5.1. Overall findings
By content analysis, it was found that the 81 papers (containing 113 studies) reviewed here discussed topics of learning such as perception, causal reasoning, social cognition, meaning-based representations, language, conceptual development, Table 3
The learning topics and their associated subtopics for content analysis.
Learning topic Associated subtopics
Perception Object/visual recognition, spatial relations, animated relations (continuous actions) Social cognition Awareness of self and others, recognizing intentional actions by others, social referencing Meaning-based
representation
Concepts of basic or natural objects, schema (true categories and episodic memories)
Language Phonological, lexical, grammatical, reading
Reasoning Causes and effects, causal principles, coordination of theory and evidence, hypothesis testing, integration of information, metacognitive reasoning
Conceptual development Basic-level/superordinate/subordinate categories or concepts (related to school subjects), conceptual change, domain-specific knowledge
Skill/strategy learning Domain-general skills, domain-specific strategies
Fig. 1. Bridging framework between eye movements and learning.
and skills/strategy learning. Among these topics, conceptual development received the most attention (40 studies), followed by perception (25 studies) and language (23 studies). Only a few studies were dedicated to causal reasoning, social cognition, and skills/strategy learning (3, 5, and 6 studies, respectively). The initial analysis suggested that when the eye movement method was applied to studies related to learning, the focus of discussion was largely on the acquisition of new information or conceptual knowledge.
As far as eye movement measures are concerned, temporal measures were the most frequently employed overall (138 times, given that some studies used more than one measure), followed by frequency measures (55 times) and spatial mea-sures (38 times). Rather than revealing basic cognitive mechanisms, eye movement meamea-sures, when applied to the studies related to learning, were used to disclose mostly the perceptual habits, mental efforts, and change or movement of attention on learning materials. In short, the eye movement data shown in the reviewed studies largely suggested real-time attention distribution. In due course, how exactly eye movement measures are related to the processes or outcomes of learning need to be further tested by studies incorporating adequate performance measurements. These measurements may include achieve-ment tests, behavioral observations, log records, rate of accuracy, response time, introspective reflections and so forth.
5.2. The bridging framework
As mentioned previously, in the last stage of the coding procedure, the cross-examinations of the research questions and eye movement indications of each study across different learning topics eventually abstracted 7 learning themes out of the 113 reviewed studies. These themes include (1) Patterns of information processing, (2) Effects of instructional strategies, (3) Reexaminations of existing theories, (4) Individual differences, (5) Effects of learning strategies, (6) Social/Cultural effects, and (7) Decision making patterns. Noticeably, some of the papers embraced two themes in their design. Accordingly, a bridg-ing framework was created as inFig. 1, showing the 7 generative learning themes that specify how the eye tracking tech-nology has been used to explore learning issues. In the following, an analysis of each learning theme is depicted.
5.2.1. Patterns of information processing
In our review, 53 studies were found to map the patterns of information processing. The majority of these works dis-cussed the learning topics of language (18 studies), perception (19 studies), and meaning-based representations (9 studies).
A few studies dealt with problems concerning conceptual development (4 studies), social cognition (2 studies), and psycho-motor learning (1 study).
The patterns of information processing investigated in these studies include how learners acquire new words or recognize learned words (Arnold, Eisenband, Brown-Schmidt, & Trueswell, 2000; Bolger & Zapata, 2011; Brandt-Kobele & Höhle, 2010;
Brown-Schmidt, 2009; Brusnighan & Folk, 2012; Chaffin, Morris, & and & Seely, 2001; Grüter, Lew-Williams, & Fernald, 2012;
Schmiedtova, 2011; Shatzman & McQueen, 2006; Wonnacott, Newport, & Tanenhaus, 2008), how learners control and dis-tribute their attention during visual searching, and how contextual, semantic or memory factors may affect object/visual rec-ognition and the processes of visual searching (Beesley & Le Pelley, 2010; Brockmole & Le-Hoa Võ, 2010; Castelhano &
Heaven, 2011; Goujon, Brockmole, & Ehinger, 2012; Guérard, Saint-Aubin, Boucher, & Tremblay, 2011; Karacan, Cagiltay,
& Tekman, 2010; Koenig & Lachnit, 2011; Mou, Liu, & McNamara, 2009; Roderer, Krebs, Schmid, & Roebers, 2012; Tremblay, 2011; van Asselen, Sampaio, Pina, & Castelo-Branco, 2011; Wang & Mitchell, 2011; Welham & Wills, 2011). At a more basic level of learning, how humans learn object-based categories and face recognition has been discussed (Blair, Watson, & Meier, 2009; Goldinger, He, & Papesh, 2009; Henderson, Williams, & Falk, 2005; Hsiao & Cottrell, 2008; Huestegge & Koch, 2012; Liu
& Chuang, 2011; Sekiguchi, 2011). As far as the higher level of learning is concerned, attention distributions during concept learning and problem solving have also been study targets (Bartolotti & Marian, 2012; Liu, Lai, & Chuang, 2011; Schneider, Maruyama, Dehaene, & Sigman, 2012). In addition, the few studies investigating social cognition and psychomotor learning reveal how learners learn to respond associatively to sensory (social) cues and how skills are obtained over time (Herwig &
Horstmann, 2011; Mu, 2010). In short, due to the process characteristic, eye tracking technology has been extensively used to depict the patterns of information processing on various levels of learning.
5.2.2. Effects of instructional strategies
As for the 26 studies focusing on the effects of instructional strategies, nearly all of these studies involved conceptual development (22 studies), while only a few involved psychomotor learning (3 studies) or language learning (1 study). Most of the studies concerned how to design multimedia for better learning support. Specifically, various cues, guidance, displays, controls and presentations were manipulated and examined for the use of multimedia, especially animations, in related studies (Boucheix & Lowe, 2010; de Koning, Tabbers, Rikers, & Paas, 2010; Johnson & Mayer, 2012; Meyer, Rasch, & Schnotz, 2010; Ozcelik, Arslan-Ari, & Cagiltay, 2010; Ozcelik, Karakus, Kursun, & Cagiltay, 2009; Schmidt-Weigand, Kohnert, & Glo-walla, 2010; Stieff, Hegarty, & Deslongchamps, 2011). Many studies focused on designs for text and graphic comprehension or problem solving (e.g.,Ariasi & Mason, 2011; Blythe et al., 2012; Hegarty, Canham, & Fabrikant, 2010; Lewis & Mensink, 2012; Liu & Shen, 2011; She & Chen, 2009). Several studies attempted to examine the use of various mind tools, such as con-cept maps (Amadieu, van Gog, Paas, Tricot, & and & Marine, 2009), video annotation systems (Mu, 2010), simulated touch function (Wiebe, Minogue, Jones, Cowley, & Krebs, 2009), focused attention training (Pradhan et al., 2011), and gaze-based tutoring systems (D’Melloa, Olney, Williams, and Hays, 2012). In sum, these studies used eye-tracking techniques to examine
student learning states while interacting with multimedia learning environments. The results of these studies can provide valuable feedback for individualized learning system designs.
5.2.3. Reexaminations of existing theories
Among the reviewed papers, we found that 14 studies aimed to reexamine the existing theories. Most of these studies were centered on the topics of conceptual development (7 studies) and perception (4 studies), while three studies were on the topics of meaning-based representation (1 study), language (1 study), and psychomotor learning (1 study). The theories under reexamination mainly included category learning (Blair, Watson, Walshe, & Maj, 2009; Rehder, Colner, &
Hoffman, 2009; Rehder & Hoffman, 2005a,b), processes of visual search in learning (Hout & Goldinger, 2011; Jones &
Kaschak, 2011), language learning (Whitford & Titone, 2012), and skill learning in complex tasks (Lee & Anderson, 2001).
These theories seem to reflect mostly the basic levels of learning.
5.2.4. Individual differences
A total of 15 studies were found to focus on examining individual differences among learners. Most of them involved con-ceptual development (6 studies) and language learning (6 studies). Others dealt with psychomotor learning (2 studies), pat-terns of decision making (2 studies) and meaning-based representation (1 study). It is noticed that the majority of these studies concerned higher-level cognition, such as conceptual development and language learning, comparing learners with different characteristics. These characteristics included prior language background (Bartolotti & Marian, 2012; Grüter et al., 2012; Tremblay, 2011), prior domain knowledge (Balslev et al., 2012; Kim & Rehder, 2011; Ooms, De Maeyer, Fack, Van Assche, & Witlox, 2012; Tsai et al., 2012), domain expertise (Balslev et al., 2012), prior experience (Sekiguchi, 2011) and age level (Blythe et al., 2012; Isaacowitz & Choi, 2012; Neider & Kramer, 2011). Those investigations which focused on back-ground knowledge or experience showed that the researchers are rooted in constructivist perspectives of learning, while those focused on age level showed that some studies examined learning from the developmental perspectives of learning.
5.2.5. Effects of learning strategies
Nine studies in this review used eye tracking technology to examine students’ learning strategies and the effects of these strategies. The learning topics in this category ranged from conceptual development (3 studies), psychomotor learning (3 studies), and perception (2 studies) to meaning-based representation (1 study). A diversity of discussions were covered in this theme, including, for example, how learners process text, graphic and spatial information (Guérard, Tremblay, &
Saint-Aubin, 2009; Smerecnik et al., 2010), how students attend to and select information during web searching (Miwa et al., 2011; Neider & Kramer, 2011), and how learners solve problems during complex tasks such as domain-specific multi-ple-choice tests (Tsai et al., 2012), multipart skill learning tasks (Lee & Anderson, 2001), and error judgment tasks (Wills, Lavric, Croft, & Hodgson, 2007). In short, studies in this theme emphasize the sequences of cognitive actions appearing in various learning and problem solving tasks.
Fig. 2. Number of studies with respect to each learning theme from 2000 to 2012.
5.2.6. Patterns of decision making
There were 3 studies discussing the patterns of decision-making in the scope of causal reasoning. Among them, one ex-plored how people retrieved memories during decision making (Renkewitz & Jahn, 2012) and the other two discussed how people of different ages and knowledge backgrounds make choices when given task relevant information or how they mon-itor their memory during a choice task (Miller & Cassady, 2012; Roderer & Roebers, 2010). In sum, studies on the theme of patterns of decision making often took into consideration cognitive and developmental constraints.
5.2.7. Social cultural effects
There were 3 studies discussing social cultural effects. Two of these studies discussed the cultural differences in process-ing faces of difference races (Kelly et al., 2011; Nakabayashi et al., 2012) and the other discussed the relation between atten-tion and social cogniatten-tion (Kim & Mundy, 2012). Obviously, the exploraatten-tion of the social-cultural effect is rare in eye movement studies related to learning.
It is worth mentioning that among the reviewed studies, we found 12 which involved more than one learning theme and, interestingly, the theme of individual differences was most often investigated with other learning themes, particularly the patterns of information processing, and the effects of learning strategies.
5.3. The research trend of the eye tracking method applied in learning-related studies
AsFig. 2shows, the number of eye movement studies related to learning has increased significantly since 2009. With re-spect to the 7 learning themes, over 40% of these studies focused on describing patterns of information processing. Mean-while, studies discussing the effects of instructional strategies have also grown significantly in the recent 3 years. Other than the abovementioned two themes, an increasing number of researchers are taking notice of individual differences. It seems that social/cultural effects and decision-making patterns are two themes that have just caught educational research-ers’ attention. Noticeably, these two themes are greatly studied in psychological research but without an educational focus.
The developmental trend of the research over a decade suggests that educational researchers may have just started to notice the wide extent to which the eye-tracking method can be applied in the study of learning issues. Our framework also points out that other than the extensive discussions on information processing and instructional strategies, there are many other learning themes that can be explored using eye tracking methods, but which still lack substantive investigations. Since eye-tracking technology is dominantly used in the domain of psychology, there should be a clear push for educational research-ers to make good use of this existing eye movement research. More convresearch-ersations or interactions between psychologists and educational researchers are needed to help extend the scope of the new research application in education.
5.4. The use of eye movement measures in learning-related studies
After analyzing the eye movement measures used in the reviewed studies, temporal measures were found to be used most widely for all the learning themes, followed by count measures, except for the themes of learning strategies and decision mak-ing patterns.Table 4shows the statistical results. This may be due to the fact that temporal and count data were easy to col-lect with the software available and were often used quantitatively. The spatial measures were the least used in the studies, which may be because the spatial indicators (e.g., fixation location and scan path) are typically analyzed in a qualitative way, which is more time and energy consuming. However, spatial measures were used more often than count measures for the themes of learning strategies and decision making patterns. This is reasonable because these two themes involve meta-cogni-tive skills, and to show meta-cognimeta-cogni-tive skills may require more sophisticated observations or analyses of fixation locations and sequences for a better conclusion. Moreover, the choice of eye-tracking measures is often motivated by specific research questions. Different research fields may focus on different types of measures. For example, in the field of reading compre-hension the focus is mostly on temporal measures, while in the field of scene perception the focus is more often on spatial measures. Since the majority of the studies in the pool of this review were related to reading-oriented learning activities, it is Table 4
Use of the eye movement measures related to different learning themes.
Learning theme Temporal Spatial Count
Patterns of information processing 47 22 24
Effects of instructional strategies 47 7 15
Reexaminations of existing theories 17 1 5
Effects of learning strategies 12 3 2
Individual differences 9 1 7
Patterns of decision making 3 3 1
Social cultural effects 3 1 1
Total 138 38 55
reasonable to obtain this result. In summary, it is important to have clear research questions or hypotheses in order to choose proper eye-tracking measures for examinations.