphical information (i.e., readers
made more saccades between text and diagram sections in articles). This
phenomenon was observed in not only fourth graders (Hannus & Hyönä, 1999;
Mason et al., 2013b), but also seventh (Mason et al., 2015) and eleventh graders
(Mason et al., 2013a). However, Norman (2012) did not verify this positive
correlation using the think-aloud protocol. We speculate that this might be due to

123
Eye-movement patterns and reader characteristics of… 1449

the different methodologies used. Although the think-aloud protocol provides

sufficient information about thinking while the participants are reading, it hampers
the process of reading and makes it unnatural. Furthermore, this method only
reflects conscious reading processes, and subconscious behavior is not measured.
Therefore, these experiments might underestimate the usefulness of graphical
information.
Methodologically, eye-tracking is a very useful technique for reading research. It
can be used to measure these reading processes objectively and immediately
(Rayner, 1998). In the past decade, eye-tracking research has been conducted to
understand the process of reading in young readers, especially in relation to
illustrated scientific text (Hannus & Hyönä, 1999; Jian, 2016; Mason et al.,
2013a, b, 2015). Additionally, the analysis of eye-tracking has significantly
progressed; the sequential analysis (Bakeman & Gottman, 1997) of eye movements
can now reveal the dynamics of reading. By combining this with the fixation time
data that previous eye-movement researchers (Hannus & Hyönä, 1999; Jian, 2016;
Jian, Wu, & Su, 2014; Mason et al., 2015) have often used, we can now study the
entire reading process.
Several studies have investigated the cognitive processes of young readers while
they read illustrated text, and most of these studies have classified participants
according to their reading ability (Harber, 1983; McTigue, 2009; Norman, 2012;
Reid & Beveridge, 1986) or intelligence (Hannus & Hyönä, 1999). Until recently,
few studies had examined groups formed according to the learning outcomes of
participants (Mason et al., 2013b). If the purpose of reading is to learn, then learning
outcomes after reading are crucial. These outcomes are determined by both
cognitive (such as reading strategy, prior knowledge, number of characters known,
and comprehension skill) and affective (reading self-efficacy, reading interest,
reading engagement, self-evaluation of learning) factors. Learning outcomes are not
necessarily restricted to the result of the readers’ reading ability. Therefore, in this
study, we investigated differences in young readers’ cognitive and affective factors,
and recorded the eye-movement patterns of those who performed well or badly on a
reading task. Knowing how young students read illustrated text not only helps
primary school teachers to understand their students in various reading-related
situations, but also allows them to use appropriate instructions to help students to
develop adequate reading strategies for the future.

Illustrated text reading in young readers

Reading of illustrated text has been investigated in recent years. Two famous
theories have been used to explain the potential beneficial effects of multimedia
materials: the cognitive theory of multimedia learning (Mayer, 2005) and the
integrated model of text and diagram comprehension (Schnotz & Bannert, 2003).
Both are based on the dual coding theory of Paivio (1990), which suggests that
verbal and pictorial information are processed by different subsystems in the human
brain to generate verbal and pictorial models; connecting these two models results
in a better mental model compared to the use of a single textual model. This theory

123
1450 Y.-C. Jian

implies that readers who read text with diagrams will have better learning
performance than those who read the same text without diagrams.
In previous studies, researchers used learning outcomes (McTigue, 2009; Rusted
& Coltheart, 1979; Segers et al., 2008; Small et al., 1993) to investigate whether
diagrams facilitate reading comprehension. The typical method was to use reading
materials with or without diagrams, and researchers then compared the test scores of
different groups. Although multimedia learning was found to be very effective in
adults, its effect was variable in young readers. Several studies have indicated that
students learn better when they read texts with diagrams (Rusted & Coltheart, 1979;
Small et al., 1993), but others showed that multiple representations (scientific text
and pictures) did not facilitate comprehension in young readers (McTigue, 2009;
Segers et al., 2008).
Several studies have used think-aloud protocols, asking young readers to report
their thoughts as they read. Moore and Scevak (1997) invited fifth-, seventh-, and
ninth-grade students to read illustrated scientific texts, and verbally report their
thoughts. Their results showed developmental differences in the reading strategies
used to understand these articles. Ninth-grade students applied diverse reading
strategies and were capable of using diagrammatic information that was relevant to
the textual information. However, younger readers rarely used diagrammatic
information, instead focusing on the detailed information presented in the text. Only
8% of fifth-grade students and 13% of seventh-grade students used the information
shown in the diagrams. Using the think-aloud protocol, Norman (2012) asked
second graders to read scientific texts with several illustrations (including
photographs, realistic drawings, captions, labels, and diagrams). As the students
read, they were instructed to think aloud about the text when they looked at the
illustration information. Then, they completed a retelling and comprehension test.
The findings showed that the number of times second graders used diagram-reading
strategies was correlated with retelling measures, but not with reading comprehen-
sion performance.

Eye-movement research on multimedia learning for children

In the past decade, eye-tracking technology has been used extensively to investigate
the cognitive processes of young readers reading illustrated scientific text. These
studies have employed not only comprehension test measures but also eye
movement data to investigate the relationship between learning outcomes and
cognitive processes.
Hannus and Hyönä (1999) investigated how illustrations guide the attention of
learners during text reading. They used eye-tracking technology to examine fourth-
grade students who had good and poor abilities in relation to reading scientific
textbook passages with several illustrations (including diagrams, tables, and
photographs). The results showed that students with good abilities spent more time
on reading the text, but this result was not statistically significant, because of the
large amount of variance. However, students with good and poor reading abilities
viewed the illustrations for a similar period of time. Surprisingly, only 6% of both
groups’ total reading time was spent on viewing the illustrations. To analyze the

123
Eye-movement patterns and reader characteristics of… 1451

transfer between the text and illustration, the authors created a five-point scale
(5 = very often to 1 = never) for two raters to assess whether the participants
looked back and forth between the illustration and text sections. The results showed
that students with good reading abilities looked back and forth more often than
students with poor reading abilities did. However, these scores were relatively low
for both groups: average ratings were 2.6 and 2.0 for students with good and poor
abilities, respectively. Additionally, Hannus and Hyönä did not find a significant
correlation between students’ comprehension, based on questions about the
illustrations, and the amount of time spent inspecting the illustrations. However, a
significant positive correlation was revealed between successfully answering
illustration-related questions and the amount of time spent reading the text section
in the scientific article. These findings imply that the young readers engaged in text-
driven reading, and did not pay much attention to the illustrations.
In recent years, Mason et al. (2013a, b, 2015) have carried out a series of
pioneering experiments using eye movement tracking to examine how young
students read illustrated text. A common characteristic of these studies is the use of
cluster analysis of the eye movements of the participants to examine reading
patterns in young readers, and to understand the association between cognitive
processes (measured by eye movement tracking) and learning outcomes (measured
by reading tests). For example, Mason et al. (2013a, b) recruited fourth graders to
read an illustrated scientific text while their eye movements were recorded. Cluster
analysis of eye movements was used to classify the participants into low,
intermediate, and high integrators of text and diagram. These authors found that
higher integrative processing of the illustrated text was associated with better
learning performance. Additionally, 47% of the participants were classified into the
high integrator group, whereas only 14% were low integrators. These data imply
that most fourth graders have the ability to use the information presented in
diagrams because the participants tried to connect this with the information in the
text, although it is unclear if they integrated both semantic representations
successfully. However, the previous studies (Hannus & Hyönä, 1999; Jian, 2016)
reported that most fourth graders did not check the relevant text and diagram in
parallel when reading illustrated scientific texts.
Mason et al. (2015), in an extended study, recruited seventh graders to read an
illustrated scientific article while their eye movements were recorded. They
performed cluster analysis and a series of hierarchical regression analyses to reveal
which eye movement indicator might predict learning outcomes most effectively.
The results indicated that only the patterns of integrative processing of text and
diagram (calculated as the number of saccades between text and diagram) during
second-pass reading (a less automatic and more targeted process) predicted learning
outcomes, after controlling for individual differences in background knowledge,
reading comprehension ability, and self-concept, regardless of whether the tests
measured factual recall or the transfer of knowledge.
Reading is a dynamic, rather than static, process; therefore, sequential analysis is
an ideal method to investigate the process of reading. Sequential analysis has
usually been used in the past by psychologists to reveal interactions between people
or to analyze moment-to-moment behavior (Bakeman & Gottman, 1997). A few

123
1452 Y.-C. Jian

recent studies have also used this analysis to investigate the reading strategies of
adults (Jian et al., 2014), adolescent readers (Cook, Carter, & Wiebe, 2008a, b), and
young readers (Jian, 2016). Jian investigated the reading strategies of young and
adult readers when reading illustrated text. The researcher asked fourth-grade
students with good reading abilities and undergraduate students to read a biological
article with two diagrams, which was aimed at elementary school-level readers,
from a scientific textbook. Eye movements in the sequential analysis showed that
adult readers demonstrated bidirectional reading pathways for both text and
diagrams, whereas children’s eye fixations only went back and forth within
paragraphs in the text and between the diagrams, and that the latter group made
fewer references to both the text and diagrams. This finding suggests that although
the fourth-grade children had good reading abilities, their visual literacy is not
advanced enough to connect corresponding features among different representa-
tions, which is crucial in reading comprehension.

Affective factors in reading performance

From the perspective of motivation theory, several affective factors (e.g., reading
self-efficacy, reading interest, reading engagement, self-evaluation of learning) may
also influence reading performance.
Self-efficacy entails an individual’s belief in his or her capacity to perform at a
desired level (Bandura & Cervone, 1983), and may influence the individual’s choice
of what to do and how much effort to expend on a given activity (Schunk, 1981).
Research has also indicated that learners’ self-efficacy is positively related to
reading comprehension (Katzir, Lesaux, & Kim, 2009). As for reading interest,
research on its relationship with reading ability remains inconclusive. A positive
correlation between the two has been reported (Baker & Wigfield, 1999; Taboada,
Tonks, Wigfield, & Guthrie, 2009), but some researchers have reported only a weak
correlation in the early elementary years (Kirby, Ball, Geier, Parrila, & Wade-
Woolley, 2011). Reading engagement relates to the extent to which a person
engages in a reading activity, such as encoding a series of words or showing reading
comprehension, as well as reading strategies applied. Cognitive dimensions of a
reading engagement index involve working hard to read and thinking deeply about
the content (Unrau & Quirk, 2014; Wigfield et al., 2008). Reading engagement has
been shown to have an impact on reading and learning outcomes (Guthrie &
Wigfield, 2000). Finally, self-evaluation of learning entails an internal comparison
between an individual’s personal standards and his or her performance (Bandura &
Cervone, 1983). To the best of my knowledge, the relationship between self-
evaluation and reading has not been investigated so far.

The present study

The theory of multimedia learning (Mayer, 2005) clearly elaborates the cognitive
processes of illustrated text reading. Successful reading comprehension requires the
reader to decode, organize, and integrate multiple representations of verbal and
pictorial information during reading. However, reader characteristics and affective

123
Eye-movement patterns and reader characteristics of… 1453

factors are not mentioned in this theory, despite their possibility of being influential
variables in multimedia learning. To further extend and contribute to the theory on
multimedia learning, this study used eye-tracking technology in combination with
sequential analysis, as well as reading tests and questionnaires, to investigate the
cognitive processes and reader characteristics of sixth graders with good or poor
performance when reading an illustrated scientific text.
From a practical perspective, although the relationship of reading comprehension
and integrative processing behavior of textual and graphical information is positive
for adult readers, the relationship is unstable and remains controversial for young
readers. Some research has shown support for a positive relationship (Hannus &
Hyönä, 1999; Mason et al., 2013b), but other studies have not (McTigue, 2009;
Norman, 2012). A possible answer to this controversial issue was another important
reason motivating the design of this study.
To investigate the cognitive processes of illustrated text reading among young
readers who differed in terms of reading performance, several eye-movement
indicators were selected based on previous studies that investigated the reading of
diagrams or illustrated texts; each indicator reflects different types of cognitive
processes (Grant & Spivey, 2003; Hannus & Hyönä, 1999; Hegarty, 1992; Jian,
2016; Johnson & Mayer, 2012). First, total reading time (the sum of all fixation
durations on an area of interest [AOI]) indicates the degree of cognitive effort
necessary for processing the information. According to dual-processing theories of
cognition (Evans, 2007), deep processing requires a large amount of mental
resources. In addition, readers have more cognitive demands when interpreting
diagrams during science text reading (Cromley, Snyder-Hogan, & Luciw-Dubas,
2010). Therefore, we hypothesized that readers with good reading comprehension
would spend longer on the reading material, not only for the text section but also for
the diagram section. Second, the proportion of fixation durations (the fixation
duration on a specific AOI divided by the total fixation duration during the learning
episode) was measured, which reflects the selective attentional focus on specific
target regions during learning. According to previous empirical research (Hannus &
Hyönä, 1999; Jian, 2016), young readers of both high- and low-ability spent a
substantial proportion of fixation durations on text sections rather than on diagram
sections in illustrated texts. Therefore, we hypothesized both groups, with high- and
low-reading performance, would showed similar patterns of fixation proportions as
found in previous research. Third, the mean fixation duration (average duration of
all fixations on an AOI) was calculated, which reflects the amount of time readers
need to decode a certain stimulus, such as words or diagrams. Young readers usually
have poor diagram literacy (McTigue & Flowers, 2011). When reading material is
difficult, readers with good comprehension need to strategically adjust the allocation
of their cognitive resources to achieve diagram comprehension, and thus slow down
their eye movements to give themselves time to complete this process (Miller,
2015). Therefore, we hypothesized that readers with good reading comprehension
would show longer mean fixation duration on the diagrams than those with poor
reading performance. Fourth, the number of saccades from the text to the diagram
(the number of times the participant changed their eye fixation from the text to the
illustration) was examined, which reflects the inference and integration of textual

123
1454 Y.-C. Jian

and pictorial information. Based on the cognitive theory of multimedia learning

(Mayer, 2005) and the integrated model of text and diagram comprehension
(Schnotz & Bannert, 2003), connecting relevant information in the text and picture
is essential for constructing a good mental model of illustrated text reading.
Therefore, we hypothesized that readers with good reading comprehension would
show more transition behavior, looking between the text and diagrams, than would
those with poor reading performance.
To investigate the characteristics of young readers who differed in reading
performance, several reading tests and questionnaires were used in this study. As
mentioned in the literature review above, learners’ self-efficacy has been found to
be positively related to reading comprehension (Katzir et al., 2009). We
hypothesized that readers with good reading comprehension would have better
reading self-efficacy than those with poor reading performance. As for reading
interest, research on its relationship with reading ability remains inconclusive
(Baker & Wigfield, 1999; Kirby et al., 2011; Taboada et al., 2009), and to our
knowledge, the relationship between self-evaluation and reading has not been
investigated thus far. Therefore, we had no specific hypotheses regarding the
degrees of article (and diagram) likeness and self-evaluation of learning of the good
and poor readers in this study.

Method

Participants

Forty-two students (22 girls and 20 boys) in grade 6 from an elementary school in
Taiwan participated in the experiment. Parental consent was sought and students
were rewarded for their participation with stationery. The average age of the
students was 12.3 years (range 11.7–12.9 years; SD = 3.2 months). None of the
participants had difficulties with character recognition, according to their scores on
the standard Chinese character recognition test (Huang, 2001), which evaluates the
number of characters known. On this test, students write 200 Chinese characters in
pinyin. The participants’ average number of Chinese characters known met the 8.1
grade norm for character recognition (range 5.1–10.0 grade level; SD = 1.65 grade
level). All participants had normal or corrected-to-normal vision.

Materials

We provided an illustrated text for the students to read. The article (see Fig. 1) was
rewritten from a seventh-year science textbook used in Taiwan (Lin et al., 2008). Its
topic was respiration and gas exchange. The article consisted of a title, text section,
and diagram section. The text section had 439 Chinese characters in three
paragraphs: the first briefly defined respiration; the second described the process of
gas exchange during cellular respiration, including examples from several animals;
and the third described the respiratory movement of the human body. The diagram
section showed two diagrams. The upper one was from the original science textbook

123
Eye-movement patterns and reader characteristics of… 1455

Title

Paragraph 1

Upper diagram

Paragraph 2

Bottom diagram

Paragraph 3

Fig. 1 Six AOIs (title, paragraph 1, paragraph 2, paragraph 3, upper diagram, and bottom diagram) of
the reading material. The participants in this study did not see these black frames

(Lin et al., 2008), and it depicted gas exchange through the skin of animals, which
was related to Paragraph 2 of the text. The lower diagram was obtained from a
website (http://www.phyworld.idv.tw), and it depicted the processes of inhalation
and exhalation during respiration in humans, which was related to Paragraph 3 of
the text.
Three experts verified that the text was suitable for sixth-grade children regarding
its difficulty and readability. The first was a professor who was an expert in reading
psychology; the second was a PhD candidate in science education; and the third was
a science teacher from an elementary school, with a master’s degree in science
education. We selected the reading material from a textbook written for seventh-
grade students rather than from a textbook for sixth-grade students to ensure novelty
of the information, which allowed us to investigate how and what students learned
while they were reading by themselves. The reading material was displayed on a
single screen, and there were no scroll bars or additional pages.

Apparatus

Eye movements of the participants were recorded using Eyelink 1000 at a sampling
rate of 1000 Hz. A chin bar was used to minimize head movements. The viewing
was binocular, but eye movements were recorded from the right eye only. Eye
movements were calibrated and validated until the average error in gaze position
was \0.5 of the visual angle. The reading material was presented on a 24-in. LCD
monitor with a resolution of 1920 9 1200 pixels. The distance between the monitor
and participants was 65 cm. The stimuli on the screen covered 46 of the horizontal
visual angle and 30 of the vertical visual angle.

123
1456 Y.-C. Jian

Procedure

This study had three sessions. In the first session, all participants completed the
standard reading comprehension screening test (Ko, 1999) and the standard Chinese
character recognition test (Huang, 2001). These tests were performed in the
classroom, and the procedure lasted for approximately 50 min.
The second session involved the eye movement tracking experiment, which
was executed 1 week later. Participants were tested individually in a quiet room
in the elementary school, where they were instructed to complete a reading self-
efficacy questionnaire (PIRLS, 2011). The questionnaire contained seven items,
such as ‘‘Reading is easy for me’’ and ‘‘I usually do well in reading.’’ Students
answered each question using a four-point scale ranging from ‘‘strongly agree’’
(1 point) to ‘‘strongly disagree’’ (4 points). Among the seven items, two tested
the lack of self-efficacy with a reversed meaning. For these, the answers were
scored in reverse (e.g., 4 points was scored as 1 point). Then, the participants
read the experimental material to test their comprehension. The experimenter
asked the participants to press the space bar on a keyboard when they finished
reading to initiate the reading test. There was no time limit set for the reading
procedure, in order to simulate natural reading conditions. Therefore, the
participants could read at their own pace. Participants first read one practice
article and answered several reading-related questions to learn the experimental
procedure. After participants indicated that they understood the procedure, the
formal experiment was executed. A 12-point calibration and validation procedure
was performed for each participant. The experimenter asked the participants to
keep their heads still throughout the experimental procedure. When the
participants had finished reading, they immediately completed the reading
comprehension test, which contained 15 yes/no questions, including five text-
based questions, five diagram-based questions, and five integrated questions
involving text and diagrams, to investigate their comprehension ability. The
questions appeared on the screen, one at a time, and then 11 essay questions
were answered on a paper sheet. The essay questions included five ‘‘which’’
questions (e.g., ‘‘Which body components of an earthworm conduct respira-
tion?’’), three ‘‘how’’ questions (e.g., ‘‘How does an organism conduct
respiration? Please elaborate on the processes’’), and three ‘‘why’’ questions
(e.g., ‘‘Why does people’s breathing becomes fast when doing strenuous
exercise?’’). The comprehension test had no time limit. The experimental
reading session lasted for approximately 20–30 min.
In the third session, we used questionnaires to measure the participants’
familiarity with the topic and their subjective opinions about the reading
material, immediately after the eye movement tracking experiment was complete.
First, the experimenter instructed the participants to press a key on the keyboard
to indicate whether they had known the meanings of a technical term before
reading the scientific article. There were 10 technical terms in the experimental
material. If the student knew the meaning of the word, 1 point was awarded;
otherwise, 0 points were awarded. One technical term was displayed at a time on
the computer screen. Finally, participants completed four questions to evaluate

123
Eye-movement patterns and reader characteristics of… 1457

the difficulty of the article, the attractiveness of the article and the diagram, and
their own learning performance, using a five-point response scale. All rating
tasks were executed on the computer screen, and the procedure lasted for
approximately 3–5 min.

Data selection, scoring, and grouping criteria

We collected different data to understand the characteristics and opinions of readers,

as well as their cognitive processes and reading performance when reading
illustrated text. The former data included a standard Chinese character recognition
test, a standard reading screening comprehension test, a self-efficacy questionnaire,
a technical vocabulary questionnaire, and an article-rating questionnaire. The latter
data included readers’ eye movements and a comprehension test after reading the
text.

Characteristics and opinions of readers

The data related to readers’ characteristics and opinions included the following:
(1) the standard Chinese character recognition test contained 200 Chinese
characters; each correct answer for writing a pinyin character was awarded 1
point. Higher scores indicate knowledge of a larger number of characters; (2) the
standard reading comprehension screening test comprised 20 multiple-choice
questions. An example is Father and mother decide to go to Kenting National
Park to go on a honeymoon once again. Which word indicates father and mother
has went on a honeymoon? Decide/go to/again/once. Another example is
Knowing him finally resolved the annoying event, I hastily dry my tears. What is
my emotion now? Sad/happy/angry/worried. Each correct answer was awarded 1
point. Higher scores indicate better reading comprehension ability; (3) the self-
efficacy questionnaire contained seven items, with a lower score indicating
higher reading self-efficacy, and a higher score indicating lower reading self-
efficacy; (4) the vocabulary questionnaire contained 10 scientific technical terms
from the reading material. A higher percentage of known terms indicates more
familiarity with the topic; and (5) the article-rating questionnaire contained four
questions, and self-evaluations of learning performance, which were made on a
five-point scale from ‘‘very easy/attractive/attractive/good performance’’ (1 point)
to ‘‘very difficult/unattractive/unattractive/bad performance’’ (5 points).

Eye movements

Eye movement data from four participants were excluded due to unsuccessful eye-
tracking (two participants) or apparent drift (two participants). Unsuccessful
recording occurred due to data transmission failure, and apparent drift occurred
when fixations were almost entirely on the blank space, rather than on the text or
illustrations. Therefore, the data from 38 participants were included in the analysis.
In addition, similar to the procedure used in previous eye-movement studies

123
1458 Y.-C. Jian

(Andrews et al., 2004; Jian, 2016), fixations shorter than 100 ms were excluded,
which comprised approximately 5% of the data.
In addition to these eye movement indicators mentioned above, we examined the
sequence of eye fixations to investigate the reading strategies adopted by the
participants. A series of matrix calculations was carried out for this analysis. First,
first-pass transitions (the proportion of first fixations transferred from the initial AOI
to the subsequent AOI) were investigated, reflecting the initial processing of target
AOIs. For example, if 20 readers first read A-AOI and 10 of them made their next
fixations on B-AOI, the transition percentage of first-pass sequences from A-AOI to
B-AOI would be 0.50. Second, total-pass transitions (the proportion of total
fixations transferred from the initial AOI to the subsequent AOI, which included
first-pass and rereading-pass reading types) were calculated, reflecting late
processing and higher-order cognitive processing during reading. For example, if
A-AOI was read 80 times by 20 readers, and there were 20 transfers to B-AOI, the
transition percentage of total-pass sequences from A-AOI to B-AOI would be 0.25.
This sequential analysis technique is frequently used to investigate moment-to-
moment behavioral sequences (Bakeman & Gottman, 1997), and it has been
recently used to analyze eye movement data (Cook et al., 2008a; Jian, 2016; Jian
et al., 2014).

Comprehension test after reading the illustrated text

The comprehension test included yes/no questions and essay questions. For the yes/
no questions, each correct answer was awarded 1 point. For essay questions, the
answers were scored by two independent raters who were blind to the purpose of the
study. Inter-rater reliability, which was evaluated with the Cohen’s kappa
coefficient for each essay question, ranged from 0.90 to 1.00. Each rating
disagreement was carefully examined and discussed by the two raters until a
consensus was reached. The comprehension test score was calculated as the sum of
scores for both question types.
To reveal differences in the characteristics and eye movements of readers with
good and poor performance, the participants were divided into two groups
according to their comprehension test scores after reading the illustrated text. We
sorted all scores of the comprehension test from the highest to the lowest. The
first half of all scores were regarded as reflecting good performance, and the latter
half poor performance. Each group had 19 participants. Mean scores for the good
and poor performance groups were 18.21 (SD = 4.54) and 9.42 (SD = 3.30),
respectively.

Results

The dependent variables were the characteristics of readers (scores on the Chinese
character recognition test, reading comprehension screening test, and self-efficacy
questionnaire) and eye movement indicators (total reading time, the proportion of

123
Eye-movement patterns and reader characteristics of… 1459

fixation durations, mean fixation duration, the frequency of saccade from the text to
the diagrams, and the sequence of eye fixations).

Characteristics of readers

The upper section of Table 1 shows that students with good performance had
significantly better Chinese character recognition test scores, t(36) = 2.67, p \ 0.05,
d = 0.87, and better reading self-efficacy, t(36) = -3.73, p \ 0.01, d = -1.21, than
students in the poorly performing group. However, both groups achieved similar
scores on the standard reading comprehension screening test (p [ 0.05).
The lower section of Table 1 shows that students with good performance rated
the diagrams as significantly more attractive, t(36) = -2.68, p \ 0.05, d = -0.87,
had higher self-evaluated learning, t(36) = -3.96, p \ 0.001, d = -1.27, and
found the article marginally more attractive, t(36) = -1.76, p = 0.088, d = -0.57,
than poorly performing students. However, technical vocabularies and ratings of the
difficulty of the article were similar in both groups (ps [ 0.05).

Eye movement analysis

To investigate processing differences between the students with good and poor
performance when reading the illustrated text, we examined various eye movement
indicators, including visual attention distribution between the text and diagrams,
reference to text and diagrams, and reading pathways. In this regard, according to
the Simple View of Reading (Hoover & Gough, 1990), word recognition is an
important component of reading comprehension. To avoid the findings for eye
movements in the present study being potentially ascribed to group discrepancies in
word recognition skills, the scores for a Chinese character recognition test were
used for a test of covariance, using ANCOVA for statistical analysis.

Table 1 Reading self-efficacy, Chinese character recognition test, and reading comprehension test for
good- and poor-performance groups
Good-performance Poor-performance t-value
group group
M (SD) M (SD)

Chinese character recognition test 135.84 (18.95) 120.53 (16.26) 2.67*

Reading comprehension screening test 0.86 (0.10) 0.82 (0.16) 1.04
Reading self-efficacy 12.26 (2.49) 15.26 (2.47) -3.73**
Know technical vocabularies 0.48 (0.13) 0.45 (0.18) 0.51
Article difficulty rating 2.79 (0.71) 3.16 (0.69) -1.62
Article likeness rating 1.84 (0.96) 2.42 (1.07) -1.76 
Diagram attraction rating 1.84 (0.83) 2.74 (1.20) -2.68*
Self evaluation of learning 2.11 (0.46) 2.89 (0.74) -3.96**
 
* p \ 0.05; ** p \ 0.01; *** p \ 0.001; p \ 0.10

123
1460 Y.-C. Jian

Whole article and detailed component analyses

We first studied the entire article as one AOI and then divided it into three AOIs:
title, text, and diagram. The means and standard deviations from the eye-movement
analysis are presented in Table 2.
Table 2 shows that the total reading time was significantly longer for students
with good performance than for those with poor performance for both the whole
article, F(1, 36) = 13.08, p \ 0.01, g2 = 0.27, the text section, F(1, 36) = 10.18,
p \ 0.01, g2 = 0.23, and the diagram section, F(1, 36) = 10.67, p \ 0.01,
g2 = 0.23. The fixation duration proportions on the text or diagram sections were
not significantly different between the groups (p [ 0.05). It is worth noting that the
mean fixation duration on the diagram was significantly longer for students with
good performance than for those with poor performance, F(1, 36) = 8.71, p \ 0.01,
g2 = 0.20. In addition, the group with good performance made significantly more
saccades from the text to the diagrams than the group with poor performance did,
F(1, 36) = 9.02, p \ 0.01, g2 = 0.21.

Analysis of eye fixations sequences

In order to examine cognitive processes and reading strategies in both groups, we

carried out a series of sequential matrix calculations (Bakeman & Gottman, 1997)
for eye fixations. The time-sequential analysis used an iterative proportional fitting
procedure (IPFP) to analyze eye-movement data. According to the definition of

Table 2 Means and standard deviations for eye-movement measures for good- and poor-performance
groups
Good-performance Poor-performance F
group group
M (SD) M (SD)

Whole article
Total reading time (s) 208.36 (90.05) 127.81 (74.39) 13.08**
Title section
Total reading time (s) 1.15 (0.80) 0.80 (1.28) 1.80
Text section
Total reading time (s) 170.81 (91.77) 107.47 (48.16) 10.18**
Proportion of fixation duration (%) 84 (10) 86 (11) 2.81
Mean fixation duration (ms) 304.73 (49.79) 290.70 (36.24) 1.64
Saccade numbers of text to diagram 5.37 (3.86) 2.68 (2.36) 9.02**
Diagram section
Total reading time (s) 36.40 (29.72) 19.54 (24.87) 10.67**
Proportion of fixation duration (%) 16 (10) 13 (11) 2.85
Mean fixation duration (ms) 287.29 (42.14) 249.26 (55.04) 8.71**

** p \ 0.01

123
Eye-movement patterns and reader characteristics of… 1461

Fiengberg (1970), ‘‘IPFP is an iterative algorithm for estimating cell values of a

contingency table such that the marginal totals remain fixed and the estimated
table decomposes into an outer product.’’
We divided the illustrated text into six AOIs: the title, paragraph 1, paragraph 2,
paragraph 3, upper diagram, and lower diagram (see Fig. 1). Calculations were
performed in accordance with previous studies (Jian, 2016; Jian et al., 2014). We
first calculated the fixation transitions among each of the six AOIs in the illustrated
text. The adjusted residuals are shown in Tables 3 and 4 in ‘‘Appendix’’. The rows
represent the initial AOIs and the columns represent subsequent AOIs. Z-values
higher than 1.96, 2.58, and 3.36, indicate that the transfer sequence reached the
p \ 0.05, p \ 0.01, and p \ 0.001 cutoff levels for statistical significance,
respectively. For example, Fig. 2a indicated that 63% readers of the good-
performance group transferred their eye fixations from the paragraph 2 to the
paragraph 1 in the first-pass reading, and this percentage was significantly exceeded
the expected value, not the outcome of random.

Fig. 2 First-pass transition

diagrams for good- and poor- Title
performance groups. The .58 .59
numbers beside the arrow
indicators show the transition
probabilities Paragraph 1 Upper
.63 diagram

Paragraph 2 .28

.67
.22 Bottom
Paragraph 3 diagram

(a) Good-performance readers

Title
.32

Paragraph 1 Upper
diagram

Paragraph 2 .28

.68
.17 Bottom
Paragraph 3 diagram

(b) Poor-performance readers

123
1462 Y.-C. Jian

First-pass and total-pass pathways are reported below. The first-pass pathway
reflects participants’ initial processing, and the total-pass pathway reflects their late
and higher-order cognitive processing.

First-pass fixation sequences in both groups

The first-pass transition diagrams for the students with good and poor performance
are presented in Fig. 2; the respective Z-value matrix is shown in Table 3 in
‘‘Appendix’’. We found that groups had both similarities and differences in the
reading pathways for the first-pass fixation sequences.
First-pass reading sequences in the groups had two similar characteristics. First,
both groups located their fixations on the upper diagram and then transferred their
next fixation to the lower diagram. The transfer probability from the upper diagram
to the lower diagram was significantly higher than the expected value (Z = 3.20,
p \ 0.01 for students with good performance; Z = 2.42, p \ 0.05 for students with
poor performance). Second, both groups located their fixations on the lower diagram
and then transferred their next fixation to Paragraph 3 of the text. The transfer
probability from the lower diagram to Paragraph 3 was significantly higher than the
expected value (Z = 4.13, p \ 0.001 for the group with good performance;
Z = 2.73, p \ 0.01 for the group with poor performance).
Differences between the groups during the first-pass reading sequences had two
characteristics. First, after reading the title for the first time, students with good
performance transferred their next fixation to the upper diagram. The transfer
probability from the title to the upper diagram was significantly higher than the
expected value (Z = 3.22, p \ 0.01); however, students with poor performance did
not prefer any particular AOIs in the article after reading the title for the first time
(p [ 0.05). Second, students with good performance read each paragraph of the text
and then returned to the previous paragraph (from Paragraph 1 to the title, from
Paragraph 2 to Paragraph 1, and from Paragraph 3 to Paragraph 2) after leaving the
target area, when they read the text for the first time (Z = 4.62, p \ 0.001;
Z = 3.22, p \ 0.01; Z = 5.43, p \ 0.001, respectively). Although students with
poor performance also had this tendency, significant effects were present only in the
transition from Paragraph 1 to the title (Z = 3.24, p \ 0.01) and from Paragraph 3
to Paragraph 2 (Z = 3.66, p \ 0.001).

Total-pass fixation sequences in both groups

The total-pass transition diagrams for students with good and poor performance are
presented in Fig. 3, and the Z-value matrix is shown in Table 4 in ‘‘Appendix’’.
Similar to first-pass fixation sequences, we also found similarities and differences in
the total reading pathways of both groups.
Total-pass reading sequence similarities between the groups also had two
characteristics. First, overall, both groups interacted more with each paragraph in
the text section. Transfer probabilities from the title to paragraph 1 were
significantly higher than the expected values (Z = 3.71, p \ 0.001 for the group
with good performance; Z = 4.46, p \ 0.001 for the poor-performing group). A

123
Eye-movement patterns and reader characteristics of… 1463

Fig. 3 Total-pass transition

diagrams for good- and poor- Title
performance groups. The .45 .35
numbers beside the arrow .14
indicators show the transition
probabilities Paragraph 1 Upper
.57 .31 diagram

Paragraph 2 .34 .40

.40 .44
.13 Bottom
Paragraph 3 diagram

(a) Good readers

Title
.48
.15
Paragraph 1 .28 Upper
.55 diagram

Paragraph 2 .47 .35

.51 .44
Bottom
Paragraph 3 diagram

(b) Poor readers

similar result was found for the reverse transfer from paragraph 1 to the title
(Z = 5.82, p \ 0.001; Z = 5.46, p \ 0.001). Both groups had higher transfer
probabilities from paragraph 1 to paragraph 2 than the expected values (Z = 6.53,
p \ 0.05 for the group with good performance; Z = 5.19, p \ 0.001 for the group
with poor performance). However, the transfer probability for the reverse transfer,
from paragraph 2 to paragraph 1, was only significantly higher than the expected
value for students with good performance (Z = 2.79, p \ 0.01). Furthermore, both
groups had higher transfer probabilities from paragraph 2 to paragraph 3 than the
expected values (Z = 7.29, p \ 0.001 for students with good performance;
Z = 6.93, p \ 0.001 for students with poor performance). A similar result was
found for the reverse transfer from paragraph 3 to paragraph 2 (Z = 6.73, p \ 0.001
for the group with good performance; Z = 6.36, p \ 0.001 for the group with poor
performance). Second, both groups transferred their fixations back and forth
between the two diagrams, as the transfer probability from the upper diagram to the
lower diagram was significantly higher than expected (Z = 8.57, p \ 0.001 for the

123
1464 Y.-C. Jian

group with good performance; Z = 8.70, p \ 0.001 for the group with poor
performance). A similar result was found for the reverse transfer from the lower
diagram to the upper diagram (Z = 7.78, p \ 0.001 for students with good
performance; Z = 5.67, p \ 0.001 for students with poor performance).
Differences between the two groups regarding total-pass reading sequences had
two characteristics. First, after reading the title, many readers with good
performance transferred their next fixation to the upper diagram, with the transfer
probability from the title to the upper diagram being significantly higher than the
expected value (Z = 3.36, p \ 0.01); however, this preference was not present in
the readers with poor performance (p [ 0.05). Second, the transfer between the text
and diagram sections was different between the groups, as the readers with good
performance read paragraph 3 and then transferred their next fixation to the lower
diagram; therefore, the transfer probability from paragraph 3 to the lower diagram
was significantly higher than the expected value (Z = 2.02, p \ 0.05). However,
readers with poor performance read the upper diagram and then transferred their
next fixation to paragraph 1; therefore, the transfer probability from the upper
diagram to paragraph 1 was significantly higher than the expected value (Z = 2.25,
p \ 0.05).

Discussion

This study used eye-tracking technology along with several reading tests and
questionnaires to investigate reading processes and reader characteristics in sixth-
grade readers with good and poor reading performance. We found both similarities
and differences between these groups.

Similarities in readers with good and poor performance

This study found several similar characteristics in students with good and poor
performance. First, both groups had a similar level of familiarity with the topic of
the article; nearly half of all scientific terms used in the article were familiar to both
groups, as they expressed that they knew the meaning of these terms before reading
the article. The method of using academic terms to test prior knowledge and article
familiarity was also used by Liebfreund (2015). Further, both groups in this study
had similar reading abilities, according to the reading comprehension screening test
(Ko, 1999). Therefore, topic familiarity and reading ability may not have
contributed to the between-group discrepancies in learning outcomes that we
observed. According to the findings of this study, reading strategies that the readers
adopted, the mental effort (reading engagement) invested in reading, and confidence
in reading abilities et al. might be the main factors that resulted in different learning
outcomes. Some aspects of this finding are consistent with those of Schroeder
(2011), who found that 8th and 9th grade students with better comprehension read
infrequent concepts more carefully and spent more time on mental model updating
while reading. Besides, both groups rated the article difficulty was approximately
medium-level difficulty.

123
Eye-movement patterns and reader characteristics of… 1465

Second, both groups had several similar reading pathways. For example, after
viewing the upper diagram, which depicted the mechanism of gas exchange through
the skin of animals, most readers in both groups transferred their eye fixations to the
bottom diagram, which depicted the process of inhalation and exhalation during the
respiratory movement, and then they referred to the relevant information in
Paragraph 3 of the text section. This result implies that sixth-grade students might
have developed the reading strategy of being able to correlate the text and the
diagram to some degree, even though their reading abilities are not yet mature.
Readers with poor performance correlated irrelevant information (eye fixation from
the upper diagram to Paragraph 1). In addition, the eye movement pathways in both
groups went back and forth within the text and between the two diagrams, but they
rarely crossed between different representations (text and diagrams). The referenc-
ing strategy between the text and diagram was, therefore, not developed fully in
sixth-grade students compared with adult readers. A recent study (Jian, 2016) found
that adult readers refer to the relevant textual and pictorial information in articles
with a very high frequency, even if they read a science textbook aimed at primary
school students.
Third, both groups mainly relied on the textual information, as opposed to the
diagram; the proportion of reading time spent on the text and diagrams were
approximately 85 and 15%, respectively. This proportion is in similar to the
results of Moore and Scevak (1997); however, an even bigger difference was
found by Hannus and Hyönä (1999). Moore and Scevak used think-aloud
protocols and found that approximately 13% of their seventh-grade participants
reported having used the diagram information when reading the illustrated text.
However, Hannus and Hyönä used eye-tracking technology, and found that fourth
grade students spent only 6% of their reading time on the diagram section,
irrespective of whether their reading abilities were good or poor. However, this
proportion was lower in student readers than in adults when reading illustrated
scientific texts, as adults spent approximately 70 and 30% of their total reading
time on the text and a diagram, respectively (Jian, 2016; Jian & Wu, 2015). These
studies suggest the presence of a developmental curve: the older a reader is, the
better is his or her diagram literacy. To conclude, these findings suggest that both
young and adult readers mainly learned new scientific information via reading
descriptions. Reading is driven by the text, rather than by illustrations; readers first
integrate the representations of related sentences in the text, and then use diagrams
to construct or check their mental model related to this integrated information
(Hegarty & Just, 1993).

Discrepancies between readers with good and poor performance

In this study, we also found several apparent differences between the eye-
movement patterns and reader characteristics of the participants with good and
poor performance. First, sixth graders with good performance were more capable
of monitoring their comprehension. Similarly, previous researchers (Oakhill &
Cain, 2007; Van der Schoot, Reijntjes, & Van Lieshout, 2012) have found that
skilled readers are capable of monitoring their comprehension, and commented

123
1466 Y.-C. Jian

that this may result in more regressive eye movements in comparison to less
skilled readers. de Leeuw, Segers, and Verhoeven (2016) also found that primary-
school students performed more and longer look backs when reading difficult
texts. In our study, we demonstrated the presence of this ability from the initial
processing stage to the late processing stage of the whole reading process.
Sequential analysis of eye movements showed that readers with good performance
went back to previous paragraphs (from Paragraph 2 to Paragraph 1, from
Paragraph 3 to Paragraph 2) to find relevant semantic information, such as the
term ‘‘respiration,’’ when their comprehension was obstructed as they read the
text. This phenomenon occurred in both first-pass and total-pass reading
sequences.
Second, the diagram literacy of sixth graders with good performance was better
than that of sixth graders with poor performance. This study showed that readers
with good performance spent twice as much time on viewing the diagrams in the
scientific article compared to readers with poor performance. Additionally, students
with good performance had longer mean fixation duration on the diagrams than that
of students with poor performance. These findings suggest that students with good
performance perceived the importance of diagrams in the science article and
invested more time and mental effort to encode the diagrammatic information.
Furthermore, readers with good performance connected the textual and pictorial
representations more frequently, and made more transitions between the text and
diagrams in the article. For example, paragraph 3 described human respiration, and
the semantically relevant diagram depicted two human bodies during the process of
inhalation and exhalation, as well as numerous important parts of the body (e.g., the
ribs, lungs, and diaphragm) and their movement status (e.g., ascend, descend,
enlarge, and deflate). Readers with good performance had higher transition
probabilities between both representations of text and diagrams. This tendency to
refer to different representations is consistent with the results of a previous study
conducted with fourth-grade student participants. In groups that were divided
according to reading ability in advance (Hannus & Hyönä, 1999) and according to
reading performance afterwards (Mason et al., 2013a, b), readers with good
performance showed more frequent referring behavior between the text and
diagrams.
Third, readers with good performance were more interested in diagrams.
Sequential analysis of eye movements showed that most readers with good
performance observed the diagram after reading the title, but this tendency was not
present in readers with poor performance. Based on the data of first-pass reading, we
cannot exclude the possibility that the eye fixation of readers was attracted to the
diagrams because of their color and shape. Nevertheless, the data of total-pass
reading (included first-pass and rereading-pass) showed the same preference in
readers with good performance. Therefore, this reading pathway (reading the title
being immediately followed by reading the upper diagram) was not only an
automatic behavior driven by perceptual stimuli, but also an intentional cognitive
behavior, because rereading is a less automatic and more targeted process (Mason
et al., 2013a, b, 2015). In agreement with this assertion, Mason et al. (2015) showed
that second-pass reading can be used to predict readers’ learning outcomes. In

123
Eye-movement patterns and reader characteristics of… 1467

addition, the subjective results of the questionnaire in this study also indicated that
readers with good performance were attracted to and liked diagrams.
Fourth, the two groups differed in terms of reader characteristics, especially in
the affective dimension. The present results indicate that readers with good
performance had higher reading self-efficacy than did those with poor performance.
This finding corresponds to that of previous research, which shows that learners’
self-efficacy is positively related to reading comprehension (Katzir et al., 2009). The
present study also shows that the readers with good performance were more
attracted to the diagrams in the learning materials and liked them. This finding
implies that reading interest may be related to reading performance (Baker &
Wigfield, 1999; Taboada et al., 2009). Furthermore, the readers with good
performance in this study were found to self-evaluate their learning outcomes more
positively, and were willing to invest more time and effort in engaging in reading
episodes.

Conclusion, contributions, and limitations

This study investigated the cognitive processes and reader characteristics of sixth-
grade students with good and poor performance when reading illustrated scientific
text. Using eye-tracking technology and several reading tests and questionnaires, we
found that various cognitive (reading strategy, diagram literacy) and affective
(reading self-efficacy, article likeness, diagram attraction, and engagement) aspects
affected the students’ reading performance.
This study has both theoretical and practical significance. In terms of theory,
the present findings extend and contribute to the field of multimedia learning. The
theory of multimedia learning (Mayer, 2005) clearly elaborates the cognitive
processes of illustrated text reading, but does not mention reader characteristics or
affective factors, which may be influential variables in multimedia learning. The
findings of this study showed that both cognitive and affective factors might
indeed influence reading outcomes, and indicated that reader characteristics like
high self-efficacy, interest, engagement, and diagram utilization might have a
positive influence on learning performance. However, this study did not determine
which variable(s) might be the main factors causing group differences in terms of
reading outcomes. Indeed, alternative explanations are possible. According to the
simple view of reading (Hoover & Gough, 1990), word recognition ability is an
important component of reading comprehension. Thus, this variable may result in
group differences in reading outcomes. Although the word recognition ability of
the two groups in this study were not paired in advance, their scores on a standard
Chinese character recognition test (Huang, 2001) was applied as a covariance in
subsequent statistical analysis. After excluding the influence of word recognition
ability, the two groups still had many differences in terms of the cognitive
processes of illustrated text reading. Therefore, word recognition may not be the
only main variable influencing the reading process and learning outcomes. To
determine which variable(s) might be a main factor(s), further research is needed

123
1468 Y.-C. Jian

to manipulate such variables systematically to determine possible causal

relationships.
This study also has practical significance in terms of its educational
implications. Using sequential analysis to analyze eye movements facilitates
greater understanding of illustrated text reading processes and readers’ strategies.
For example, the findings showed that readers with good performance had higher
transition probabilities of eye fixations to the diagram section after reading the
title. Knowing how young students read illustrated texts not only helps primary
school teachers to understand the process of reading comprehension among
students under various conditions, but also allows them to instruct students more
appropriately and help them to develop adequate reading strategies for the future.
For example, students might have to learn how to decode pictorial information or
to use reference strategies. In sum, we suggested that teachers in the classroom
should not only improve students reading strategies but also design teaching
activity to trigger students’ reading motivation and interests and cultivate their
self-efficacy to help them overcome difficulties in reading comprehension.
Some limitations to this study have to be taken into consideration. We used
only one biological text with two diagrams; therefore, our findings cannot be
generalized to other kinds of diagrams. Visual representations in a scientific
article might be formatted as, for example, diagrams, illustrations, photographs,
or flow charts. These representations carry different amounts of information
displayed in various forms. Because of this diversity, it is difficult to evaluate the
effect of several scientific graphs. However, this diversity is typical in scientific
articles or textbooks. Future studies need to overcome this limitation. In addition,
future research might take advantage of combining eye-tracking technology and
the retrospective think-aloud technique (Sung, Wu, Chen, & Chang, 2015; van
den Haak, De Jong, & Schellens, 2003). Replaying videos of eye-movement
tracks to participants and instructing them to report on what they were thinking
might provide more information to help understand reading processes and
strategies.

Acknowledgements This study is supported by the Grants MOST103-2511-S-003-065-MY3 and

MOST105-2628-H-003-002-MY3 from the Ministry of Science and Technology in Taiwan.

Appendix

See Tables 3 and 4.

123
Eye-movement patterns and reader characteristics of… 1469

Table 3 Z-value matrix of the first-pass sequences for good- and poor-performance groups
Target AOI Title Paragraph 1 Paragraph 2 Paragraph 3 Upper Bottom
diagram diagram

Start AOI
Good-performance group
Title 0.26 -2.05 -1.01 3.22** -1.36
Paragraph 1 4.62*** -0.88 -1.01 -0.88 -1.36
Paragraph 2 -1.81 3.22** -1.06 -0.47 -0.50
Paragraph 3 -2.04 -2.65 5.43*** -1.67 -0.15
Upper diagram 0.64 1.10 -2.75 -1.06 3.20**
Bottom diagram -1.58 -2.33 1.09 4.13*** -0.11
Poor-performance group
Title 1.58 -0.75 -0.74 0.40 -1.21
Paragraph 1 3.24** 0.41 -1.01 -0.15 -1.65
Paragraph 2 -1.12 1.58 0.00 0.97 -1.12
Paragraph 3 -1.80 -2.50 3.66*** -1.67 1.28
Upper diagram 0.79 1.23 -2.39 -1.01 2.42*
Bottom diagram -1.65 -1.65 -0.71 2.73** 1.65

* p \ 0.05; ** p \ 0.01; *** p \ 0.001

Table 4 Z-value matrix of the total-pass sequences for good- and poor-performance groups
Target AOI Title Paragraph 1 Paragraph 2 Paragraph 3 Upper Bottom
diagram diagram

Start AOI
Good-performance group
Title 3.71*** -2.12 -2.74 3.36** -2.01
Paragraph 1 5.82*** 6.53*** -6.34 -0.64 -4.43
Paragraph 2 -4.14 2.79** 7.29*** -4.64 -5.61
Paragraph 3 -2.87 -4.48 6.73*** -3.66 2.02*
Upper diagram 0.63 1.27 -2.88 -5.01 8.57***
Bottom diagram -1.45 -4.20 -3.07 1.58 7.78***
Poor-performance group
Title 4.46*** -1.88 -2.44 1.38 -1.06
Paragraph 1 5.46*** 5.19*** -4.75 -0.67 -3.66
Paragraph 2 -3.04 0.11 6.93*** -3.46 -4.88
Paragraph 3 -2.46 -4.15 6.36*** -2.17 0.24
Upper diagram 0.23 2.25* -1.90 -3.97 8.70***
Bottom diagram -1.58 -1.63 -4.60 -0.17 5.67***

* p \ 0.05; ** p \ 0.01; *** p \ 0.001

123
1470 Y.-C. Jian

References

Ainsworth, S. (1999). The functions of multiple representations. Computers and Education, 33(2),
131–152. doi:10.1016/S0360-1315(99)00029-9.
Andrews, S., Miller, B., & Rayner, K. (2004). Eye movements and morphological segmentation of
compound words: There is a mouse in mousetrap. European Journal of Cognitive Psychology, 16(1/
2), 285–311. doi:10.1080/09541440340000123.
Bakeman, R., & Gottman, J. M. (1997). Observing interaction: An introduction to sequential analysis
(2nd ed.). Cambridge: Cambridge University Press.
Baker, L., & Wigfield, A. (1999). Dimensions of children’s motivation for reading and their relations to
reading activity and reading achievement. Reading Research Quarterly, 34(4), 452–477. doi:10.
1598/RRQ.34.4.4.
Bandura, A., & Cervone, D. (1983). Self-evaluative and self-efficacy mechanisms governing the
motivational effects of goal systems. Journal of Personality and Social Psychology, 45(5),
1017–1028. doi:10.1037/0022-3514.45.5.1017.
Cook, M. P. (2006). Visual representations in science education: The influence of prior knowledge and
cognitive load theory on instructional design principles. Science Education, 90, 1073–1091. doi:10.
1002/sce.20164.
Cook, M., Carter, G., & Wiebe, E. N. (2008a). The interpretation of cellular transport graphics by
students with low and high prior knowledge. International Journal of Science Education, 30(2),
239–261. doi:10.1080/09500690601187168.
Cook, M., Wiebe, E. N., & Carter, G. (2008b). The influence of prior knowledge on viewing and
interpreting graphics with macroscopic and molecular representations. Science Education, 92,
848–867. doi:10.1002/sce.20262.
Cromley, J. G., Snyder-Hogan, L. E., & Luciw-Dubas, U. A. (2010). Cognitive activities in complex
science text and diagrams. Contemporary Educational Psychology, 35, 59–74. doi:10.1080/
104132001753149883.
de Leeuw, L., Segers, E., & Verhoeven, L. (2016). Role of text and student characteristics in real-time
reading processes across the primary grades. Journal of Research in Reading, 39(4), 389–408.
doi:10.1111/1467-9817.12054.
Evans, J. B. T. (2007). Dual-processing accounts of reasoning, judgment, and social cognition. Annual
Review of Psychology, 59, 255–287. doi:10.1146/annurev.psych.59.103006.093629.
Ferk, V., Vrtacnik, M., Blejec, A., & Gril, A. (2003). Students’ understanding of molecular structure
representations. International Journal of Science Education, 25(10), 1227–1245. doi:10.1080/
0950069022000038231.
Fienberg, S. E. (1970). An iterative procedure for estimation in contingency tables. Annals of
Mathematical Statistics, 41(3), 907–917. doi:10.1214/aoms/1177696968.
Grant, E. R., & Spivey, M. J. (2003). Eye movements and problem solving: Guiding attention guides
thought. Psychological Science, 14(5), 462–466. doi:10.1111/1467-9280.02454.
Guthrie, J. T., & Wigfield, A. (2000). Engagement and motivation in reading. In M. L. Kamil, P.
B. Mosenthal, P. D. Pearson, & R. Barr (Eds.), Handbook of reading research (Vol. III,
pp. 403–422). Mahwah, NJ: Erlbaum.
Hannus, M., & Hyönä, J. (1999). Utilization of illustrations during learning of science textbook passages
among low- and high-ability children. Contemporary Educational Psychology, 24(2), 95–123.
doi:10.1006/ceps.1998.0987.
Harber, J. N. (1983). The effects of illustrations on the reading performance of learning disabled and
normal children. Learning Disability Quarterly, 6(1), 55–60. doi:10.2307/1510866.
Hegarty, M. (1992). Mental animation: Inferring motion from static displays of mechanical systems.
Journal of Experimental Psychology. Learning, Memory, and Cognition, 18(5), 1084–1102. doi:10.
1037/0278-7393.18.5.1084.
Hegarty, M., & Just, M. A. (1993). Constructing mental models of machines from text and diagrams.
Journal of Memory and Language, 32(6), 717–742. doi:10.1006/jmla.1993.1036.
Hoover, W., & Gough, P. (1990). The simple view of reading. Reading and Writing: An Interdisciplinary
Journal, 2(2), 127–160. doi:10.1007/BF00401799.
Huang, H. S. (2001). Chinese character recognition test. Psychological Publishing Company (in
Chinese).

123
Eye-movement patterns and reader characteristics of… 1471

Jian, Y.-C. (2016). Fourth graders’ cognitive processes and learning strategies for reading illustrated
biology texts: Eye movement measurements. Reading Research Quarterly, 51(1), 93–109. doi:10.
1002/rrq.125.
Jian, Y. C., & Wu, C. J. (2015). Using eye tracking to investigate semantic and spatial representations of
scientific diagrams during text-diagram integration. Journal of Science Education and Technology,
24(1), 43–55. doi:10.1007/s10956-014-9519-3.
Jian, Y. C., Wu, C. J., & Su, J. H. (2014). Learners’ eye movements during construction of mechanical
kinematic representations from static diagrams. Learning and Instruction, 32, 51–62. doi:10.1016/j.
learninstruc.2014.01.005.
Johnson, C. I., & Mayer, R. E. (2012). An eye movement analysis of the spatial contiguity effect in
multimedia learning. Journal of Experimental Psychology: Applied, 18(2), 178–191. doi:10.1037/
a0026923.
Katzir, T., Lesaux, N., & Kim, Y. (2009). The role of reading self-concept and home literacy practices in
fourth grade reading comprehension. Reading and Writing, 22(3), 261–276. doi:10.1007/s11145-
007-9112-8.
Kirby, J. R., Ball, A., Geier, B. K., Parrila, R., & Wade-Woolley, L. (2011). The development of reading
interest and its relation to reading ability. Research in Reading, 34(3), 263–280. doi:10.1111/j.1467-
9817.2010.01439.x.
Ko, H. W. (1999). Reading comprehension-screening test (in Chinese). Psychological Testing, 46, 1–11.
Kress, G., & van Leeuwen, T. (1996). Reading images: The grammar of the visual design. London and
New York: Routledge.
Liebfreund, M. D. (2015). Success with informational text comprehension: An examination of underlying
factors. Reading Research Quarterly, 50(4), 387–392. doi:10.1002/rrq.109.
Lin, Y. C., Lee, C. S., Huang, N. T., Chang, Y. T., & Tsai, S. F. (2008). Living science and technology
textbook. Kang Hsuan Company Press.
Mason, L., Pluchino, P., Tornatora, M. C., & Ariasi, N. (2013a). An eye-tracking study of learning from
science text with concrete and abstract illustrations. Journal of Experimental Education, 81(4),
356–384. doi:10.1080/00220973.2012.727885.
Mason, L., Tornatora, M. C., & Pluchino, P. (2013b). Do fourth graders integrated text and picture in
processing and learning from an illustrated science text? Evidence from eye-movement patterns.
Computers and Education, 60(1), 95–109. doi:10.1016/j.compedu.2012.07.011.
Mason, L., Tornatora, M. C., & Pluchino, P. (2015). Integrative processing of verbal and graphical
information during re-reading predicts learning from illustrated text: An eye-movement study.
Reading and Writing, 28(6), 851–872. doi:10.1007/s11145-015-9552-5.
Mayer, R. E. (2005). The Cambridge handbook of multimedia learning. New York, NY: Cambridge
University Press.
McTigue, E. M. (2009). Does multimedia learning theory extend to middle-school students?
Contemporary Educational Psychology, 34(2), 143–153. doi:10.1016/j.cedpsych.2008.12.003.
McTigue, E. M., & Flowers, A. C. (2011). Science visual literacy: Learners’ perceptions and knowledge
of diagrams. The Reading Teacher, 64(8), 578–589. doi:10.1598/RT.64.8.3.
Miller, B. W. (2015). Using reading times and eye-movements to measure cognitive engagement.
Educational Psychologist, 50(1), 31–42. doi:10.1080/00461520.2015.1004068.
Moore, P. J., & Scevak, J. J. (1997). Learning from texts and visual aids: A developmental perspective.
Journal of Research in Reading, 20(3), 205–223. doi:10.1111/1467-9817.00033.
Norman, R. R. (2012). Reading the graphics: What is the relationship between graphical reading
processes and student comprehension? Reading and Writing, 25(3), 739–774. doi:10.1007/s11145-
011-9298-7.
Oakhill, J., & Cain, K. (2007). Issues of causality in children’s reading comprehension. In D. McNamara
(Ed.), Reading comprehension strategies: Theories, interventions, and technologies (pp. 47–72).
New York, NY: Erlbaum.
Paivio, A. (1990). Mental representations: A dual coding approach (pp. 53–83). New York, NY: Oxford
University Press.
Rayner, K. (1998). Eye movements in reading and information processing: 20 years of research.
Psychological Bulletin, 124(3), 372–422. doi:10.1037/0033-2909.124.3.372.
Reid, D. J., & Beveridge, M. (1986). Effects of text illustration on children’s learning of a school science
topic. The British Journal of Educational Psychology, 56(3), 294–303. doi:10.1111/j.2044-8279.
1986.tb03042.x.

123
1472 Y.-C. Jian

Rusted, J., & Coltheart, M. (1979). Facilitation of children’s prose recall by the presence of pictures.
Memory and Cognition, 7(5), 354–359. doi:10.3758/BF03196939.
Schnotz, W., & Bannert, M. (2003). Construction and interference in learning from multiple
representation. Learning and Instruction, 13(2), 141–156. doi:10.1016/S0959-4752(02)00017-8.
Schroeder, S. (2011). What readers have and do Effects of students’ verbal ability and reading time
components on comprehension with and without text availability. Journal of Educational
Psychology, 103(4), 877–896. doi:10.1037/a0023731.
Schunk, D. H. (1981). Modeling and attributional effects of children’s achievement: A self-efficacy
analysis. Journal of Educational Psychology, 73(1), 93–105. doi:10.1037/0022-0663.73.1.93.
Segers, E., Verhoeven, L., & Hulstijn-Hendrikse, N. (2008). Cognitive processes in children’s multimedia
text learning. Applied Cognitive Psychology, 22, 375–387.
Slough, S., & McTigue, E. (2010). Introduction to the integration of verbal and visual information in
science texts. Reading Psychology, 31(3), 206–212. doi:10.1080/02702710903241397.
Small, M. Y., Lovett, S. B., & Scher, M. S. (1993). Pictures facilitate children’s recall of unillustrated
expository prose. Journal of Educational Psychology, 85(3), 520–528. doi:10.1037/0022-0663.85.3.
520.
Sung, Y. T., Wu, M. D., Chen, C. K., & Chang, K. E. (2015). Examining the online reading behavior and
performance of fifth-graders: Evidence from eye-movement data. Frontiers in Psychology, 6, 1–15.
doi:10.3389/fpsyg.2015.00665.
Taboada, A., Tonks, S. M., Wigfield, A., & Guthrie, J. T. (2009). Effects of motivational and cognitive
variables on reading comprehension. Reading and Writing, 22(1), 85–106. doi:10.1007/s11145-008-
9133-y.
Unrau, N. J., & Quirk, M. (2014). Reading motivation and reading engagement clarifying commingled
conceptions. Reading Psychology, 35(3), 260–284. doi:10.1080/02702711.2012.684426.
van den Haak, M. J., De Jong, M. D. T., & Schellens, P. J. (2003). Retrospective vs. concurrent think-
aloud protocols: Testing the usability of an online library catalogue. Behavior and Information
Technology, 22(5), 339–351. doi:10.1080/0044929031000.
Van der Schoot, M., Reijntjes, A., & Van Lieshout, E. C. M. D. (2012). How do children deal with
inconsistencies in text? An eye fixation and self-paced reading study in good and poor reading
comprehenders. Reading and Writing, 25(7), 1665–1690. doi:10.1007/s11145-011-9337-4.
Wigfield, A., Guthrie, J. T., Perencevich, K. C., Taboada, A., Klauda, S. L., McRae, A., et al. (2008). Role
of reading engagement in mediating effects of reading comprehension instruction on reading
outcomes. Psychology in the Schools, 45(5), 432–445. doi:10.1002/pits.20307.

123