Desktop version

Home arrow Philosophy

  • Increase font
  • Decrease font


<<   CONTENTS   >>

Can Eye Movements Shed Light on Processing Strategies and Strategy Use?

In eye movement registration the location of the eye gaze is recorded with short time intervals (e.g., a low-precision eye-tracker with a sampling frequency of 60 Hz collects pictures of the eye gaze every 16.67 milliseconds, and a high-precision eyetracker with a sampling frequency of 1200 Hz collects pictures of the eye gaze every .83 milliseconds). In a next step, the location of the eye gaze is related to the stimulus a participant is looking at (e.g., a text, a picture, a video, a questionnaire). This technique allows us to investigate to what parts of the learning material a student allocates visual attention and for how long (Holmqvist & Andersson, 2017). Eye movement research has been used extensively to better understand reading processes at the word and sentence level, the text level and the level of multiple documents (Hydna, Lorch, & Rinck, 2003; Jarodzka & Brand-Gruwel, 2017; Rayner, 2009). Eye movement research focuses on the micro-processes of reading (e.g., how much time is needed to process certain words and from which word a reader starts rereading a sentence). Theorists have emphasized that the text-related processing strategy adopted by a reader/student will have strong effects on micro-processes and on the construction of the mental representation (Hydna, Lorch, & Kaakinen, 2002; Kintsch, 1998; Kintsch & van Dijk, 1978). Focusing on text learning, first pass and second pass reading times are often used as eye movement duration measures (Hydna et al., 2003; Jarodzka & Brand-Gruwel, 2017). First pass reading time refers to the summed duration of all the fixations on the target region (e.g., a sentence) before exiting it. Second pass reading time refers to the duration of all regressions back to the target region (e.g., a sentence) after the first pass reading time has been terminated (Hydna et al., 2003). First pass reading times are an indication of early processing and object recognition (Hydna et al., 2003). Second pass reading times or rereading times reflect processes happening later in comprehension (Holmqvist & Andersson, 2017), such as high-level or deeper cognitive processing (Ariasi & Mason, 2011; Holmqvist & Andersson, 2017; Penttinen, Anto, & Mikkila-Erdmann, 2013) and attempts to reinstate text information into working memory in order to elaborate on it or rehearse it (Hyônâ & Lorch, 2004). Second pass reading times are thus more strategic in nature than first pass reading times.

Different student characteristics shape how students build up their mental representations during text learning (Alexander & Jetton, 1996; Fox, 2009; Jarodzka & Brand-Gruwel, 2017). Influential models on deep and surface processing strategies and strategy use stress the importance of the interplay between learner characteristics and the nature of that processing or strategy use (Alexander, 1997; Dinsmore & Hattan, this volume; Richardson, 2015; Vermunt & Donche, 2017). Important learner characteristics that affect the nature of processing during learning are the students’ general disposition towards strategy use, interest, motivation, prior knowledge, working memory capacity, personality, regulation and emotions (Baeten, Kyndt, Struyven, & Dochy, 2010; Vermunt, 2005; Vermunt & Donche, 2017). Students’ general disposition towards deep and surface processing strategies can have an important influence on how they learn from texts (Kirby, Cain, & White, 2012), especially on what they perceive to be relevant or important in the text (Kendeou & Trevors, 2012). The study of Catrysse et al. (2018) combined general self-report questionnaires on deep and surface processing strategies with eye tracking data of reading one expository text. They showed that students with a general disposition towards combining deep and surface processing strategies, as measured with self-report questionnaires before reading the text, reread the text more thoroughly than students who were lacking in the use processing strategies, as reflected in longer second pass fixation durations. We believe this reflects the uni-dimensional effect of the students’ general disposition towards processing strategies on their actual processing during learning from a text. However, the study of Catrysse, Gijbels, and Donche (2018) provided evidence for the multidimensional nature of strategy use during text learning. In this study, general self-report questionnaires on deep and surface processing strategies, task-specific cued retrospective think-alouds and eye tracking were combined. They showed that highly interested students, who use deep processing strategies (both as measured with a general self-report questionnaires and cued retrospective think-alouds), reread key sentences in a text for longer than detailed sentences and thus process these key sentences more deeply. The study did not take students’ learning outcomes into account, so it is unclear whether processing these key sentences more thoroughly was related to better reading comprehension. This study emphasizes the importance of the interplay between processing strategies and other learner characteristics in order to fully understand the micro-processes of learning. In addition, it shows that the selectivity in processing information in the text is the result of a more complex interplay between different learner characteristics and is not solely determined by processing strategies. Although this was already assumed in theoretical frameworks and empirical research focusing on learning outcomes (Alexander & Jetton, 1996; Schiefele, 1996, 1999,2012; Schiefele & Krapp, 1996), eye movement registration now allows us to show these effects during the online learning and reading process.

Early work on eye movements and vision showed the important influence of the learning task on participants’ eye movements (Yarbus, 1967). Also, reading and learning from a text may occur with diverse tasks in mind, such as reading in order to give a presentation, reading in order to answer closed-ended questions, reading in order to answer open-ended questions, reading for entertainment, reading in order to find relevant information, etc. (Kaakinen & Hydna, 2005, 2007; Kaakinen, Hydna, & Keenan, 2002; Yeari, Oudega, & van Den Broek, 2016; Yeari, van Den Broek, & Oudega, 2015). In addition, researchers examining students’ processing strategies agree that one of the most salient contextual variables influencing processing strategies is the assessment method. This is also known as the backwash-effect of assessment (Baeten et al., 2010; Gielen, Dochy, & Dierick, 2003; Segers, Nijhuis, & Gijselaers, 2006). The study of Catrysse et al. (2016) showed that the assessment demands influence how the students process different types of information in the text. More specifically, students who were expecting reproduction-oriented questions processed the details more thoroughly and repeated these details more often. However, students in the deep condition did not look longer at essentials in the text, but this can be explained by the fact that incorporating key information in the mental representation can be achieved mentally, or can result in overt behavior in which students actively reread essential parts (Hydna et al., 2003; Kaakinen & Hydna, 2008). Other research showed that key information or central ideas in the text are learned regardless of strategy use and are necessary for building a mental representation of the text (Lonka, Lindblom-Ylanne, & Maury, 1994; van Dijk & Kintsch, 1983). In other eye-tracking studies on students’ processing strategies, all students received the same instruction for learning the text, namely, they needed to learn the text in order to answer questions on the content afterwards. It was not further specified which type of questions students could expect (Catrysse et al., 2018). Therefore, it may come as no surprise that students processed details and key sentences in a similar way in order to prepare for all kinds of questions. In another study, all students received the instruction to study the text material like they would do when preparing for exams. Again, it was no surprise that students using different processing strategies were not selective in processing key sentences and details, as the learning task was quite general (Catrysse et al., 2018).

Is fMRI a Bridge Too Far?

With regard to brain imaging methods, there has been an explosion in neuro-educational research since the beginning of the 21st century (Huettel, Song, & McCarthy, 2014). Brain imaging methods are used to localize deep and surface processing strategies in the brain, more specifically, to examine which brain regions are activated during deep and surface processing strategies (Galli, 2014). In 1997, Bruer published a paper in which he claimed that neuroscience was a bridge too far for educational research (Bruer, 1997), meaning that we cannot draw implications for educational practice directly from neuroscientific research. However, together with other researchers, he believed that a two-way path is possible in which education can be linked to cognitive science in fields such as educational and cognitive psychology, and cognitive science can be linked to neuroscience (Bruer, 1997; Mason, 2009; Mayer, 1998). And as student learning takes place in the brain, neuroscience is a relevant research area to examine further (Mayer, 1998). In the educational psychology literature, researchers have emphasized that we should not see deep and surface processing strategies as a pure dichotomy, but rather as being at the ends of a continuum (Dinsmore & Alexander, 2016; Lonka, Olkinuora, & Makinen, 2004). Most of the time students combine several processing strategies while learning, and, consequently, how students learn cannot by characterized by one single processing strategy (Donche & Van Petegem, 2009; Vanthournout, Coertjens, Gijbels, Donche, & Van Petegem, 2013). Previous eye movement studies in the field have also provided evidence for the fact that deep and surface processing strategies are often combined when learning (Catrysse et al., 2016, 2018; Catrysse et al., 2018). Neuroscientific research can provide more insight into whether differences between deep and surface processing strategies are qualitative (i.e., activations in different brain regions) or quantitative (i.e., overlapping brain regions but with differences in the level of activation) in nature (Galli, 2014). More specifically, it has the opportunity to indicate whether deep and surface processing strategies are overlapping constructs or not (Catrysse, Gijbels, & Donche, 2019).

In accordance with what other researchers have mentioned as being challenges for neuroscientific research (Mason, 2009; Mayer, 1998; Varma, McCandliss, & Schwartz, 2008; Willems, 2015), processing strategies have mostly been examined using very basic learning tasks at the word level (Catrysse et al., 2019; Galli, 2014). Subjects receive deep-level and surface-level learning tasks in order to evaluate words when they are being scanned. An example of deep-level tasks is the animacy judgement task in which subjects need to decide whether a word is a living or non-living object. The case judgement task is an example of a surface-level task where subjects need to decide whether a word is printed in uppercase or lowercase (Galli, 2014). The levels-of-processing effect was mostly examined at the word level and we suggest two possible explanations for this. Firstly, the practical constraints of the MRI scanner result in the use of very basic learning tasks. Secondly, Craik and Lockharts (1972) levels-of-processing framework was extensively investigated in experimental research at the word level on memory (Gallo, Meadow, Johnson, & Foster, 2008; Sporer, 1991; Weinstein, Bugg, & Roediger, 2008). In addition, this line of experimental research provided clear and convergent outcomes; namely, that deep processing strategies lead to better recall performance than surface processing strategies (Richardson, 2015). These robust findings at the behavioral level serve as a good start to setting up neuroscientific research, because clear hypotheses can be tested (Mayer, 1998). We can conclude that fMRI research is no bridge too far when it is used to gain more insight into processing strategies at the word level. However, we want to emphasize that a great variety of encoding tasks is used in this line of research, making it hard to compare studies (Catrysse et al., 2019). The review of Galli (2014) indicated that it is hard to precisely distinguish between deep and surface encoding tasks. Galli (2014) suggested that this is one of the main reasons why neuroscientific research is offering mixed evidence on whether the distinction between deep and surface processing strategies is qualitative or quantitative in nature.

 
<<   CONTENTS   >>

Related topics