Using Eye-Tracking in Education—A Review Study

Abstract

Visual perception is a complex psychological operation that is used to understand our environment. Its measurement gradually penetrates various areas of human life as well as the educational process. The presented study has a theoretical–analytical character; its aim was to identify the nature of publications that are devoted to the issue of eye-tracking focused on the educational process. The selection of the analyzed articles was carried out in the Web of Science database in January 2021, and all peer-reviewed articles (marked articles) that were in the Education category were included in the Educational Research database. Those texts were further analyzed with the use of the generally recognized PRISMA guidelines. The number of studies that are focused on eye-tracking had an upward trend, studies from Europe were the most often represented, the authors focused on reading comprehension, and the most common sample were university students. Implications for practice and suggestions for further research can be found in the conclusion.

1. Introduction

Visual perception is a complex psychological process used to understand our environment. Its measurement is gradually penetrating various areas of human life, including the educational process. Essentially, it can be assumed that tracking eye movements primarily aims to analyze the ability to read and process information, as described in the article by Merchie et al. (2021), which addresses reading difficulties among children in late primary education in relation to mind maps.

Today, eye-tracking devices are so precise that researchers can investigate whether students fixate on specific words, sentences, or parts of texts (Jarodzka & Brand-Gruwel, 2017). This capability enables the use of eye-tracking across educational domains to examine, for example, whether students focus on relevant or irrelevant text information, central or peripheral ideas, or the topics of leading or concluding sentences in a text (Merchie et al., 2021).

The use of eye-tracking in education is connected to various theories. One example is Cognitive Load Theory, which addresses how working memory—a limited-capacity processing system—impacts learning. It suggests that excessive information (cognitive load) in working memory can hinder learning and retention (Schnotz & Kürschner, 2007). Another relevant theory is Dual Coding Theory, which proposes that humans process information through two distinct systems: a verbal system for language and a nonverbal system for images. This theory posits that combining verbal and visual information enhances memory retention and comprehension because the interconnected nature of these systems strengthens encoded information (Paivio, 1991). Lastly, the Cognitive Theory of Multimedia Learning (Mayer, 2024) seeks to explain how people learn academic material from words and graphics and has developed significantly over the past four decades.

2. Eye-Tracker

In the 1950s, the first contact lenses, including coils, were used to accurately record where a subject was looking (Holmqvist et al., 2011). This method was quite invasive and demanding. One of the earliest studies in history tested army pilots during landing (Fitts et al., 1950). It is possible to differentiate between the interactive and diagnostic uses of eye-tracking. In the context of interactive use, eye-tracking refers to computer control, where gaze replaces or supplements common peripherals, such as a keyboard or mouse. In diagnostic applications, eye-tracking was primarily used in psychology. Nowadays, eye-tracking is widely used across various domains due to its significant advantages. For example, a participant’s point of view can be calculated from the collected data and displayed in a first-person video documentation known as a “scan path”. Comparing different objects and points of interest can be further explored by defining areas of interest (AOIs). Using quantitative AOI-specific measures, such as dwell time—the total time spent looking at an AOI—it is possible to draw conclusions regarding visual behavior and attention (Weiss et al., 2021).

Another reason for the recent boom in eye-tracking is easier access to equipment (King et al., 2019), which naturally leads to an increase in research in this area (Titz et al., 2018). Despite its long history, eye-tracking can still be considered a relatively new method (Jarodzka & Brand-Gruwel, 2017; Wyss et al., 2021). Eye-tracking devices can be classified as either fixed or mobile. The advantage of mobile devices, such as eye-tracking glasses, over fixed ones is that they do not restrict the participant’s movement, allowing more natural behavior. For example, in education, eye-tracking is primarily used to study the professional vision of teachers (Wolff et al., 2016). Head-mounted displays (HMDs) have become more affordable and lighter in recent years, facilitating wider use of virtual and augmented reality (VR/AR) applications (Kapp et al., 2021). This enables entry into mixed reality environments, as these devices can optimize display quality through techniques such as foveated rendering. Regarding the aforementioned topic, it is interesting to consider the study by Stein et al. (2021), who compare the eye-tracking latencies of several commercial head-mounted displays. The authors note that a number of VR HMDs with integrated eye trackers have recently become commercially available.

Eye-Tracking in Educational Research

The process of eye-tracking is used across various domains of education, for example, in both cognitive and affective areas (Konieczny & Döring, 2003), and it influences multiple aspects of the educational process, including subjects such as mathematics (King et al., 2019; Schindler & Lilienthal, 2019; Strohmaier et al., 2020). Researchers typically begin with the assumption that mathematical objects are not directly accessible but only through their representations (Duval, 2006). From this premise stems the uniqueness of mathematics, which lies in how texts (axioms, theorems, and proofs), mathematical symbols, visualizations, and their usage are processed (Andrá et al., 2015). Individual representations are thus directly dependent on cognitive processing, for which eye movements can provide valuable insights (Holmqvist et al., 2011).

The use of video in teaching (Beege et al., 2017; Tarchi et al., 2021; Deng & Gao, 2023) is another area where eye-tracking is applied. Typically, researchers focus on their own video designs rather than on individual differences, as evaluating and mapping eye movements in this context is quite complex. A key question discussed in detail is “How does the use of eye-tracking technology and interpretation of eye-tracking metrics advance our understanding of the mechanisms underlying effective video-based learning?” (Deng & Gao, 2023). Based on an analysis of 44 studies, they concluded that eye-tracking metrics are not a panacea for analyzing unconscious processes during video-based learning. Referring to Smith et al. (2019), they note that the most significant limitation is that situational interest is a psychological state characterized by increased affect, attention, and aconcentration during student engagement. On a broader level, eye-tracking can be used to manage a classroom (Coskun & Cagiltay, 2021) by analyzing individuals’ non-verbal expressions (Lemmer et al., 2012; Barati, 2015), with the lecturer’s eye movements considered a key non-verbal cue. The analysis is conducted in synchronization with cameras, and the RTA (Retrospective Think-Aloud) sessions are recorded. It has been shown that wearable eye trackers are promising tools for classroom research and, more importantly, can serve as feedback tools to enhance teacher performance (Cortina et al., 2015).

Vocabulary development (Kang et al., 2022; Pellicer-Sánchez & Siyanova-Chanturia, 2018) in various linguistic subjects is yet another area where eye-tracking is applied. Researchers note a strong contrast between the eye-tracking method and traditional reading methods, such as reading accompanied by underlining or note-taking. Eye-tracking is thus better suited to revealing an individual’s natural reading behavior. For example, it has been found that readers pay more attention to low-frequency words than to frequently occurring ones (Mohamed, 2018).

There are many educational applications of eye-tracking. Examples include the following:

(a)

Investigating students’ geometric misconceptions (Uygun et al., 2024);

(b)

Special education (Donmez, 2023);

(c)

Predicting the difficulty level of spatial visualization problems (Li et al., 2020);

(d)

Interactive virtual environments (Ugwitz et al., 2022), among others.

3. Research Focus

Research studies on eye-tracking have a rising trend and touch more and more areas. With the growing number of research studies focused on the mentioned topic, the scientific field is gradually asking for summarizing studies of a theoretical–analytical nature. Based on the above, the following aims were set:

-

To find out whether there was an increase in the number of publications with respect to the year of publication, and also to determine the influence of other categories with respect to the year of publication;

-

To determine the representation of studies with regard to the nationality of the respondents and determine the influence of other categories with regard to the nationality of the respondents;

-

To define the representation of studies with regard to the level of education and determine the influence of other categories with regard to the level of education;

-

To determine the representation of studies with regard to the focus of the topic;

-

To define the representation of studies with regard to psychological phenomena.

4. Methods

4.1. A Selection of Articles

A selection of articles was conducted within the Science database in January 2021. Due to the relative novelty of eye-tracking, no time limit was set (the first article is from 2007), and all peer-reviewed articles (marked article) that were in the database were found in the education category Educational Research. Data collection for the research part took place during 2021, when 150 studies were found according to the specified criteria. The data was then supplemented in January 2023 with studies that were published in 2022, a further 61 studies (Figure 1 and Figure 2). The checklist for PRISMA was filled and a PRISMA model was created according to the guidelines (Page et al., 2021).

The following key word combinations were used to find the texts:

Eye tracking (Title);

Eye tracking (Title) AND Eye-tracking (Title).

The found texts were further analyzed using the generally recognized PRISMA guidelines (Page et al., 2021), intended for the implementation of review studies and meta-analyses. The final criterion for the selection of the analyzed articles was whether they were original research texts or not. For each of the searched studies, it was firstly examined, according to the abstract, whether it is an original research study or not. If it was not obvious from the abstract at first glance, the study was read in its entirety. Out of 211 studies, 11 studies were excluded based on this criterion.

4.2. Geographical Restrictions and the Coding Process

It was the aim to obtain the most comprehensive data possible. Therefore, the studies were not limited geographically, and all available studies from around the world were included. The coding process was carried out in two phases: first, open (exploratory) coding, during which the main themes and variables (e.g., type of research, educational level, and application of eye-tracking) were identified; then, categorization (axial) coding was performed, where individual occurrences were assigned to standardized categories. Codes were created iteratively and adjusted according to the occurrence of recurring patterns. Microsoft Excel was used to manage and organize the coding data with a matrix of coding fields that included more than 15 variables. Excel was supplemented by manual triangulation between two authors, who conducted a parallel check of a random sample of 20% of the studies to increase the reliability of the coding.

4.3. Time Limit

Given the relative novelty of eye-tracking, no time limits were set either. The last search of articles was carried out in January 2023. At the same time, this overview study had no restrictions on primary research. The oldest study found is therefore from 2006, and the latest studies are from the end of 2022. This is an overview of studies that were carried out over a period of 17 years.

4.4. Language Restrictions

Regarding the language of the research studies, we included articles in all languages in our analysis. For the most part, these were studies published in English. In particular, newer studies recently added to the database have so far been published only in their original languages. This meant that it was necessary to work with studies in other languages as well. Concerning research in Czech, at the time of our study, only two Czech studies were available in the database. No biases were observed regarding differing perspectives on the issue of eye-tracking. Part of the text was published in English, so no translation was required.

4.5. Data Analysis

In the first step, articles were searched using the selected database. The articles were gradually all tracked down and downloaded. Subsequently, individual articles were studied and determined whether it was an original research study. This was followed by a list of data by following categories: year of publication, magazine, nationality of respondents, number of respondents, level of education, focus of the article, and eye-tracker used. Examples of coding can be found in

Table 1. Example of coding of analyzed studies.

Category	Explanation	Code Examples
Year of publication	Year of publication in which the article was published	2006 and 2011
Magazine	The magazine where the article was published	Computer-Assisted Language Learning and Second Language Research
The nationality of the authors	Country where the authors come from	Germany, Finland, China, and Japan
Number of respondents	The number of participants who participated in the research as respondents	15, 25, and 326
Level of education	Level of education of the participants at the given time	University, Kindergarten, Primary School, and Secondary School
Article content	What topic was the article about? Aimed at what was examined by the eye-tracker	Geometry, bilingualism, metacognition, and the process of reading of the pupils in kindergarten
Used eye-tracker	Which machine they used to research	Tobii T120 and SMI RED 250

.

After data processing, the article content categories were decoded by dividing them into two additional subcategories. First, we examined the focus from the perspective of the topic, which included languages, mathematics, reading, multimedia, pedagogical approach, special education, natural sciences, and medicine. Second, the studies were further classified according to the psychological phenomena investigated, such as learning, attention, reading comprehension, thinking, behavior, metacognition, and evaluation. Subsequently, the results from all categories were processed and coded numerically. Their graphical representation can be seen in the results chapter. Furthermore, the relationships between the selected categories were tested using the chi-square test. This test assesses whether the observed frequencies in the data collected from pedagogical settings differ significantly from the theoretical expected frequencies—i.e., whether some categories are affected more or less than expected. There is only a brief mention of inter-coder reliability. Inter-coder reliability ensures consistency among researchers when categorizing or labeling data in qualitative analysis. This is particularly crucial in review studies where multiple coders interpret complex datasets. By measuring agreement, it safeguards against subjective biases and enhances the credibility and reproducibility of findings. In this study, inter-coder reliability was ensured through inter-rater agreement on the nominal codes, with an agreement rate of 100%.

Additional results were obtained through inferential statistics, specifically the chi-square test of independent samples (χ²). The chi-square test of independence is used to determine whether two categorical variables are related or independent. It is a non-parametric test that examines whether the observed frequencies in a contingency table differ significantly from the expected frequencies if the variables were truly independent (Franke et al., 2012).

5. Results

Regarding the set aims of the research, the monitored variables are presented in this chapter. The first variable is the year of publication. As mentioned earlier in the article selection process, the earliest study dates back to 2006, and the most recent studies are from 2022. Thus, the sample covers eye-tracking research focused on education over the past 17 years. In 2006, only one study was published. Similarly, in 2007, there was again just one published study. In contrast, in the most recent year, 34 studies have been recorded in the database so far. A clearer picture of the gradual increase in the number of studies can be seen in the figure, where publications are grouped by intervals. Figure 3 shows that during the first interval, research in the field of eye-tracking focused on education was quite minimal. However, in the most recent four-year interval, these studies accounted for nearly 60% of all the studies found. To put this in numbers, that represents 118 studies out of a total sample of 200.

Another element this study focused on was the origin of the research teams. For each study, the countries of the authors were identified—that is, which countries are attempting to use eye-tracking in educational research. Most studies originated from Europe. An interesting fact is that the oldest study included in the sample, from 2006, was published by authors from the USA. It should also be noted in detail for each continent. In Europe, authors from Germany had a clear lead. Out of a sample of 200 studies, 33 were authored by German researchers or, to a lesser extent, mixed teams involving collaboration between authors from the USA, Sweden, and Switzerland. From the Americas, the largest number of authors came from the USA, with 21 studies. This country ranked second overall in terms of the number of studies by authors’ nationality. Among Asian countries, Taiwan had the largest representation, ranking third in the list of countries by authorship, with 20 author teams.

Furthermore, it was interesting to observe the thematic focus of the research articles. During the investigation, the direction and topics of each study were recorded in detail and then categorized. Unfortunately, some studies did not fit into any of the selected categories due to vague or uncategorizable topics. Therefore, 12% of the studies were classified as ambiguous data. At first glance, it is clear that a large percentage of studies focus on education—27% specifically on languages. This category included studies investigating how eye-tracking can aid in teaching grammar and vocabulary in both foreign and, in some cases, native languages.

Multimedia was another prominent category, with 19.5% of the articles exploring how to create multimedia teaching aids and how to integrate multimedia into education effectively. An interesting category, representing about 6% of the studies, was the pedagogical approach category, which included research on where teachers direct their attention while teaching.

We also applied another categorization based on psychological phenomena monitored in the studies. In this categorization, only 8% of the samples were non-specific, meaning it was not possible to clearly assign them to a category. Studies focusing on the learning process predominated, representing 39% of the sample. Another significant category was reading comprehension, accounting for 21.5% of the studies. These studies addressed reading comprehension in both native and foreign languages, mainly focusing on early childhood and primary education levels. In contrast, studies investigating behavior were underrepresented, comprising only 0.5% of the sample.

Another aspect analyzed was the education level of the participants—i.e., the type of school the respondents attended during the research. This information was often missing from abstracts, so it was researched in detail within the full texts. Even so, as shown in Figure 4, this data was unavailable in 38% of the studies—the highest percentage of missing data among all categories examined.

The “mix” category included studies conducted across multiple educational levels, often combining high school and university students, or involving both students and teachers simultaneously. These studies typically used eye-tracking to investigate where teachers and students focus their attention during lessons. The “adult” category included studies with practicing teachers and research on specialized educational courses, such as military training. Overall, aside from the unavailable data, the higher education category was the most represented, with 35.5% of the studies. The fewest respondents came from kindergartens, accounting for only 2.5%.

The mutual relationships between individual variables were examined using inductive statistics. The aim was to determine whether certain variables were more prevalent in published studies based on the authors’ origins or the educational level of the respondents at the time of the research.

Regarding the variable of the year of publication, no significant differences were found in relation to the other observed variables, which are listed in

Table 2. χ² values of the goodness-of-fit test taking into account the relationship of the investigated variables with respect to the year of publication.

	The Nationality of the Authors	Level of Education	Topic Focus	Psychic Phenomena
χ2	7.82	23.15	18.72	25.08
p	0.55	0.08	0.60	0.12
Cramer’s V	0.114	0.196	0.177	0.204

. As shown in Table 1, the relationships between variables were insignificant, indicating that the representation of the examined factors remained consistent regardless of when the study was published.

Another variable was the authors’ country of origin, but no significant difference was identified with regard to the variables of interest (

Table 3. Values of the χ² test of agreement taking into account the relationship of the investigated variables with respect to the origin of the authors.

	Number of Respondents	Level of Education	Topic Focus	Psychic Phenomena
χ2	12.23	15.13	28.42	31.14
p	0.25	0.27	0.13	0.07
Cramer’s V	0.143	0.112	0.154	0.149

).

While the p-values presented in Table 2 and Table 3 do not indicate statistically significant relationships (all p > 0.05), it remains informative to consider the effect sizes as indicated by Cramer’s V. In social science research, a Cramer’s V of approximately 0.10 is typically interpreted as a small effect, around 0.30 as medium, and 0.50 or higher as a large effect (Cohen, 1994). In our data, the strongest observed association was between the nationality of authors and the psychological phenomena studied (Cramer’s V = 0.149, p = 0.07), suggesting a small but potentially meaningful trend. This is supported by the findings in Figure 5, where U.S.-based studies more often focused on evaluation-related processes, while Asian studies prioritized reading comprehension—patterns that may reflect underlying cultural or academic emphases. Another noteworthy relationship appears between the year of publication and educational level (Cramer’s V = 0.196, p = 0.08). Although not statistically significant, this small-to-moderate effect may imply a gradual diversification in study populations over time, with growing interest beyond university student samples. These observations suggest that even in the absence of statistically significant chi-square values, certain categorical associations in the dataset may hold interpretative relevance. Therefore, effect sizes like Cramer’s V provide essential additional insight into the data structure, complementing p-values by indicating the strength of association between variables. From Figure 5, however, it can be read that within the framework of the psychological phenomena variable, using the z-score, it was found that in studies by authors of American nationality, evaluation-oriented processes were investigated through the eye-tracking method more than expected (p < 0.01). Further, in the studies by the authors from the Asian environment, the processes aimed at reading comprehension were investigated using eye-tracking research more than expected (p < 0.05).

No significant difference was identified in the analysis of other variables, so further results are not presented.

6. Discussion

The main aim of this study was to map the existing research on the use of eye-tracking in education and thereby create an overview that acquaints readers with current research trends in this field. Based on the available information, it appears that no recent comprehensive mapping has been conducted. A summary study by Lai et al. (2013) included research on eye-tracking in education from 2000 to 2012. Additionally, only review studies focusing on the use of eye-tracking in specific subjects were found, such as mathematics (Strohmaier et al., 2020), special education (Chita-Tegmark, 2016; Papagiannopoulou et al., 2014), teacher behavior (Beach & McConnel, 2019), medicine (Ashraf et al., 2018), and driving license exams (Kapitaniak et al., 2015). Data for this study were collected using the Web of Science database, a widely used resource for publishing research articles across all disciplines. This broad scope was especially important given the thematic focus of the study. The overall picture of eye-tracking use in education is detailed through the monitored variables, which are presented in the next section of this chapter.

An upward trend in the number of studies was observed within the publication year category. A similar growth trend was noted by Hahn and Klein (2022) in their study on eye-tracking research in physics education. Their earliest paper was published in 2005, and their mapping extended to 2022, covering multiple databases simultaneously. Comparable findings were reported by Kiefer et al. (2017), who investigated pupils’ spatial perception using eye-tracking. These authors also noted an increasing number of research studies as eye-tracking becomes a more widely used research technique.

Building on this upward trend, Lai et al. (2013) also observed a steady increase in the use of eye-tracking in education. This growing interest may be partly due to eye-tracking being a relatively new research method, with its potential applications still evolving. The technological development of eye-tracking is advancing steadily, which likely contributes to its increasing use. For more information on technological advancements in eye-tracking, see Harezlak and Kasprowski (2018) or Klaib et al. (2021). However, as mentioned in the theoretical section, eye-tracking research remains expensive, particularly due to the cost of laboratory equipment. It can be assumed that fewer laboratories were equipped for this research in earlier years, resulting in fewer studies. The number of eye-tracking studies in education is influenced by this factor rather than by typical educational processes, such as maintaining attention (Brunye et al., 2019).

The findings presented in the results section of the manuscript are connected with cognitive theories introduced in the theoretical background: Cognitive Load Theory, Dual Coding Theory, and the Cognitive Theory of Multimedia Learning. Cognitive Load Theory (Sweller, 1988; Sweller et al., 1998) provides a valuable lens through which to understand how learners process visual and textual information. It suggests that instructional design must manage the limited capacity of working memory, and that excessive cognitive load—such as processing overly complex visuals or poorly integrated media—can hinder learning. This is particularly relevant in studies using eye-tracking to evaluate whether learners attend to extraneous information or split their attention between misaligned text and visuals. Dual Coding Theory, developed by Paivio (1986, 1991), offers another useful perspective, proposing that verbal and visual information are processed through separate but interconnected cognitive channels. Studies that monitor gaze behavior during reading or multimedia interaction often align with this framework, as they explore how learners coordinate written language with graphical or spatial representations. Finally, Mayer’s Cognitive Theory of Multimedia Learning (Mayer, 2001; Mayer & Moreno, 2003) synthesizes and extends both previous models by emphasizing the need for coherence, redundancy avoidance, and temporal and spatial contiguity between multimedia elements. Eye-tracking metrics—such as fixation duration or saccade paths—provide empirical insight into whether learners engage with relevant instructional elements or are distracted by poorly designed materials. By incorporating these frameworks, we aim not only to interpret the reviewed studies more robustly but also to situate our findings within a broader interdisciplinary dialogue among cognitive psychology, instructional design, and educational technology.

However, many researchers focus on success in conjunction with eye-tracking. The mentioned studies revealed the influence of focused attention on success in selected subjects, and because this proved to be a highly beneficial area, the number of research studies on this topic has increased (Tsai et al., 2019; Yang et al., 2018; Zawoyski et al., 2015). A closer examination of the reviewed studies reveals several thematic and methodological inconsistencies that warrant critical reflection.

For instance, Deng and Gao (2023) emphasized that while eye-tracking is increasingly used to study video-based learning, the diversity of experimental setups and inconsistent interpretations of eye-tracking metrics limit generalizability. Similarly, Ashraf et al. (2018), in a systematic review of eye-tracking in medical education, highlighted the scarcity of standardized procedures and the overreliance on descriptive findings, which makes cross-study comparisons difficult. Furthermore, Alemdag and Cagiltay (2018) pointed out that studies in multimedia learning often fail to align theoretical constructs (e.g., cognitive load or multimedia principles) with the metrics used, leading to interpretative ambiguity. These contradictions and limitations suggest the need for more theory-driven research designs, greater methodological transparency, and efforts toward standardizing the use of eye-tracking in educational contexts.

Why is methodological transparency crucial for advancing eye-tracking research in education? The nationality of the research teams in the studies reviewed was also examined, specifically identifying the countries from which the researchers originated. The results showed that most authors were from Europe. This may be because the earliest research cases, which can be considered predecessors of modern eye-tracking studies, originated in Europe. As mentioned in the theoretical section, this method was first used by the French ophthalmologist Javal (Leggett, 2010).

In 2018, Alemdag and Cagiltay published a systematic review mapping eye-tracking research with a focus on multimedia learning (Alemdag & Cagiltay, 2018). An interesting finding emerged when comparing their review with ours: the distribution of studies by the nationality of authors was similar, showing comparable representation across continents and countries. This suggests that eye-tracking research is concentrated in countries that invest substantial financial resources in research. Consequently, studies from African countries or some Asian countries with lower gross domestic product are scarce. These observations indicate that eye-tracking research is predominantly conducted in developed European countries and the USA.

Regarding the education level of participants, most respondents were university students. This aligns with Alemdag and Cagiltay’s (2018) findings, where university students also represented the largest group in multimedia learning studies. Other participant groups had minimal representation compared to university students.

To facilitate a more systematic categorization of future research, we propose a tentative framework for classifying educational eye-tracking studies along two core dimensions:

• Educational Level: This groups studies based on the target learner population, ranging from early childhood education, primary and secondary schooling, to university and adult education.
• Cognitive Function Examined: This includes functions such as attention allocation, reading comprehension, visual reasoning, metacognition, and decision-making.

To further support this conceptual framework, we visualized the intersection between education levels and cognitive functions of the studies analyzed in the present review (see Figure 6).

Mapping existing studies along these two dimensions reveals clear patterns in the current literature and brings to light underexplored intersections—as illustrated in Figure 6. The heatmap visualization provides a structured overview of how eye-tracking research is distributed across educational levels and cognitive functions. It confirms that areas such as reading comprehension and attention have been extensively studied, particularly at the university and primary education levels. In contrast, other combinations—for example, metacognitive strategies in elementary education or decision-making processes in adult learners—remain notably underrepresented. This visualization not only reinforces the conceptual framework but also serves as a practical tool to guide future research design and synthesis in the growing field of educational eye-tracking. These visual patterns resonate strongly with the statistical results obtained from the chi-square analyses presented earlier (Table 1 and Table 2; Figure 5). For instance, the heatmap’s emphasis on the intersection between university-level education and reading comprehension is mirrored in the statistically significant overrepresentation of this pairing in the chi-square analysis (p < 0.05). Likewise, Asian-authored studies showed a disproportionate focus on reading comprehension (p < 0.05), a pattern also visible in the heatmap via the high intensity in the university–READ cluster.

Conversely, the sparsely populated zones in the heatmap—such as metacognition at the kindergarten level or decision-making in secondary education—align with statistically negligible frequencies (Cramer’s V < 0.12), reinforcing the notion that these topic–population intersections remain severely underexplored. By synthesizing inferential statistics with visual pattern recognition, the heatmap effectively validates the statistical conclusions and highlights systemic gaps in current research emphases. This integrative approach underscores not only the reliability of the statistical models used but also the diagnostic utility of the heatmap as a strategic tool for future research planning. Research with university students tends to be more straightforward, as these participants generally understand study requirements and can maintain attention throughout experimental procedures. Younger participants, by contrast, often struggle to sustain attention over extended periods, making eye-tracking studies with them more complex and demanding. Consequently, many researchers prefer university students as their study sample. This tendency raises an important question: could the use of eye-tracking with younger learners influence their motivation to learn or even their value systems? Phenomena such as motivation and values can indeed be explored using eye-tracking technologies. Although the body of research in this area remains limited, some studies have begun to investigate value systems through various eye-tracking methodologies (e.g., Fiedler et al., 2013). When comparing the thematic focus of our review with that of Alemdag and Cagiltay’s (2018) review on multimedia learning, both highlight multimedia learning as a dominant area. However, their domain-specific analysis showed that studies in physics, biology, and other natural sciences were the most frequently represented, whereas language studies—predominant in our review—were absent from theirs. Additionally, while both reviews included studies focused on reading, which has traditionally been a core focus of educational eye-tracking research, the scope of such studies remains limited. Psychological phenomena, like those mentioned above, continue to be rarely explored and remain marginal within the field—a trend reflected in both the heatmap (Figure 6) and our broader analysis. This observation helps to explain the overall scarcity of related research and highlights opportunities for future investigation.

Limitations of the Study

One limitation of this study is the exclusive use of the Web of Science database, which may have resulted in the omission of relevant studies indexed in other databases, such as Scopus or ERIC. This choice was made to ensure consistency, maintain publication quality, and utilize tools for efficient data analysis. Future research should consider including additional databases to achieve broader coverage of the research field. Another limitation is the search strategy, which focused exclusively on article titles. This approach may have excluded relevant studies that address the topic less explicitly. However, this limitation was deliberately chosen to ensure high thematic relevance and to provide an overview of mainstream research. Additional limitations arose during the analysis and evaluation of the data. Nearly all the data collected were nominal in nature, which justified the use of the chi-square test. The interval variable involved the number of respondents in the selected studies; however, this was only a marginal variable with minimal influence on the overall and specific findings. For this reason, other parametric statistical methods, such as Analysis of Variance (ANOVA) and Spearman’s correlation coefficient, were not applied.

7. Conclusions

7.1. Implementation in Pedagogical Practice

The findings of this review suggest several practical directions for the educational use of eye-tracking. First, eye-tracking research offers valuable insights into how students engage with texts and multimedia content, particularly regarding visual attention and cognitive load. For example, placing related text and images close together on a page has been shown to support more efficient processing and learning (Van Gog et al., 2009). In teacher education, wearable eye-trackers are increasingly used to help future teachers develop what is known as professional vision—the ability to notice and interpret important classroom interactions. Watching and reflecting on one’s own gaze data has proven effective in raising teachers’ awareness of where their attention goes during instruction (Wolff et al., 2016; Cortina et al., 2015). Moreover, eye-tracking has the potential to support adaptive learning systems. By analyzing how learners focus on specific parts of the screen, these systems can provide personalized feedback or adjust the pacing of content delivery in real time (Jaarsma et al., 2022). This opens up promising applications for digital learning platforms, especially in environments where individualization is essential. As these technologies evolve, deeper collaboration between educators, cognitive scientists, and instructional designers will be necessary to turn empirical insights into usable teaching tools (Lai et al., 2013).

In addition to its theoretical implications, the review also suggests several concrete ways in which educators and educational technology developers can apply the findings of eye-tracking research. For teachers, one of the most direct applications lies in the development of professional vision—a skill that can be cultivated by analyzing where attention is directed during classroom interactions (Wolff et al., 2016). By using gaze-based feedback, novice teachers can train their awareness of student engagement, board scanning patterns, and classroom dynamics, which can improve instructional decision-making in real time. For instructional designers and developers, the implications are equally promising. Eye-tracking data can inform the design of more intuitive and cognitively efficient learning materials—for example, by identifying which layouts lead to information overload or which multimedia combinations support better retention (Mayer & Moreno, 2003; Jaarsma et al., 2022). In adaptive systems, real-time eye-tracking can be used to adjust difficulty, highlight overlooked content, or pause instruction when attention drifts—making personalized learning environments more responsive and effective.

Eye-tracking is not only a research tool but also a driver of pedagogical and technological innovation. The use of eye-tracking could be helpful in incorporating instructional design principles to create effective and engaging learning experiences. These principles include understanding your audience, setting clear objectives, using diverse learning materials, incorporating interactivity, and ensuring accessibility. Additionally, focusing on learner experience, utilizing storytelling techniques, and continuously evaluating and refining the design are key. Teaching interventions informed by eye-tracking could enhance student learning processes, reading skills, and cognitive engagement, potentially improving personalized instruction and identifying learning difficulties. Platform development can also benefit from eye-tracking technology, which can be used to enhance software interfaces, tools, and development processes. Eye-tracking data can assess software artifacts, tools, and techniques, as well as inform developer recommendations and software traceability tasks.

7.2. Recommendations for Further Research

The work offers many stimuli for further pedagogical research. Future research paper authors may focus on any of the following points:

• To involve age groups other than university students, so that it is possible to compare results in other levels of education as well. This proposal is based on the use of the eye-tracker especially among university students;
• To explore the area dedicated to children with autism spectrum disorders more. Research activity in this area has already been started, but it is possible to further deepen it and focus especially on the diagnostic area;
• To continue in special pedagogy and explore the area where the eye-tracker serves as a sensing part of the screen and thanks to it, individuals with limited mobility can work with the computer using eye movements as if they were also involving their hands, which are often immobile.