Reading Behavior in Science Comics and Its Relations with Comprehension Performance and Reading Attitudes: An Eye-Tracker Study

One of the ways to acquire scientific knowledge is by reading science comics. This study aims to investigate the attitudes (e.g., reading habits, interest, motivation) of university students toward reading science comics, and how they read science comics for acquiring scientific knowledge reflected by an eye tracker. Sixty-five undergraduates were invited to complete an attitudes and habits questionnaire of reading comics, after which they read a science comic where their reading processes were recorded by an eye tracker; finally, they completed a reading comprehension test. The results showed that most undergraduates had a positive attitude for reading comics, and were more likely to learn science by reading comics rather than texts. In addition, the analysis results of the linear mixed-effect models indicated that fixation with regard to re-reading durations could promote post-reading comprehension. The readers who particularly re-read the important information in relation to the boxed-in texts and graphics for a longer period scored higher in the post-test (e.g., the cause and variation processes of cancer). The analysis of variance also indicated that readers who exhibited a good test performance allocated more re-reading time on informational texts and diagrams. This meant they were more intentional and selective in re-reading the core and concept-intensive information; however, concerning the reader who performed poorly in their test, this reading pattern was unapparent. Therefore, those who exhibited a better grasp of the knowledge in the science comics were likely to fully process the areas that presented relevant information concerning important science concepts.

Keywords: Science comics; Eye movements; Reading comprehension; Reading attitudes

Copyright comment Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Scientific knowledge is full of abstract concepts and has been communicated predominantly through language as the medium. Its exposition is often conducted neutrally, which requires readers to extract and understand abstract concepts out of written texts. Comics have recently emerged as a medium of mass communication and scientific knowledge dissemination. Farinella ([4]) believed that comics could be considered a more direct vehicle of communicating education for adults and children alike. Science education researchers have often regarded comics as a substandard form of scribbling, composing, and printing (Hosler & Boomer, [6]; Spiegel et al., [20]). Recently, however, many comic works, such as Dr. Stone and Cells at Work, have been created to communicate scientific knowledge to the public. A growing number of scientists have also become devoted to using this attractive and humorous medium for introducing scientific principles and stories (Hosler & Boomer, [6]; Lin et al., [13]; Tatalovic, [22]). Research into the efficacy of science learning through comics, and its influencing factors, has begun to gain ground (Farinella, [4]; Hosler & Boomer, [6]; Lin & Lin, [12]; Spiegel et al., [20]; Tatalovic, [22]).

From the perspective of cognitive science, the dual-code theory postulates that visual and textual stimuli could give rise to separate mental representations that benefit learning (Paivio, [17]). Presenting abstract scientific knowledge through comics allows readers to grasp the concepts via both visual and textual channels, thus, facilitating the formation of mental models.

Based on cognitive principles, Farinella ([4]) asserts that the three major advantages of using comics for education lie in visualization, storytelling, and metaphors. First, as scientific knowledge is often formed by a combination of relevant concepts, rather than a linear structure, a sequential art form like manga could construe a more concrete structure for abstract concepts, which could be further subdivided into smaller and easily understandable units, before merging into coherent pictures in the comics. Second, considering storytelling, readers could be engaged and immersed in the scientific knowledge along with the narrative. Third, as the narrative in manga is based on visual metaphors, such as the lines accompanied by character movements representing speed, beads of sweat symbolizing stress, etc., they could transform abstract concepts from the textual form into concrete, tangible graphics, allowing for more faithful representations of knowledge, and facilitating better understanding among laypersons.

Empirical studies on science comics have focused on the relationship between the type of text and the learning attitude and motivation of students (Hosler & Boomer, [6]; Lin et al., [13]; Spiegel et al., [20]; Tatalovic, [22]).

Tatalovic ([22]) reviewed science education research pertaining to science comics for education, which the author considered a genre of its own right, targeting both students and the public. Moreover, the pictorial representations and humorous narrative style were found to improve students' interest in learning science.

Hosler and Boomer ([6]) proposed that the interplay of textual and visual information in comics could serve the dual purpose of effectively conveying scientific concepts and arousing students' interest. Specifically, comics could build positive impressions on scientific knowledge among students. Their experiment compared students' attitude towards learning biology through comic books before and after instructions, with participating students comprising both biology majors and non-majors. Results showed that non-biology majors attained very low scores in both, relevant biological knowledge and attitude towards the subject matter; however, after receiving instructions, they showed significant improvements in both knowledge and attitude, which could be attributed to the use of comic books in the instructions. The science comics led to increased scores in content knowledge among non-majors, suggesting that these texts fostered the understanding of relevant scientific knowledge. It could be inferred from this study that the comics could lead to learning effects that were at par with conventional texts; along with facilitating learning, comics could produce the effect of enticing interest in the subject.

In a study by Spiegel et al. ([20]), a total of 873 grade 10 students of biology were asked to engage in reading science comics. Results showed that the informal genre induced positive impacts among teenage learners of science, who were more engaged because of the comics. This effect was more pronounced among those who tended not to identify themselves as a "science kind of person." Compared with traditional learning materials, comics could more effectively induce interest in science among teenagers. In a previous study by Lin et al. ([13]) that employed both qualitative and quantitative methods, the effects of reading conventional science texts and comics were compared in a sample of 300 non-student adults, using stratified sampling method. Results revealed that the effect of comics on science learning was comparable with that of pure text reading. For those with limited prior knowledge, the motivation to learn science had increased. Contrarily, the pure-text group showed reduced interest in science. Interview data also suggested that participants preferred learning science from comics. The author reasoned that these comics introduced scientific knowledge with humor, which induced positive emotions in the learning process, thus increasing the level of engagement and improving the learning effects.

Empirical studies on comic reading have shown positive effects in both learning outcome and motivation. Aleixo and Sumner ([1]) invited adult participants to read content on the basic principles of sleep and assigned them to one of the three reading conditions: pure text, normal comics, and incongruous comics. Post-reading memory tests returned significantly higher scores in the normal comics group than the other two, with the pure text group performing better than the incongruous comics group. Lin and Lin ([12]) compared the effects of comics and pure text among students of different abilities. Results showed that conventional texts and comics brought the strongest learning effects in the high-ability group and the average-ability group respectively, while the effects of both formats were similar among those in the low-ability group. They concluded that four major factors affected the differential learning effects: perceived difficulty, interest in the subject matter, emotions towards science learning, and the duration of learning. To average achievers, comics reduced perceived difficulty in the understanding of scientific concepts, resulting in higher willingness to read. The comics provided scaffolding for learning, and assisted learning through contextualizing, visualizing, and humorizing the content, from which mental imageries were constructed. To high achievers, who possessed superior prior knowledge and reasoning abilities, the lower level of sequentiality in conventional texts allowed them to apply their skills in learning; however, comics appeared to hinder their understanding of abstract concepts. As for low achievers, given their limited prior knowledge and reading comprehension, spending prolonged periods on reading either format would only produce limited learning outcomes.

Thus, empirical studies majorly revealed positive learning effects of science comics. However, some researchers have aptly noted the importance of an appropriate proportion between comics and scientific content in science comics (Jee & Anggoro, [7]; Tatalovic, [22]). Tatalovic ([22]) reminded educators about the possibility of pursuing the value of humor at the expense of science education. Jee and Anggoro ([7]) highlighted the dual purpose of science comics as being entertainment and education. Nevertheless, to strike a balance between simplifying and comprehensively introducing scientific knowledge is challenging in practice. Jee and Anggoro concluded that the advantages of using comics as a medium of science education are as follows: (1) concrete visualization; (2) adjacency between graphical and textual information and facilitating integration; (3) narrative that assists understanding; (4) cross-reference between characters and readers; and (5) potential of personification for logical reasoning. Meanwhile, they stressed that the benefits might be accompanied by these potential drawbacks: (1) faulty deductions; (2) strengthened stereotyping and weakened thinking; (3) unnecessary moral attribution due to personification; and (4) mistaken understanding resulting from over-simplification.

With regard to the process of reading, which refers to the processing of real-time information, eye-tracking technology has proven to be a suitable method (Rayner, [19]). Recently, as the technology has become more sophisticated, it has been applied to the study of a vast range of disciplines, such as pure text reading, scientific text reading (Jian, [9], [10]; Jian et al., [8]; Jian, [11]; Tsai et al., [23]; Liu & Wu, [14]; Yang, [26]), and mathematical reading (Wu et al., [25]). However, limited studies were concerned with reading comics with the use of eye trackers (Jee & Anggoro, [7]; Tabassum et al., [21]).

Tabassum et al. ([21]) deployed the technology in the study of how university students (N = 60; age: 18–27) read terms and conditions documents in the format of pure text, illustrated text, and comics. Each participant was asked to read two of the formats and perform a post-test on the content; their preference was inquired about. Regarding reading time, comics were read for the longest duration; however, in terms of comprehension, none of the formats showed significant difference. Regarding survey results, twelve participants stated that the comic version was more visually appealing; three of them opined that the massive information on the comic version hindered reading. Based on the results, while reading comics could improve an individual's visual attention during the reading process, it would barely lead to the enhancement of their comprehension of the comics. It could possibly result in other cognitive costs, as suggested by Jee and Anggoro ([7]). In another study, von Reumont and Budke ([24]) examined the use of a combination of comics and maps in geography learning among teenage students (N = 36; age: 10–14). Using a pre-test to rule out prior knowledge, the researchers examined the process of teenage participants reading four sets of comics related to international rose trade, each of which was associated with some features on a map, followed by a post-test on comprehension and memory. Results revealed that among all areas of interest on the reading materials, textual elements received the most attention, with the highest fixation counts, number of revisits, and longest dwell time, compared with graphical elements. Attention to the map, measured by fixation counts and dwell time, was associated with post-test performance. Participants who made more eye movements between the map and graphics outperformed those who did not, supporting the value of multimedia in reading.

In conclusion, few studies examined the reading processes in relation to reading science comics with the use of eye trackers, as well as whether the readers' attitudes (e.g., reading habits and interest) and eye-movement patterns are relevant to scientific reading and its comprehension, which was unclear. This study aimed to investigate the attitudes of university students toward reading scientific comics, and how they read them to acquire scientific knowledge reflected by an eye tracker. In addition, since comics are composed of textual and graphical regions, and these regions can be divided into smaller regions according to their different functions, such as the conceptual description of the science concept, non-conceptual description of conversation (e.g., I am not interested in this...; Who actually wants to know about cancer?), boxed-in text, organizational diagram (e.g., sequential changes of cells), and decorative graphics (e.g., personification of cells). However, according to the researcher's knowledge, there was no research that analyzed these detailed regions to investigate if the different performance readers had different eye movements on the different regions of the scientific comics. Therefore, this study will analyze these detailed regions by providing multiple data of temporal and spatial readings to reflect sophisticated reading processes of scientific comic reading. There were four specific research questions in this study. First, what are the attitudes (e.g., reading habits, interest, motivation) of university students toward reading scientific comics? Second, is the performance of reading comprehension different among individuals with different reading habits (never, occasional, and often) concerning reading comics? Third, are eye-movement indicators concerning reading scientific comics associated with post-reading comprehension performance and the readers' prior knowledge about the reading material? Fourth, are eye-movement patterns concerning the detailed regions of the scientific comics associated with post-reading comprehension performance?

There were 65 undergraduate and postgraduate students (aged between 18 and 25, 38 females) recruited in this study. The students were from universities in Northern Taiwan and did not major in science. The students that majored in science were not recruited for this study to avoid the floor effect, as the content of the scientific comics may be too easy for them. All participants were native readers of Traditional Chinese and had normal or correct-to-normal vision. They provided written informed consent prior to the experimental procedure.

The reading materials were taken from the Encyclopedia Britannica (BomBom Story, [2]). Comprising six pages, the topic of the comic selection was cancer—its epidemiology, basic understanding of human cells, cell division, cancer cells, metastasis, types of cancer, cancer treatment, and prevention. The original layout was retained, and the comic selection was presented on a 27-inch LED monitor (1920 × 1200 pixels), one page at a time, using the Experiment Builder.

The assessment tools consisted of a pre-test, post-test, and questionnaire related to the student's attitude towards reading comics. The pre-test consisted of an open-ended question ("What do you know about cancer? Please elaborate as much as you can, for example, causes, the change of human cells, symptoms, etc."); the post-test comprised four questions, including "(1) Describe and explain the onset of cancer"; "(2) Compare and contrast division of normal cells and malignant cells"; "(3) Describe and explain the spread of cancer"; and "(4) Describe and explain the diagnosis of cancer." Expert review on the test questions had been conducted. These tests were created by the researcher and two experts with science matter degrees who were invited to confirm the tests' content validity regarding whether the assessments covered the important science concepts contained in the scientific comics.

A survey questionnaire on the attitude towards comics is comprised of two questions on the comic book reading habit and five statements on their views on comic reading. The questionnaire's reliability regarding the students' attitude towards reading comics was 0.801. The questions included the following: "(1) How often do you read 'recreational comics'?" and "(2) How often do you read 'educational comics'?"; responses were based on a 3-point Likert scale (rarely, occasionally, and quite often). The statements on the views were as follows: "(1) I am interested in reading comics"; "(2) It is easy to identify the reading order of comic books"; "(3) It is easier to learn about science from comic books"; "(4) It is more motivating to read educational science comic books than to read non-fiction science books"; and "(5) Irrelevant information in comics could be distracting, crowding out attention to crucial information"; the extent of agreement with these statements were rated on a 5-point Likert scale.

The eye-tracking system used in the present study was the EyeLink Portable DUO, with a sampling rate of 1000 Hz. The participant's head position was fixed using a chinrest placed 75‒80 cm away from a 27-inch screen (with a resolution of 1920 × 1200 pixels), from which the reading materials were presented.

Before the eye-tracking procedure began, the participant completed the questionnaire on the attitude towards comics, and the pre-test on cancer. Calibration with the eye tracker was performed before the participant commenced reading the comics, during which their eye movements were tracked. Reading was self-paced by the participant, but return to the previous page was not allowed. The participant was instructed to read the comics carefully to prepare for a post-reading assessment, which was conducted immediately after the reading process. The entire procedure lasted for approximately 25‒30 min.

After excluding two cases that showed failure of calibration, there remained 63 valid cases (age: M = 21.24, SD = 1.84; 38 females). On the pre- and post-tests, one score was awarded for one concept or content point. Typos were accepted if the meaning was recognizable. Scoring was conducted by two rates; discussions between the raters were carried out before the final scores were separately determined. The inter-rater consistency of the pre-test and post-test was 0.88 and 0.83, respectively.

Regarding the analysis of the eye movement data, the areas of interest (AOIs) should be defined first. The science comics were spatially divided into two levels of regions (Fig. 1), namely, textual and graphical. The textual region was broken down into content-specific regions including "conceptual," "non-conceptual," and "boxed-in" areas. The difference between non-conceptual texts and conceptual texts was that the former consisted of mainly dialogues expressed in first-person, while the latter often adopted a third-person point of view in explaining phenomena and concepts. Boxed-in texts were extra texts presented in separate boxes, serving as additional information to the topic covered in the dialog. The graphical region, like the textual region, was composed of "conceptual," "non-conceptual," and "boxed-in" areas. Conceptual diagrams provided support to the texts in describing and explaining the relevant concepts, systems, or processes; for instance, sequence diagrams and organizational diagrams. While non-conceptual graphics could either be depicting the characters and the scene, or simply be decorative. Boxed-in graphics accompanied boxed-in texts to provide additional information on relevant scientific knowledge.

Graph: Fig. 1The schematic representation of the six AOIs, defined by representational mode (textual vs. graphical) and content-specific areas (conceptual, non-conceptual, and boxed-in)

Several eye-movement indicators used in this study were defined as follows: (1) the mean fixation duration, being the average duration of all fixations, which represents the ease of decoding the text and diagram (Jian, [9], [10]; Mason et al., [15]; Miller, [16]); (2) the total fixation duration, being the sum of all fixation durations on a specific area, indicating the overall difficulty and the degree of cognitive effort required to process the reading materials (Chiou et al., [3]; Jian, [9]; [10]; Wu et al., [25]; Liu & Wu, [14]; Tsai et al., [23]; Yang, [26]); (3) the first-pass reading time, being the total duration of all fixations on a specific area during the initial reading and before exiting it, which represents the initial reading process, including the decoding of words or objects and the preliminary extraction of meaning from the reading material (Jian et al., [8]; Henderson et al., [5]); and (4) the re-reading time, being the total duration of all fixations on an AOI excluding the first-pass total fixation duration, as well as the duration of fixations leaving the AOI and then returning to re-read it, which represents more purposeful and deeper cognitive processing, which involves aspects such as comprehension and integration (Jian, [9], [10]; Jian et al., [8]; Mason et al., [15]).

Linear mixed-effects models were fitted using the nlme package (Pinheiro et al., [18]) on R (version 4.1.2).

Figure 2 exhibits interesting findings, which were obtained from the self-report questionnaires on science comic books: First, a majority of the undergraduate sample (73%) expressed interest in reading recreational comics, but, overall, those who reported no habit of reading science comics also constituted the majority (59%). Second, 60% of the respondents expressed liking comic books, 24% were neutral about them, while 16% held a negative view. Third, although comic books did not stipulate a clear or linear reading order similar to conventional texts, the overwhelming majority (86%) reported that they could derive the order during the reading process. Fourth, almost half of the respondents (47%) agreed or strongly agreed that they grasped scientific knowledge from a comic. Fifth, majority of them (65%) opined that comic books were more effective in inducing university students to engage in science learning, with a mere 16% expressing disagreement. However, 43% of the respondents perceived that they could be more easily distracted by irrelevant information when reading educational comics, than when reading conventional, non-fiction science texts.

Graph: Fig. 2Results of survey on the habit, interest, attitude with regard to comic reading

A one-way ANOVA on the post-reading scores among three comic reading habit groups (never, occasional, and often; Table 1) returned no significant difference; ps > 0.05.

Table 1 Post-reading comprehension performance by three levels of comic reading habits (never, occasional, and often reading)

Correlation analyses (Table 2) revealed that prior knowledge (i.e., pre-test scores) was negatively correlated with the total time spent on reading the comics, r = − 0.30, p = 0.02, and with the total number of eye fixations on the comics, r = − 0.31, p = 0.02. Evidently, participants with weaker prior knowledge compensated for the weakness with a higher level of engagement with the comics. Additionally, results showed that the longer the reading time, the better the post-reading comprehension; r = 0.25, p = 0.045.

Table 2 Correlations among pre-test scores, post-test scores, and eye-movement indicators on the entire comics

*indicates p <.05. ** indicates p<.01

We examined whether the mean fixation duration, total fixation duration, first-pass reading time, and re-reading time for the two main representational mode regions (textual vs. graphical), and the three associated content-specific areas (conceptual, non-conceptual, and boxed-in), could predict post-reading comprehension performance (Table 3). The two representational modes and three content-specific areas gave rise to six AOIs (see Fig. 3 as an example) altogether. Values were normalized before modelling.

• To examine whether the speed of encoding the six AOIs could be accounted for by post-reading comprehension performance, a linear mixed-effects model was based on post-test score with mean fixation duration as the fixed effect, and representational mode (textual vs. graphical) and content-specific areas as crossed random effects. Results showed that mean fixation duration did not significantly predict post-test scores, β = 0.03, t(371) = 0.50, p = 0.62. For the random effects, compared with separate models that removed the crossed design (i.e., with the six AOIs entered as random effects), the model with crossed design was not superior (model with crossed design: AIC = 1090, BIC = 1109; model without crossed design: AIC = 1088, BIC = 1103; log likelihood test (full model vs. representative-mode-only): χ2(1) = 2.70, p = 1.00), thus suggesting that crossed design of AOIs did not improve the prediction on post-test scores. Separate a-priori simple linear regressions showed that the mean fixation duration of none of the six AOIs could predict post-reading comprehension performance, ps > 0.05.

• To examine whether post-reading comprehension performance could be accounted for by the processing level of the six AOIs, a linear mixed-effects model was based on the post-test scores with total fixation duration as the fixed effect, and representational mode (textual vs. graphical) and content-specific areas as crossed random effects. Results showed that the prediction of total fixation duration was only marginal, with the p-value adjusted using the false discovery rate method for multiple comparisons, β = 0.12, SE = 0.06, t(371) = 2.10, p = 0.055. For the random effects, compared with separate models that removed the crossed design (i.e., with the six AOIs entered as random effects), the model with crossed design was not superior (model with crossed design: AIC = 1086, BIC = 1106; model without crossed design: AIC = 1084, BIC = 1100; log likelihood test (full model vs. representative-mode-only): χ2(1) = 0.28, p = 0.59), thus suggesting that the crossed design of AOIs did not improve the prediction on post-test scores. Separate a-priori simple linear regressions, with post-test scores as the dependent variable, and total fixation duration of individual AOIs as the predictor, revealed that, with p-values adjusted using the false discovery rate method, post-test scores only predicted the total fixation duration of the boxed-in text (adjusted R2 = 0.07), and boxed-in graphic areas (adjusted R2 = 0.08) with marginal significance, ps = 0.057. The significant fixed effect of post-test score in the main linear mixed-effects model might be largely attributed to the boxed-in text and boxed-in graphic areas.

• To examine whether post-reading comprehension performance was attributable to the level of initial processing on individual AOIs, a linear mixed-effects model was based on post-test scores with first-pass fixation duration as the fixed effect, and representative mode (textual vs. graphic) and content-specific areas as crossed random effects. Results showed that first-pass fixation duration was not a significant predictor of post-reading comprehension performance, β < − 0.001, SE = 0.05, t(371) = − 0.003, p = 1.00. Considering the random effects, compared with separate models that removed the crossed design (i.e., with the six AOIs entered as random effects), the model with the crossed design was not superior (model with crossed design: AIC = 1090, BIC = 1110; model without the crossed design: AIC = 1088, BIC = 1104; log likelihood test (full model vs. representative-mode-only) log likelihood test: χ2(1) < 0.001, p = 1.00), thus suggesting that the crossed design of AOIs did not improve the prediction on post-test scores. Separate a-priori simple linear regressions, with post-test scores as the dependent variable, and first-pass fixation duration of individual AOIs as the predictor, revealed that the first-pass fixation duration of none of the six AOIs significantly predicted the post-test scores, ps > 0.05.

Table 3 Summary statistics on eye-movement indicators on the six AOIs of the science comics

The number of percentages in brackets indicates the ratio of the specific area in the whole scientific comic

Graph: Fig. 3The examples of the AOIs (Areas of the Interests) in the experimental material

To examine whether post-reading comprehension performance could be explained by the level of processing during the re-reading of individual AOIs, a linear mixed-effects model was based on post-test scores with re-reading fixation duration as the fixed effect, and representative mode (textual vs. graphic) and content-specific areas as crossed random effects. Results showed that re-reading fixation duration could significantly predict post-test performance; β = 0.16, SE = 0.05, t(371) = 3.13, p = 0.006, with p-value corrected for multiple comparisons. Regarding the random effects, compared with separate models that removed the crossed design (i.e., with the six AOIs entered as random effects), the model with crossed design was not superior (model with crossed design: AIC = 1081, BIC = 1110; model without crossed design: AIC = 1079, BIC = 1094; log likelihood test (full model vs. representative-mode-only) log likelihood test: χ2(1) = 0.08, p = 0.78), thus suggesting that the crossed design of AOIs did not improve the prediction on post-test scores. Separate a-priori simple linear regressions, with post-test scores as the dependent variable, and the re-reading fixation duration of individual AOIs as the predictor, revealed that with p-values adjusted using false discovery rate method, only in the boxed-in textual area and the boxed-in graphic area did the re-reading fixation duration significantly predict the post-test performance (boxed-in textual area: β = 0.24, SE = 0.09, t(61) = 2.83, p = 0.03; boxed-in graphic area: β = 0.62, SE = 0.25, t(61) = 2.48, p = 0.048). The re-reading fixation duration of the other four areas did not significantly predict post-reading comprehension performance, ps > 0.05.

In accordance to the above analysis showing that the re-reading fixation durations could promote the post-test performance, which varied depending on the different AOIs in the scientific comics, this study further analyzed if the good and poor performances of the post-test could distinguish which information regarding the different AOIs is important. The two-way ANOVAs were conducted based on the total fixation durations as dependent variables, test performance groups (good and poor) as a between-subject independent variable, and the reading orders (first-pass and re-reading) as a within-subject independent variable.

Figure 4 exhibited that the boxed-in textual and boxed-in graphic AOIs were main effects of the test performance groups, F (1, 61) = 4.20, p < 0.05, η2 = 0.06; F (1, 61) = 6.93, p < 0.05, η2 = 0.10, and the reading orders, F (1, 61) = − 4.97, p < 0.05, η2 = 0.08; F (1, 61) = − 22.73, p < 0.001, η2 = 0.27, as well as the interactions between the test performance groups and the reading orders was significant, F (1, 61) = 4.97, p < 0.05, η2 = 0.08; F (1, 61) = 5.23, p < 0.05, η2 = 0.08. The simple effect tests showed that the high-test-performance group's total fixation durations of the re-reading were significantly longer than that of the first-pass regardless of reading the AOI of the boxed-in textual, t (32) = 3.14, p < 0.01, or reading the AOI of boxed-in graphic, t (32) = 4.52, p < 0.001. However, the low-test-performance group had similar total fixation durations of the first-pass and of the re-reading regarding reading the AOIs of the boxed-in textual and of the boxed-in graphic, ps > 0.05. For the AOIs of the conceptual and non-conceptual textual, there were main effects of reading orders, F (1, 61) = 151.34, p < 0.001, η2 = 0.71; F (1, 61) = 36.11, p < 0.001, η2 = 0.37, as both test performance groups had total fixation durations significantly longer regarding the first-pass reading than that of the re-reading, which had no main effect towards the test performance groups nor the interaction of the test groups and the reading orders, ps > 0.05. For the AOI of the conceptual graphic, there was a main effect regarding the reading orders, F (1, 61) = 28.15, p < 0.001, η2 = 0.32, as both test performance groups' total fixation durations of the re-reading were significantly longer than that of the first-pass. However, there was no main effect of the test performance groups nor the interaction of the test groups and the reading orders, ps > 0.05. As for the AOI of the non-conceptual graphic, there were no main effects nor interaction with regard to the test performance groups and the reading orders, ps > 0.05.

Graph: Fig. 4The total fixation durations of the first-pass and the re-reading on the six AOIs in the scientific comics for the two groups of good versus the poor test performances

This study aims to investigate if the attitudes (e.g., reading habits, interest, motivation) of university students towards reading scientific comics is relevant to their reading comprehension of its content, and if readers' eye-movement patterns regarding the detailed regions of the scientific comics are associated with their post-reading comprehension performance. The four (4) major findings of this study are discussed as follows:

First, the results from the questionnaire analysis, the attitude of university students towards comics is largely positive. This finding was similar with a previous study, which indicated that regardless of being an adult or teenage reader, they had a positive attitude towards reading comics (Hosler & Boomer, [6]; Lin et al., [13]; Spiegel et al., [20]; Tatalovic, [22]). Despite this, the habit of reading recreational comics in this study does not match the level of interest professed. Majority of the participants expressed interest in comic reading; however, they were limited to reading for leisure, as only a small proportion reported a habit of reading comics to acquire knowledge. Only half of the participants in this study agreed that it is easier to learn about science from comic books as opposed to learning from text. This was different from the previous study, which indicated that when compared with scientific text, scientific comics could increase the readers' motivation and decrease the subjective perceived difficulty of the knowledge in relation to science (Lin et al., [13]; Lin & Lin, [12]; Spiegel et al., [20]).

Second, the undergraduate students in this study reported a moderate overall level of self-efficacy regarding learning from comics, which might be attributed to the concern that they could be distracted by information, which is of unrelated subject-matter in the text. Nevertheless, the eye-movement results revealed that they were competent in distinguishing between crucial and unimportant information from the scientific comics. It was supported by the evidence that the readers in this study spent more time on total fixations regarding the important information of the conceptual text and the boxed-in text areas (e.g., the causes of cancer, the process of metastasis, possible treatments) than on the less important information of non-conceptual text (e.g., conversation between people) and non-conceptual areas (e.g., people, building, decorative graphs), even though the non-conceptual text and graphic areas had larger areas in the scientific comics. Additional evidence was that the re-reading time for irrelevant dialogs (i.e., non-conceptual texts) was much shorter than the first-pass fixation durations, suggesting that the participants did not see the need to re-read irrelevant or unimportant information when it was presented alongside useful messages in educational comics.

Third, no difficulty with regard to encoding was observed among the participants of this research study in relation to the selected science comics, despite the use of scientific vocabulary. This finding is supported by the comparable mean fixation durations for the dialogue between the characters (i.e., non-conceptual texts, which did not require conceptual understanding), elaborations on the diseases (i.e., conceptual texts), and extra medical knowledge (i.e., boxed-in texts), which were about 236 ms. This duration was very similar to a classical literature review of the eye movement study, which reported that 225–250 ms is an average fixation duration for adult readers while reading (Rayner, [19]).

Fourth, regarding the association between the eye-movement patterns and the post-reading comprehension performance, the total fixation durations did not significantly promote post-reading comprehension. However, separating the total durations into the first-pass and the re-reading fixation durations revealed that only the re-reading fixation durations could promote post-reading comprehension. Particularly, those who scored higher in the post-test exhibited a tendency to have re-read the boxed-in texts and graphics for a longer time. It indicated that besides the text information, graphs are also crucial for reading comprehension for learning science knowledge. Von Reumont and Budke ([24]) also found that the reading time on the graphs of the map was associated with post-test performance.

This study collected subjective (e.g., the questionnaire of measuring the readers' attitudes, habit, and motivation) and objective data (e.g., the post-test and eye movements) regarding scientific comic reading and investigated the relationships between these data. In conclusion, this study had some theoretical and instructional contributions.

For the theoretical contribution, this study provided empirical evidence for depicting Farinella's ([4]) statements, which stated how there were three advantages of using comics for education, namely, visualization, storytelling, and metaphors through online eye movement data. Regarding the visualization and metaphors, Farinella explained that the scientific comic constructs a more concrete structure for abstract concepts and can be divided into smaller and understandable units, as many narratives in comics are based on visual metaphors. In this study, compared to the poor-test performance readers, the good-test performance readers spend much more time on the areas of boxed-in textual (e.g., cell division in cancer cells; normal cells in our body grow, divide, and die; the cycle keeps the cell number at an optimal level) and graphical (e.g., depiction of two contrast graphs of normal cell divisions and cell divisions in cancer cells) information, which describes and depicts important science concepts for comprehension. This result indicated that good readers were capable of distinguishing which information is more or less important regarding the scientific comics. The multiple representations that correspond to the dual-coding theory could also be used (Paivio, [17]) regarding reading scientific comics and comprehension of their content. For the storytelling, the readers in this study spent lot of time reading on the narrative of the conceptual text area (e.g., As we grow up, the number of cells increases, so does our height and weight.) as opposed to the non-conceptual text area (e.g., Oh no... This means cancer could spread to other parts of the body!), which reflected that the undergraduate readers could be engaged and immersed in scientific knowledge as well as in the important narratives. In Table 3, the mean of the total fixation durations spent, for all the participants, on the conceptual descriptions (occupied 15% areas of the whole science comic) was 71.04 s, and only 28.68 s on the non-conceptual descriptions (occupied 11% areas of the whole science comic).

For the instructional contribution, the questionnaire survey of the readers' attitudes towards scientific comic reading gave a chance for the teachers to know how often the students read (knowledgeable) comics, how interest in reading comics developed, and how easy it was for them to learn about science from reading scientific comics et al. In addition, the data of this study indicated that science graphs are crucial for reading and comprehension with regards to learning scientific knowledge. This was confirmed by a previous study (von Reumont & Budke, [24]). Therefore, it is essential for teach students to pay attention to the science graphs and decode them, especially for those students with poor-test performance.

Despite the above contribution, this study experienced a few limitations and shed some light in further research with regard to reading comics. Firstly, although some researchers make use of comics to perform scientific knowledge which might had some advantages (e.g., concrete visualization, narrative that assists understanding, potential of personification for logical reasoning), there are some drawbacks which might be a concern, such as faulty deductions and misconceptions (Jee & Anggoro, [7]; Tatalovic, [22]). However, this study did not measure the readers' misconceptions or mistaken understanding. Furthermore, this research may add a measurement to investigate this problem. Secondly, due to limited participants in this study, only a few participants (N = 10) expressed that they disliked reading comics. Further research may recruit more participants to be examined if the readers have different interests (not interested, neutral, interested) towards reading comics and had different eye movements and test performance results while reading scientific comics.

This research was financially supported from the grant MOST 111–2636-H-003–009- under Young Scholar Fellowship Program by Ministry of Science and Technology in Taiwan, NSTC 111–2410-H-003–013-MY3 by National Science and Technology Council in Taiwan and the "Institute for Research Excellence in Learning Sciences" and "Higher Education Deep Cultivation Project" of National Taiwan Normal University (NTNU), sponsored by the Ministry of Education, Taiwan.

The data that support the findings of this study are available from the corresponding author, upon reasonable request.

The author declares no competing interests.

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.