Table 1. Reading Levels, Word Recognition Rates and Reading Comprehension Criteria
First language literacy is a prerequisite both for proper formal education and second language learning. When children start school, not all of them succeed in learning reading perfectly; for some of them, it may take more time to reach the desired reading proficiency. This study aimed to examine the relationship between the first language oral reading behaviors of poor readers among 3rd graders. For this purpose, reading comprehension levels and reading behaviors along with eye movements of eight poor readers were investigated. A narrative text suitable for the grade level was utilized to determine the reading skills and analyzed through IRI and Ekwall and Shanker Reading Inventory. Additionally, the eye movements of 8 adult skilled readers were recorded for comparison. Reading speed, word recognition rates, reading errors and comprehension scores were analyzed along with dwell time, fixations and total time spent on each word. Results showed that poor readers had viable difficulties in comprehension and made frequent reading errors which were also accompanied by a low word recognition rate. Poor readers also exhibited inflated eye movement parameters when compared to adult skilled readers. The results were discussed regarding poor reading ability, dyslexia and possible pedagogical interventions.
Reading is a complex process in which cognitive and physical processes go hand in hand to create meaning. It involves various functions of the eye and brain such as seeing, perceiving, vocalizing, understanding and mental structuring (Güneş, 2012). Akyol (2017) defines it as a dynamic meaning-making process that requires interaction between the author and the reader. When the definitions are examined, it is inferred that reading cannot be considered separately from the concept of comprehension. When readers perform word recognition and distinction while reading the text, they also perform comprehension processes by using their prior knowledge and several reading comprehension strategies (Baştuğ et al., 2019). Hence, reading comprehension can be elaborated as reconstructing meaning created by examining, sorting, classifying, relating, questioning and evaluating the information obtained by the individual through reading by combining them with the previous knowledge of the reader (Güneş, 2009). Throughout this process, a product is created which is a combination of competency, word recognition, reading fluency, syntactic processing and word meaning knowledge (Rayner et al., 2006). Since the primary aim of reading is comprehension, poor readers commonly have serious problems in obtaining information from texts. Especially, young learners with reading problems in their native language are not only incompetent in reading comprehension but also have specific reading problems. Even though they have normal intelligence levels and pedagogical experience, their problems with the text itself persist. They generally experience inversion of letters or words, weak word skills and inability to comprehend (Cited in Lefton, 1979; Pavlidis, 1978). The aim of this study was to reveal time-based eye movements of young learners who learn reading in L1 during reading and compare it to reading and comprehension performance. The study was significant since eye movement analysis of young learners during reading in L1 was rare in the Turkish context.
Poor readers commonly have problems in decoding the text and a majority of them have difficulty in understanding what they read when they do not experience decoding problems (Landi & Ryherd, 2015). On the other hand, skilled readers constantly evaluate their understanding while reading; if they cannot, they use strategies to construct meaning. They monitor comprehension to identify contradictory expressions or inconsistencies in the text (van der Schoot et al., 2012). It was known that readers with high levels of comprehension also tend to have vocabulary and word reading skills. Less fluent readers need to shift their limited attention between word identification and comprehension processes because being able to do both at the same time exceeds readers' attention skills. This results in strenuous reading and poor understanding. On the other hand, fluent readers can recognize words almost effortlessly and maintain constant attention to understanding the text (Nelson, 2010). Several related studies showed that comprehension was closely linked to word decoding skills and vocabulary (Verhoeven & Leeuwe, 2008; Beck et al., 1982; Stahl & Fairbanks, 1986; Braze et al., 2007). Around the age of 6 many children are considered beginner readers (Juel et al., 1986). Several studies reported that with the increasing age, children start to employ shorter fixation durations, make fewer fixations and regressions per sentence along with smaller saccade amplitudes (Taylor, 1965; Rayner, 1985; McConkie et al., 1991). Hence children learning to read were expected to have a higher total fixation duration along with several fixations, more regressions and longer saccades. For poor young readers,cthese symptoms were more observable as they encounter several problems during reading.
As for reading skill and age increase, the decrease in cognitive processing difficulties related to text comprehension will be reflected in changes in eye movements. In this case, it will reveal differences in reading skills among children of the same age (Blythe & Joseph, 2011). It is observed that children with reading problems fall behind their peers according to their age and grade levels and make various errors while reading. The common errors are additions, word/letter skips, repetitions, wrong reading and inversion of letters (Akyol, 2016). The excessive amount of errors make the child unable to make sense of the text, as reading comprehension and text analysis skills are interrelated. A healthy analysis of text helps students to identify the words in their vocabulary and realize that this definition is important to understand (Polloway et al., 2014).
It has been observed that poor readers cannot distinguish important information from unimportant information during reading, cannot control their comprehension processes, and therefore have difficulties in understanding what they read as well as in reading (Kuruyer & Özsoy, 2016). These difficulties lead psychologists, educators, ophthalmologists and neurologists to fundamental problems such as neurological basis of reading, perception and understanding of higher developmental level patterns (Adler-Grinberg & Stark, 1978).
While reading the text it feels like the eyes are gliding over the text page, as it performs a series of rapid movements. The eyes make leaps (saccades) and pause in the text from one place to another. For skilled readers, eyes typically move in the range of about seven to nine letters per saccade (Rayner, Chace, Slattery & Ashby, 2006). Between saccades, our eyes remain motionless during fixings of approximately 200-300 ms (Rayner, 1998). It was thought that information was obtained during these fixations. Although our participants were not diagnosed as dyslexic, “poor reader” would be more suitable, several previous reading research proposed abundant evidence stating that both poor and dyslexic readers tend to have longer fixations (pauses), short saccades, more fixations and more regressions (Rayner, 1998, 1978; Vagge et al., 2015; Rayner et al., 2006; Eden et al., 1994; Seassau & Bucci, 2013). The change in oculomotor behaviors is observable through eye-tracking which indicates serious improvement in reading skills (Kirkby et al., 2008). Thus, eyetracking studies have a crucial role in understanding the problems experienced in reading and provide pedagogical solutions.
The study aimed to examine the differences regarding the reading and reading comprehension levels of poor readers via eye-tracking. In line with this purpose, answers will be sought for the following research questions:
In this study, the relational scan model was carried out to reveal the reading aloud behaviors regarding the reading and reading comprehension levels of young poor readers through eye-tracking. The general purpose of the relational scan model is to make a description by taking a picture of the current situation related to the research subject (Büyüköztürk et al., 2017). We adopted this model to describe reading level, reading comprehension and eye movement differences across different reading proficiency levels and to build a rationale between them.
The beginner participants were selected from the pupils from the third grade of a state elementary school with a purposeful sampling method. While selecting beginner participants, we collaborated with the teacher of the students and confirmed that all participants had observable reading difficulties. After the students were identified, the families were informed about the study, and a parent consent form was signed containing the terms of the study. While determining the study group, the criteria for students' reading levels to be at anxiety or teaching level were sought. The study cluster of the study consisted of 12 pupils with the age of ten but only 8 of them completed all the procedures (2 males, 6 females). 4 participants were excluded due to poor eye calibration during experiments and inability to understand instructions properly. 8 (5 males and 3 females) adult skilled readers in the age range of 25 to 34 voluntarily participated in the study. They had at least a bachelor's degree and were employed in a state university. All participants had normal or corrected to normal eyesight.
Ekwall & Shanker Reading Inventory developed by Ekwall and Shanker (1988) adapted into Turkish by Akyol (2017) was used to determine the performance of the participants regarding their reading level and reading aloud performance. With this inventory, which is also informative about what kind of errors students make, three types of reading levels are determined: anxiety level, teaching level and free level (Akyol, 2016).
Informal Reading Inventory-IRI (Karasu et al., 2013) was used to determine the students' level of reading comprehension. The inventory consisted of 10 questions about a predetermined narrative text. The total score that students can get is 100 points. In the current study, IRI was applied after a narrative text suitable for 3rd-grade level. The questions about the narrative text used included different difficulty levels. Three types of questions were used to measure reading comprehension: textual open questions (4), textual closed questions (4), and questions based on knowledge and experience. Reading comprehension level and word recognition rates were determined with the criteria given in Table 1 below:
Table 1. Reading Levels, Word Recognition Rates and Reading Comprehension Criteria
To determine the reading behaviors and reading comprehension levels of the beginner students, a narrative text consisting of 177 words (see Appendix 1) and a reading comprehension form related to the text suitable for the 3rd grade level of primary school was applied.
Appendix 1. Text Stimulus
To ensure validity, the text stimulus was chosen in accordance with the Ekwall and Shanker Reading Inventory. The text was appropriate for a third-grader with a word count between 100 and 200 words (Akyol, 2006). In this study, a narrative text consisting of 177 words suitable for the third-grade level was selected. The teaching and comprehension of narrative texts are easier than informative texts due to their structure (Akyol, 2006) In addition, the structure of the selected text (story parts) and readability levels (number of words, t-unit number, number of clauses, and different words) and reliability studies were conducted and inter-rater reliability scores were found to be 100% (Karasu et al., 2013).
To determine the reading comprehension levels of the readers, 10 questions were asked about the text titled "ÖmerveGüvercin". The number of questions was determined regarding the grade levels by paying attention to the fact that the questions are textual open, textual closed, and questions based on knowledge and experience, including the order of the main event/thought in the text and ensure that they are suitable for text structures, as well as that the questions include the three types of questions mentioned. Taking the students' acquisition of reading strategies and the development of their usage skills into account, 10 questions were prepared consisting of 4 open textual, 4 closed textual questions, and 2 knowledge-experience questions (Karasu et al., 2013). In addition, the answers given by the students to the reading comprehension inventory questions were scored by a primary school teacher and the corresponding author. The interrater reliability was calculated by using the formula "Reliability = number of consensus/total agreement + number of disagreements" (Miles & Huberman, 2015). The reliability between raters was determined as 88%.
The eye movements were recorded through a remote GP3 eye tracker with a 60 hz speed. This eye tracking device can register a sample every 16 milliseconds with 0.5–1° of visual angle accuracy and a 25 cm (horizontal) x 11 cm (vertical) head movement flexibility. Eye movement data were analyzed with GP3 Professional Software. GP3 is a noninvasive remote eye tracker.
Appendix 2. The Eye Tracker Device used in the Study (GP3 Remote Eye Tracker)
Each beginner participant took the same procedures individually. Eye movements and voices of beginner learners were registered while they read the text aloud. This procedure was followed by the Informal Reading Inventory- IRI and Ekwall and Shanker Reading Inventory. For skilled adult readers, we only recorded eye movement data since they were already proficient in their L1 and the inventories were not suitable for them.
With the GP3 eye tracker, dwell time and fixation count values were obtained. Dwell time refers to the total time on any given Area Of Interest (AOI). In our study, the text as a whole was taken as a single AOI. Fixation refers to each pause employed and seems like a bubble (see Figure 1). Fixation count is the number of pauses employed. To examine dwell time and fixation count on each word, these values were divided into word count in the text (177). Eye movement data were analyzed via GP Professional software. A Mann-Whitney U test was adopted to see any significant difference between beginner and adult skilled readers regarding eye movements.
Figure 1. Eye Tracking Data Sample
Recorded oral reading performances were deciphered and Words Speed per Minute (WPM), the number of words read incorrectly, and the rate of word recognition was analyzed depending on the Ekwallve and Shanker Reading Inventory. Informal Reading Inventory-IRI was used to examine the level of reading comprehension.
The results of reading and reading comprehension and results of poor readers at 3rd grade were presented in Table 2.
Table 2. Findings Related to Reading Level and Reading Comprehension of Beginner Readers
Table 2 was examined, and none of our beginner participants were at teaching level on both categories. Thus, we assumed that all the participants had observable reading problems. Findings regarding the types of errors made by students are presented in Table 3.
Table 3. Common Error Types
Table 3 confirmed Table 2. All learners were observed to have committed incorrect reading errors at different levels. A majority of them also made additions while reading and skipped some words while reading. S2 and S8 made the least errors when compared to other participants who were in anxiety level. According to the findings, S1, S4 and S7 who made excessive number of errors had serious reading problems. Examples of error types observed during reading were given as follows:
Incorrect Reading: All participants have read some words incorrectly by making errors in decoding typo such as reading the words “oynarken” and “kurtarmam” as “oynanırken” and “kurtulmam” respectively.
Addition: Most of the participants tended to add an extra morpheme(s) to the words they could not decode properly such as reading the words “yavru” and “uçuyor” as “yavrun” and uçuyordu” respectively.
Elision: All participants showed erroneous elision behavior by omitting required morpheme(s) such as reading the words “çıkarak” and “güvercini” as “çıkar” and “güvercin” respectively.
Skipping: More than half of the participants skipped several words erroneously and randomly.
In this section, we also used related data from (Özer, 2019) to make results more interpretable. We initially compared real-time eye movement patterns of the poor readers and adult skilled readers as shown in Figure 2.
Figure 2. Real-Time Heatmaps and Gaze Plots*
When the eye movement patterns were analyzed (please see Appendix 3 for detailed gaze plots for each participant), we initially observed more intense hot regions on the poor reader heatmap (on the left) when compared to adult skilled readers (on the right). Additionally, a higher density of fixations was observable in the poor readers' gaze plot along with bigger fixations which referred to higher fixation duration.
Appendix 3. Detailed Gaze Plots for each Participant
We assumed that the poor readers employed frequent fixations in which they spent a considerable amount of time. On the other hand, adult skilled readers made fewer fixations with less duration. Mean values are given in Table 4.
Table 4. Eye Movements across Reading Levels
First of all, as expected, we found a significant difference between our beginner readers and adult skilled readers regarding all eye movement measures at p< .05 level with a Mann-Whitney U Test. Also, according to Table 4, our beginner participants slightly fell behind the anxiety level participants in Özer (2019) regarding dwell time. Hence, we assumed that they even performed worse than poor beginner readers in Özer (2019). On the other hand, all beginner readers including free level readers had inflated eye movement values when compared to adult skilled readers. While adult skilled readers employed no more than a single fixation on each word, other levels made at least two fixations for each word. Our findings showed that our beginner readers had serious problems in processing words as they have spent a considerable amount of fixation duration on each word. The same metric for other beginners was found to have been under 1 second. For adult skilled readers, it was observed to be less than half of a second.
The general purpose of this study was to reveal the reading proficiency gap between beginner and skilled readers in L1. For this purpose, eye movement data, WPM, word recognition rate and comprehension performances of weak readers were examined through a narrative text.
To determine the reading levels of the beginner participants, a narrative text consisting of 177 words was given and their oral reading performances were evaluated through the Ekwall and Shanker Reading Inventory. The results showed that two students were at teaching level in one category only while six of the participants were at the anxiety level in both categories. The word recognition rate of the anxiety level beginners was observed to have been below 90%, while the students at the education level are below 98%. The number of words the students read incorrectly was determined to be between 4 to 50 words.
When the number of words read by the participants per minute was examined, the most successful performance was between 70 WPM and 25 WPM. 3rd Graders needed to read 100 words per minute (Güneş, 2007) and our results showed no better performance than 70 WPM. Based on these results, we could propose that beginners had moderate word recognition rates while their error numbers persisted which crippled their reading speed. Also, none of the beginner participants met the expected criteria in terms of WPM. Our findings related to low WPM and high reading errors were also in line with several recent studies in literacy development in Turkish as L1 (Erden et al., 2002; Yıldız et al., 2009; Akyol & Temur, 2006; Akyol et al., 2014; Dündar & Akyol, 2014; Çeliktürk-Sezgin & Akyol, 2015; Akyol & Sever, 2019; Akyol & Ketenoğlu-Kayabaşı, 2018).
To determine the reading comprehension level of the poor readers, questions related to the narrative text presented were prepared and the answers given by the students were scored according to the IRI. The results of IRI showed that the reading comprehension level of most of the students was below 75%. Hence, we assumed that the poor readers' reading comprehension levels were also inadequate which is in parallel with their low reading levels accompanied by poor word recognition and WPM. Poor phonological processing was emphasized as a cause of reading difficulties in the field of dyslexia. However, beginner readers who struggled to read also experienced difficulty with semantics, syntax and discourse in the process of literacy acquisition (Bishop & Snowling, 2004). Many studies examined the relationship between reading and reading comprehension (Tazebay, 1995; Acat, 1996; Çaycı & Demir, 2006; Ellis, 2009; Egmon, 2008; Palmer, 2010; Miller & Schwanenflugel, 2006). In a study conducted on the phonological processing skills of English and German dyslexic children, the reading levels of students decreased to around 50% when they come across texts containing polysyllabic long words (Landerl et al., 1997). Our results were also in line with these related findings, proposing that beginners with structural problems in reading also had deficiency in reading comprehension.
According to the findings of the second sub-problem of the study, we found out that the poor readers spent a considerable amount of time on the text and employed several fixations on each word. When compared to adult skilled reader eye movements, these numbers were observed to have been rather inflated. Our results also coincided with the one of Özer (2019) regarding anxiety level readers. We assumed that the slight differences would be related to the different eye trackers used in the two studies. Eye fixations on each word during silent reading English as L1 were found to last about 250 milli seconds and the average saccade size of the eye was in the range of 7 to 9 letters (Rayner, 1998). For Turkish, Baştuğ et al. (2019) examined eye movements according to reading types (reading aloud and silent), and concluded that the participants had more pause time and several pauses during reading aloud than silent reading. Lefton et al. (1979) examined third and fifth-Grade primary school students and adults with good and poor reading skills. The results showed that weak readers made longer pauses and fixations, and at the same time, they performed more regressions than other groups. In a study conducted on German children, it was concluded that the reading speed was slow, the number of fixations was higher, and regressions were higher in the group of individuals with dyslexia (Trauzettel-Klosinski et al., 2010). In a study conducted with students with dyslexia, it was found that students with dyslexia exhibit more fixation and longer fixation time during text reading (Hutzler & Wimmer, 2004), although eye movement mean values differed to some extent across these studies, including ours. We also confirmed that reading difficulties could be observed through eye-tracking. Time spent on the text and fixation count gradually decreased as the proficiency increased, since learners started to improve their syntactic, morphological and comprehension skills in time.
The results of the study showed that poor reading behavior could be observed through eye-tracking methodology, as reading difficulties were found to have been reflected on eye movement parameters. We concluded that beginner readers in this study had some observable difficulties and had a long way to go to become skilled readers. Most importantly, examining eye movements during reading can contribute to understanding literacy development. Especially, longitudinal studies in which L1 learners' reading behavior is tracked progressively with eye-tracking would provide much more evidence on how literacy develops in time. This technique is even more important for learners who had observable reading difficulties as it may help to develop better pedagogical interventions. Although reading inventories works fine with tracking learner reading development, eye movements such as how much time spent on words and how many fixations are made can give a detailed picture of the L1 reading development of poor readers. These learners can be given appropriate feedback with the support of eye movement data which may help teachers, instructors and researchers to come up with better pedagogical solutions.
This study is not with its limitations. The number of participants would have been higher but finding suitable students with reading problems at the given age range and applying a successful eye-tracking experiment with young participants was challenging. It is always harder to track the eyes of children as they cannot keep their concentration for a longer period and they keep moving during the experiments.
A longitudinal developmental design may be advised for more detailed results which would come up with more pedagogical implications. Reading development of a small group of young participants with reading problems can be tracked with eye-tracking and classroom observations with proper pedagogical intervention. Differences in eye movements for a longer period (i.e. a semester) can be triangulated with observation notes which may result in invaluable insights for L1 pedagogy.