Размещено на http://www.allbest.ru/

Правительство Российской Федерации

Федеральное государственное автономное образовательное

учреждение высшего образования

Национальный исследовательский университет

«Высшая школа экономики»

Факультет гуманитарных наук

Образовательная программа

«Фундаментальная и компьютерная лингвистика»

Eye-movements during reading in children with hearing loss

Anastasia Kaprielova

Москва 2020



TABLE OF CONTENTS

• 1. Introduction
• 1.1 Eye-tracking studies
• 2. Background
• 2.1 Literature review
• 2.1.1 Deaf readers
• 2.1.2 Typically developing hearing children
• 2.1.3 Children with dyslexia
• 2.2 Current study
• 3. Methods
• 3.1 Participants
• 3.2 Materials
• 3.2.1 Design and stimuli: reading experiment
• 3.2.2 Design and stimuli: visual search task
• 3.2.3 Apparatus
• 3.3 Procedure
• 4. Results
• 4.1 Reading experiment
• 4.2 Visual search task
• 4.3 Vocabulary test
• 4.4 Raven's Progressive Matrices
• Discussion
• Conclusion
• Acknowledgments
• References
• Appendix 1: Data for the other participants
• Appendix 2: Stimuli
• Appendix 3: Reading experiment results


1. Introduction

Indisputably, reading is a fundamentally important skill for each individual in modern society. Therefore, it is essential to know how different groups of population acquire this skill. There is a substantial number of studies that investigate reading patterns in healthy adults, teenagers, and children. There is also a moderate body of research is describing the process of reading in children with dyslexia and adults with aphasia, or groups with other speech (or mental health) disorders. At the same time, research on reading process in deaf population is remarkably scarce. dyslexia children's peripheral vision

The majority of studies of reading in deaf populations were conducted with the speakers of American Sign Language (ASL). According to these papers, learning to read is difficult for deaf individuals because they cannot rely on the phonological codes; instead, they must rely exclusively on spelling (Bйlanger, 2013). At the same time, they have certain advantages, such as highly developed peripheral vision which allows them to discern 18 characters to the right of the current fixation - while hearing readers discern only 14 (Bйlanger, 2015). For that reason, deaf readers can catch up in reading speed with hearing readers or even outperform them.

Last year we investigated reading in eight adult native speakers of Russian Sign Language (RSL) as compared to 96 hearing Russian readers. In line with their unique status, RSL speakers exhibited some patterns characteristic of proficient readers (fixation durations and skipping rates similar to those of the control group) as well as some patterns characteristic of poor readers, such as saccade landing position closer to the beginning of the word (saccade is a rapid movement of the eye between fixations), longer total reading times and a greater number of fixations on a word, decreased sensitivity to word frequency and predictability, and increased slowdown in reading longer words. A possible explanation for these results may be that deaf lack reading experience and have lower reading skills. It could also be that the reading experience is comparable between deaf and hearing participants, but the deaf still lack exposure to oral language use, so that frequent and predictable words are not so frequent of predictable in their personal language use. dyslexia children's peripheral vision

The goal of this study is to find out whether these patterns are also true for RSL-speaking children. We are going to investigate the mechanisms of reading acquisition in primary school children with hearing loss who communicate in Russian Sign Language since their birth or early age. Furthermore, we are going to compare their eye-movements during reading to those of dyslexic children and to those of the control group of typically developing children without hearing loss and reading disorders. We expect that reading level of the group of children with hearing loss will be close to the control group of typically developing children, but their reading comprehension may be on the level of the group of children with dyslexia. The particular comparison with children with dyslexia is interesting also because it was demonstrated that while deaf adults have a larger-than-normal parafoveal preview, adults with dyslexia have a smaller-than-normal one, consistently with a visual deficit hypothesis of dyslexia. We will test whether similar dissociation could be found in the reading patterns of the two groups of children.

Children with hearing loss participated in four tests: (a) an online vocabulary test; (b) an eye-tracking experiment with the silent reading task; (c) an IQ test based on Raven's Colored Progressive Matrices; (d) another eye-tracking experiment testing the size of visual attention span during visual search.

The results of our study may potentially help to develop school programs for evaluating and correcting reading and reading comprehension skills in children with hearing loss.

1.1 Eye-tracking studies

We'll start with a brief overview of eye tracking.

In 1879, ЙmileJaval was the first who noticed that our eyes do not move smoothly in reading, but make a series of rapid eye-movements in one direction (saccades) and short stops (fixations). Moreover, J. M. Cattell in 1886 was working with Wundt in Leipzig with a new device which demonstrated that words during reading can be recognized from a single glance. That means that we do not need to read the word letter by letter or move our eye across the word in order to read it.

As we already mentioned, our eyes do not move smoothly, but in a series of small rapid movements called saccades. Planning a saccade is usually done in parallel with comprehension process in reading. Between the saccades there are fixations - periods of short stops - that last about 200-300ms. It is important to know that we do not obtain new information during saccades, because our eyes move too quickly; all visual information is extracted during fixations (Uttal& Smith, 1968). But some studies claim that lexical processing is not suppressed during saccades (Irwin, 1998).

Regression saccades move the eyes back to the parts of the text that were already read. It is widely accepted that regressions are typical for complicated parts of the text. Sometimes we make regressions due to the too long saccades. In this case, we first skip a few words and then come back to process them. Keith Rayner (1998) mentioned that short saccades to the left may be necessary for reading to proceed efficiently. Long regressions that span more than 10 letter spaces back happen because the reader experiences some difficulties in processing the text. Proficient readers are more precise in sending their eyes back to the most complicated part of the text (Frazier & Rayner, 1982; Kennedy, 1983; Kennedy & Murray, 1987; Murray & Kennedy, 1988). On the contrary, poor readers go back through the text in order to find the part of the text that caused them difficulty.

The reason why we make saccades so often is that our visual acuity has limitations. Our visual field is divided into three sections (see Figure 1).

Figure 1. Regions of the visual field

The highest acuity is in the foveal area (the central 2? of vision), it gives us a clear detailed image. In the parafovea (extends out to 5? on both side of fixation) the ability to distinguish details decreases, and in periphery (the area beyond the parafovea) it is even poorer. This part is highly important to our current study.

Rayner (1998) suggested that when there is more information located in the peripheral area, the more fixations we need to understand the text, and thus the speed of reading decreases. However, more recent studies do not lend support to this statement (Bйlanger, 2015). If it is difficult to process the information in periphery, we can do not use it all in order not to confuse ourselves. Nevertheless, some information still comes from peripheral vision. Nathalie Bйlanger and her colleagues demonstrated that more developed peripheral vision could be even an advantage for those individuals who have it. More developed peripheral vision could even increase the reading speed. We expect that children with hearing loss have this advantage. Even though, we aimed to test this explicitly with a separate study targeting peripheral vision, but were unable to collect the data due to pandemic-related restrictions.

Generally, in reading experiments we ask our participants to read a word, a sentence or a paragraph. Sometimes participants have to answer the comprehension questions after sentences.

Eye-tracker records the number of saccades and fixation durations during reading, so we can understand how long it takes to process a word or a sentence. In such experiments total reading times, saccade length and other reading measures can change depending on the level of text complexity or on the frequency of the certain words and constructions.

The vast majority of reading experiments investigate process of silent reading. We will briefly review some basic facts about eye movements during reading in Russian. Laurinavichyute et al. (2017) conducted a study investigating the reading process in monolingual Russian-speaking adults (N = 96). The task was to read 147 sentences and answer three-choice comprehension question after 1/3 sentences.

They found that participants skipped 34% words during reading, while 56% words were fixated only once. This is a common proportion, short and frequent words are processed parafoveally and do not need a separate fixation, therefore, they are skipped. Function words are more likely to be skipped, because they are short and have high predictability in most cases. Here we can see the general relationship between word length and probability of skipping this word. As length increases, the probability of fixating a word increases.

Higher predictability of a word in a context increases the chances that there will be a fixation on this word. This is unusual, because it seems that the probability of skipping a word should increase when the word is more predictable in the context. It was previously investigated in German and other languages. However, in Russian there is no strict relationship between the word predictability and location of this word in the sentence.

In Russian, word length increases several reading measures: first fixation duration - duration of the fist fixation on a word if there were no other fixations; single fixation duration; gaze duration - total duration of all fixations on a word before our eyes move to the next word on the right; total reading time - total duration of all fixations on a word, including second reading and regressions.

Probability of regressive saccade increases when the word is more predictable. Word predictability decreases gaze durations and total reading times.

This study illustrates that reading in Russian is similar to reading in other alphabetic languages, such as English or German.

In the next section we will discuss how the reading process may differ in deaf adults, typically developing children without reading disorders and children with dyslexia.



2. Background

2.1 Literature review

2.1.1 Deaf readers

As it was already mentioned, learning to read is more difficult for deaf individuals. Some studies have shown that only 15% of deaf or hard of hearing students read above the sixth-grade level comparing to typically developing children (Allen, 1994). Children with hearing loss graduating from school are unable to outperform secondary school children in reading. In order to become a successful reader, it is usually required to understand the mapping between the spoken language and the printed word. Nonetheless, children with hearing loss are unable to learn this mapping between the spoken language and the printed word. We will discuss below how profoundly deaf children can become more efficient in reading.

Children with hearing loss born to deaf parents learn sign language as their first language. Sign languages are not based on spoken languages. As an example, American Sign Language is not English, and it is not related to British Sign Language. Both American Sign Language and English are autonomous and have different structure. That is why we can consider people with hearing loss to be bilinguals. The language they learn from birth is not the language they are learning to read in (Goldin?Meadow & Mayberry, 2001, p. 223).

Goldin?Meadow and Mayberry (2001) investigated how deaf children learn to read and how proficient they can be. Moreover, Susan Goldin?Meadow described the differences in reading process between children with hearing loss born to hearing or deaf parents.

Hearing-impaired children born in families without hearing loss are more likely to be not very proficient in sign language. It is because their parents do not know sign language, and a child will not be exposed to an appropriate language environment since birth. While in the deaf families it is easier to adjust to the fact that the child is deaf and such children are more likely to be placed earlier into appropriate educational environments.

In sign languages there are unique signs for words and expressions, but at the same time there is fingerspelling, when each letter in the word is represented by a different hand configuration. Children without hearing loss need to learn the mapping between printed words and sounds, while children with hearing loss need to learn the relation between the language which they already know (sigh language) and printed text. At first, they learn signs and hand configuration for spelling a word and only afterwards they learn how to read.

Children with hearing loss learning to read not necessarily receive oral training (Hanson & Fowler, 1987). Deaf readers not always rely on phonological information in reading tasks (Bйlanger, 2015). It could be helpful for hearing children at the early stages when they learn to read out loud. But there is no such task for deaf readers. Moreover, Nathalie Bйlanger pointed out that both skilled and less-skilled readers with hearing loss do not necessarily need to rely on phonological codes in reading tasks. Put it otherwise, phonological codes do not determine whether an individual will be a skilled or a less-skilled reader.

It should be mentioned that there is another difference between children with hearing loss born in deaf and hearing families. Mayberry (2001) conducted a longitudinal study with two groups of ASL-speaking children. The children were tested with the stories presented in three different formats. Firstly, there was an ASL story in order to find out how familiar they are with the sign language. Secondly, a MCE story to understand the level of their knowledge of English conveyed in signs. And the last one was an ordinary story printed in English. The task was to “read” these stories and answer some comprehension questions.

Children with hearing loss born to deaf parents as they grew older tend give more correct answers for all three types of stories, while children with hearing loss born to hearing parents did not give many correct answers to the ASL story. The possible reason is that they were unable to learn ASL since their birth or at early ages. However, they performed as well as the deaf children of deaf parents in MCE stories. The most significant difference between these two groups was in the printed English stories. At the ages of 7-9 years old, children from both groups got less than to 50% of questions correctly. But at the age 13-15 years old, children of hearing-impaired parents answered almost all questions correctly, while children with hearing loss of hearing parents got only 50% of correct answers. Thus, the hearing-impaired children of parents with hearing loss had a better progress in reading printed English. `Continued growth in a language related skill such as reading appears to depend on successful and steady language acquisition throughout early childhood and elementary school. Signing skills turn out to be the best predictors of reading skill. Apparently, knowing a language - even a manual language with different structure from the language captured in print - is better for learning to read than not knowing any language' (Goldin?Meadow & Mayberry, 2001, p. 226).

A large number of prominent studies of eye movements during reading in deaf population were conducted by Nathalie Bйlanger and her colleagues. In 2012, Nathalie Bйlanger with colleagues conducted an eye-tracking experiment where they found that there is a difference in peripheral vision between the groups of deaf and hearing participants. Enhanced peripheral vision allows deaf individuals to discern 18 characters to the right from the current fixation, while the participants without hearing loss discern only 14 characters. Three years later, the same research group investigated the role of phonological codes in reading for deaf individuals and demonstrated that deaf individuals do not rely on phonological codes, no matter whether they are skilled or less skilled in reading (Bйlanger, 2015). Therefore, the absence of reliance on phonological codes alone cannot explain reading skill variation within the group of deaf individuals.

Apart from reading studies, there are also several papers describing the benefits of an enhanced peripheral vision in participants with hearing loss. Daphne Bavelier and her research group conducted a series of experiments focusing on the visual search tasks. The main finding of these studies is that enhanced peripheral vision does not help skilled readers with hearing loss and less skilled in the same way. For a group of skilled readers, enhanced peripheral vision is a benefit only when it comes to reading. For a group of less-skilled deaf readers, enhanced peripheral vision is advantageous in visual search tasks more than it is for hearing or skilled readers with hearing loss (Bavelier, 2006). The research group also found that hearing-impaired participants respond faster when they need to find a target image among distractors. They were quicker than the participants without hearing loss to predict the direction of movement of the object located in the peripheral zone. Another finding reported that deaf individuals are more sensitive to the distractors in the peripheral zone, while participants without hearing loss do not pay any attention to the distractor when it is not in the central zone.

To summarize, deaf individuals largely tend to experience serious problems with reading, most likely because they are not native speakers of spoken languages.

2.1.2 Typically developing hearing children

In general, reading patterns are not the same for adults and children, and there is a hypothesis about the differential contribution of phonological and spelling information to the reading process in children (Grainger et al., 2012; Ziegler, Perry, and Zorzi, 2014). At the early stages of learning to read, children rely heavily on phonological information (phonological codes associated to meaning prior to reading), but over time, more experienced readers increase their reliance on spelling and context. These two stages are associated with the earlier maturation of effective phonological processing, normally formed by the beginning of school, and the subsequent gradual formation of the spelling lexicon, expanding as the reading experience increases (Bowers and Newby-Clark, 2002).

Hazel Blythe and Holly Joseph describe the basic and developmental changes in reading patterns for children (Blythe & Joseph, 2011). First of all, the ability to control eye movements improves with age: adult readers are more accurate and capable of controlling their reading speed. Due to adults being better in oculomotor control, they read a line of text with a sequence of extremely quick saccades, while children need more fixations and longer fixation durations due to the lack of experience and lower oculomotor control. Normally, the main characteristics of children's eye-movement behaviour reach the level of the adult at the age of 11. For example, Vitu et al. (2001) showed that children over 11-years-old did not appear to differ from adults in the location of their first and second fixations on a word.

There are also differences in reading proficiency and reading patterns between children of the same age (Bishop &Snowling, 2004; Grant, & Karmiloff-Smith, 2001). For example, less experienced and less skilled readers have a slower reading speed and, presumably, limited perceptual span or poor attention span. Another good predictor of the skilled reader is lexical processing ability. It generally correlates with the behavior of the eye movements while reading sentences, for example, total reading times are longer on low-frequency words than on high-frequency words.

2.1.3 Children with dyslexia

Dyslexia is a neurodevelopmental reading disability that impedes reading fluency and text comprehension (Benfatto et al., 2016). It may be caused by impairment of phonological processing (Olson et al., 1983), oculomotor impairment (Martos& Vila, 1990) or the combination of both (Eden et al., 1994).

Eden et. al. (1994) found that there are some differences in saccades between children with dyslexia and children without diagnosed reading disorders. Children with dyslexia experience difficulties in taking their eye movements under control due to the oculomotor deficit. They typically make shorter saccades and more fixations during reading compared to typically developing children (Rayner, 1998). Moreover, they tend to have more regressions because they are unable to plan and control their saccade landing position - children with dyslexia have to reread more often. However, it is also important to keep in mind that there are various types of dyslexia: while most children diagnosed with dyslexia have problems with phonology, in some children these problems are accompanied by visual and oculomotor deficits.

Nevertheless, children with dyslexia can be closer to the reading level of typically developing children when the text consists of short and more frequent words. Rello et al. (2013) sought to establish whether frequent or short words improve readability and understandability for children with dyslexia. They found that participants with dyslexia read more frequent words faster; they also made fewer fixations than in less frequent words. Besides, when reading a text with shorter words, they were also faster, and the level of understandability was higher than in the texts which contained longer words.

Children with dyslexia make more fixations and regressions during reading and shorter saccades than children without reading disorders. Binocular coordination during and after saccades in children with dyslexia is worse than in the control group of children without reading disorders. “Dyslexic children showed a larger standard deviation of their fixation disparity during fixations than non-dyslexic children, and this effect was more pronounced for the close reading distance, reflecting a remarkable demand on fusional processes to obtain a single clear vision of the words” (Jainta&Kapoula, 2011, p. 7).

2.2 Current study

As we already know, learning to read is more difficult for deaf than for hearing children because the deaf cannot rely on the phonological codes and must rely exclusively on spelling (Bйlanger, 2013). Moreover, they read in a foreign language as their first language is sign language. At the same time, the hypothesis about the differential contribution of phonological and spelling information to the reading process should not work in this case at all. Children with hearing loss could not rely on phonology even at the early stages of learning to read.

When we compare reading patterns in children with hearing loss, children with dyslexia, and the control group of typically-developing children, children with hearing loss may have lower reading skills than the typically developing children. We expect children with hearing loss to not have the same level of proficiency in Russian as typically-developing children, so it is more likely that comprehension response accuracy for hearing impaired children will be lower in comparison with hearing children. Moreover, children with hearing loss used to learn sign language since birth. That means that they read in a foreign language. In this case they will probably have longer total reading times, lower comprehension accuracy and have more regressions than typically-developing children.

At the same time, RSL-speaking children might have more developed peripheral vision which allows them to see more characters to the right from the current fixation than typically developing hearing children. Children with hearing loss are more efficient in visual information processing, for that reason they are more likely to need fewer fixations to extract the visual information than hearing children and children with dyslexia. We hypothesize that children with hearing loss may be more proficient in using spelling: they probably will have shorter fixations and gaze durations and the reading times for longer words may be less than for our control group.

Overall, children with hearing loss may have both advantages and disadvantages in reading. RSL-speaking children are less familiar with Russian language than hearing children or children with dyslexia. Even if hearing impaired children are proficient in spelling, we still cannot expect them to have shorter total reading times as they are reading in a foreign language.

Current study explores whether children learning to read already have the advantage in reading due to more developed peripheral vision. Additionally, we decided to compare the results of children with hearing loss with the results of children with dyslexia. As we already mentioned, the main deficit of children with dyslexia is believed to lie in phonology, but it can also be accompanied by an oculomotor deficit. Both children with dyslexia and children with hearing loss have different patterns in visual information processing which should be represented in their eye movements during reading. We expected that children with hearing loss should have shorter fixations and gaze duration, as well as shorter total reading times for longer words compared to the group of children with dyslexia because children with hearing loss might have more developed peripheral vision compared to typically developing children and children with dyslexia. However, total reading times for longer words may be the same for hearing impaired children and children without hearing loss and reading disorders. Typically developing children do not need more developed peripheral vision in order to read longer words, they could already know them and it is easier for such children to make a guess. Children with dyslexia should have longer fixations, shorter saccades, more fixations and more regressions than children with hearing loss due to the oculomotor deficit. Reading comprehension was hypothesized to be on the same level for children with hearing loss and children with dyslexia.



3. Methods

3.1 Participants

We collected eye movements while reading in primary school children with (N = 4, Mage = 8.75, range 8 - 10 years old, 3 female; data collection is ongoing) and without hearing loss (N = 38, Mage = 8, range 7 - 9 years old, 19 female; data was collected as a part of a different project, and was made available to us by research fellow A. Lopukhina). Children with hearing loss communicate in Russian sign language (RSL) on a daily basis from birth or early age. For more detailed information, see Table 1.

Table 1. Children with hearing loss

ID	Age	Gender	Grade	Vocabulary	Raven	Degree of hearing loss
1	8	Female	2	16000	32	Severe
2	9	Female	3	14000	35	Moderately severe
3	8	Male	2	13000	NA	Severe
4	10	Female	5	19000	NA	Profound

Moreover, we compared the reading patterns of children with hearing loss with the reading patterns of children with diagnosed dyslexia (N = 29, Mage = 9.55, range 7 - 11 years old, 10 female). This data was collected as a part of a different project, and was made available to us by our colleagues in National Research University “Higher School of Economics”.

For more detailed information about control group of typically developing hearing children and children with dyslexia see Table A1 in the Appendix section.

3.2 Materials

3.2.1 Design and stimuli: reading experiment

All children read the same 33 sentences comprising the child version of the Russian Sentence Corpus (Korneev et al., 2017) and answered two-choice comprehension questions after every 3 sentences.

This corpus includes 30 sentences selected from National Russian Corpus (http://www.ruscorpora.ru). Sentences from the child version of Russian Sentence Corpus consist of 6-9 words, and have various syntactic structures: simple sentences, sentences with coordinated parts, sentences with participle clauses, with verbal adverb phrase or sentences with subordinate clause, see examples in the Table 2.

Table 2. Stimuli

Sentence type	Sentence
Simple sentence	Коробку с подарками украшал бант огромного размера. The gift bow was decorated with a huge bow.
Sentence with coordinated parts	У них был уютный, спокойный дом, крепкая семья. They had a cozy, quiet home, and a close-knit family.
Sentencewithparticipleclause	Я отдал последнюю монету, найденную в кармане. I gave the last coin which I found in my pocket.
Sentence with verbal adverb phrase	Дорога вела в глухой лес, петляя по склонам. The road led to a dense forest curving along the slopes.
Sentence with subordinate clause	Чтоб было удобнее, поправь ремешок своего рюкзака. To make it more comfortable, adjust the strap of your backpack.

3.2.2 Design and stimuli: visual search task

This experiment is a classic visual search task with so called `alien road' sign stimuli (see Figure 2). In general, we had 64 signs which could be located on squares (2.5° Ч 2.5°) or diamonds (square rotated 45°) colored in blue (RGB: [0, 79, 162] on 0 to 255 range) or brown (RGB: [162, 84, 0]). All features (the symbol itself, color and shape) were chosen in a random way so that each icon is unique (Chetverikov et al., 2018). For example, if we are searching for blue square Yoda, there will be no brown or diamond Yodas.

3.2.3 Apparatus

Participants were tested individually with the EyeLink 1000+ desktop mount eye-tracker using a chin rest. They were seated at a comfortable distance of 55 cm from the camera and 90 cm from the monitor. In this setup, one character subtended 0.29° visual angle. Only the right eye was tracked, at a rate of 1000 Hz. The experiment was carried out on a 24-in. ASUS VG248QE monitor.

3.3 Procedure

All the participants took part in the experiment with the written consent of their parents. The procedure started with the instructions and a questionnaire about a child, his or her reading experience and the use of sign language in the family. All the forms were filled out by the children`s parents. After this the experimental procedure began.

Firstly, we invited our participants to take part in a vocabulary test. As we already mentioned, the vocabulary test, Raven's Colored Progressive Matrices and a visual search experiment targeting peripheral vision were available only for our target group of children with hearing loss. The data for two other groups of children was collected by our colleagues that is why we do not have vocabulary test data and visual search experiment for the children with dyslexia and for the control group of children without hearing loss and reading disorders. Nevertheless, for the group of typically developing children, we also have data for nonverbal intelligence test (Raven's Progressive Matrices).

The reason why we decided to give our participants the vocabulary test is that it was extremely important to include more information about their reading experience. It took about 5 minutes to complete the task. The test is available online: https://www.myvocab.info/. In this test participants` task was to figure out if they know the meaning of the word presented on the screen. There were two options `знаю' (I know) if they knew the meaning of the word, and `незнаю' (I do not know) if they do not know the meaning. After some trials, participants were asked to clarify the meaning, i.e., to choose the closest synonym or description out of four presented on the screen. For example, for the word `оттенок' (shade) there were four options: `подарок' (gift), `билет' (ticket), `цвет' (color), `длина' (length).

Afterwards, participants took part in two experiments with eye-tracker.

We communicated the instructions on how to participate in the experiment with the help of participants' parents. The parents received the instructions in written form, and then they explained the instructions to their children in Russian Sign Language. Our participants were able to ask any questions and clarify the procedure with the help of their parents.

We decided to give the reading task first, because reading could be more exhausting for primary school children than visual search task. The stimuli were displayed in the black font on a light grey background so that the contrast with white background would not make reading physically uncomfortable. The experiment began with the instructions for the calibration, followed directly by the 9-point calibration procedure itself, followed by the instructions for the experiment. The calibration was performed before the beginning of the experiment, after the training part (3 sentences), and after every 15 sentences afterwards.

After reading the instruction, participants read three training sentences. If they had no questions regarding the experimental procedure, the experiment began. Stimuli were presented in two blocks, between which the participant could take a break. One block consisted of 15 trials.

Each trial began with a black fixation point at the position of the first letter of the first word in the sentence. If the participant fixated it for at least 500 ms, the sentence presentation automatically commenced; otherwise, in 5000 msthe 9-point calibration was repeated. Sentences were presented in one line in the middle of the screen against a light gray background. After finishing reading the sentence, participants were instructed to look at the red dot in the lower right-hand corner of the screen. If the participant did not fixate this dot in 150000 ms, the recalibration process was repeated. If the participant successfully fixated the red dot, the experiment continues with the comprehension question or with the next sentence. To ensure that participants read the sentences for comprehension, 1/3 of the sentences were followed by an easy two-choice comprehension question; the response was recorded from a mouse click.

For the visual search experiment, the procedure was different. The task for participants was to search for a unique target image on the screen (see Figure 2). After 9-point calibration, participant saw 36 trials. For every trial, a participant had to press the `space' when looking at the center dot, after that he or she was shown the search target sign for 1 second. Once it was removed, it was replaced by a matrix of images in 6 rows and 6 columns, and participant's task was to search for the target sign and report as soon as the target was found among other pictures present on the screen. If the target was present (the target was present in 66% of trials), participants pressed the right mouse button; if the target was absent (it was absent in 33% of trials), the left mouse button. Each trial ended when participants have responded, or when 8 seconds have elapsed, whichever comes first. Location of the target image on the screen during the search phase was selected randomly.

This experiment consisted of three blocks. The difference between them is in the immediately viewable area for each block: one block with full search array visible (Figure 2), and the other two blocks when only a part of the screen is visible through a participant-controlled aperture. The difference between two aperture blocks was only in the diameter: 5 or 9 (see Figures 3 and 4) degree diameter visible (Chetverikov et al., 2018). The baseline condition was run first for all participants to familiarize them with the task and search stimuli, while two aperture blocks were presented afterwards.

Figure 2.No aperture (baseline)

Figure 3.5?aperture

Figure 4.9?aperture

The last experiment was Raven's Progressive Matrices - a nonverbal intelligence test. The children's version for 5-11 years old children consists of 36 colored patterns. The task is to identify the missing part that makes a given pattern complete. Almost all of the patterns were presented on a colored background while the very last few items are in black and white. This experiment was send to the parents of our participants in pdf version via email with the detailed instructions. Due to pandemic-related restrictions, we were unable to arrange another meeting to complete the experimental procedure - that is why we asked our participants to do it in this way. Parents kindly agreed to help us and did this test with their children by themselves. They sent us back the list of answers so we could analyze them and give a detailed feedback. See Table 3 for the results of typically developing children.

Table 3. RPM results

Age	Average value	Range	Age	Average value	Range
4.5 - 5.5 years old	14	8 - 22	7.5 - 8 years old	23	16 - 31
5.5 - 6 years old	17	12 - 24	8 - 8.5 yearsold	24	17 - 32
6 - 6.5 years old	18	13 - 27	8.5 - 9yearsold	26	18 - 34
6.5 - 7 years old	20	14 - 29	9 - 10yearsold	29	20 - 35
7 - 7.5 years old	22	15 - 30	10 - 11 yearsold	32	21 - 35

According to these results, two participants from our target group of children with hearing loss who completed this test had normal nonverbal intelligence. Moreover, both of them had higher results than the average value for their age.



4. Results

4.1 Reading experiment

For the group of children with hearing loss saccade landing position was closer to the center of the world than for the control group of typically developing children (see Figure 5). As we mentioned, saccade is a rapid movement of the eye between fixations. Shorter saccades and their landing positon closer to the beginning of the word are more typical for less experiences readers, while landing position near the middle of the word allows the reader to extract visual features of the word most efficiently and recognize the word's identity based on single fixation. This result may confirm the hypothesis of the early benefits of developed peripheral vision and greater parafoveal preview.

Figure 5.Landing position

On top of that, children with hearing loss exhibited other characteristics of more efficient readers in comparison to the group of typically developing children. For example, children with hearing loss had higher probability of skipping a word in comparison with the control group of typically developing children without hearing loss (see Figure 6). Moreover, children with the hearing loss have higher probability of making a single fixation on a long word, while children from the control group are more likely to fixate longer words more than once. For typically developing children, with increasing word length, the probability of fixating this word more than once increases, while children with hearing loss displayed overall lower probability of fixating a word more than once.

Figure 6. Mean probabilities of skipping the word, fixating it once or twice & times depending on word length

Figure 7 illustrates the relationship between the number of fixations on the sentence and its length. Children without hearing loss have more fixations on the sentences compared to the RSL-speaking children. All children make more fixations on longer words, but in children with hearing loss this increase is smaller than in hearing children. There were no significant differences in the number of fixations in other measures for children with hearing loss and typically developing children without hearing loss and reading disorders.



Figure 7. Number of fixations

Moreover, hearing impaired children slowed down on longer words less than the hearing ones (see Figure 8). However, results for first fixation duration did not reach a statistically significant value (for all predictors p-value was greater than 0.05). Best predictors for total reading time for a word were word length (p-value < 0.001) and group (p-value = 0.018) of reader. Total reading time was significantly lower for children with hearing loss than for typically developing hearing children.

Figure 8.Reading measures



However, comprehension response accuracy was a little bit lower for children with hearing loss. Typically developing children had 94% of correct answers, while children with hearing loss answered 87% of questions correctly. Figure 9 illustrates individual comprehension response accuracy results for each participant with hearing loss.

Figure 9. Comprehension response accuracy for children with hearing loss

More detailed analysis and comparisons between children with hearing loss, children with dyslexia and typically developing children without hearing loss and reading disorders have shown that children with hearing loss have demonstrated the highest probability of skipping a word in comparison with other groups (see Figure 6 for comparison between children with hearing loss and typically developing children). This characteristic is typical for proficient readers and likely to stem from more developed peripheral vision or more successful extraction of visual information during fixations in general. Both children with hearing loss and children with dyslexia have shown longer gaze duration for longer words, although this effect was significantly lower in children with hearing loss. Among all three groups, children with dyslexia had the lowest question response accuracy rate (83%), but it not significantly different from that of children with hearing loss (87%). Both children with hearing loss and children with dyslexia seem to have similar comprehension levels.

4.2 Visual search task

As for the visual search task, our predictions were the following: first of all, if children with hearing loss have improved covert attention, they should perform better in the baseline condition (no aperture). We expect them to have more correct answers when the target image is present and shorter reaction times. However, both children with hearing loss and typically developing children are more likely to have longer reactions time when the target image is absent. If our assumption is true, children with hearing loss should be more hindered (it is possible that they may demonstrate a larger drop in performance) for smaller apertures. They probably will have longer reaction times compared to the baseline condition. This may simply mean that there will be no difference between the groups for the smallest aperture.

However, there were technical problems; because of them we were unable to collect the data for all our participants. Moreover, due to the pandemic restrictions we had the data for only one participant and no data for control group of typically developing children. We are going to collect the data for our target group of children with hearing loss (target = 15) and data for the control group of hearing (target = 15).

4.3 Vocabulary test

We expect that the level of vocabulary proficiency in our target group of children with hearing loss will be lower than in typically developing hearing children. The main reason for this prediction is that children with hearing loss presumably do not have enough experience in reading in general.

As it was already mentioned, the receptive vocabulary was measured by an online test. We compared results of our participants with the available results for primary school children on a website of the experiment.

Our first participant had 16000 words in her receptive vocabulary. According to the results of her peers she is better than 16% of other 8-years-old children. The average number of words for a typically developing child should be around 23000 - 24000 words.

Figure 10.Vocabulary test participant 1

Another 8-years-old participant had 13000 words in his vocabulary. That means that his results are better than those of 11% of the other participants of this test (Figure 11).

Figure 11.Vocabulary test participant 3

Both 8-years-old children have poor vocabulary comparing to their peers.

For 9-years-old children the average number of words in their vocabulary is around 26000 words. However, our 9-years-old participant is estimated to have a vocabulary of 14000 words - this result is higher than that of only 6% of her peers (see Figure 12).

Our last participant is a 10-years-old girl. She had 19000 words in her receptive vocabulary. But the average number of words for typically developing children in the age of 10 is 28000 words. Results of our last participant are higher than for 14% of her peers (Figure 13).

Figure 13.Vocabulary test participant 4

These results confirm our hypothesis about low vocabulary level for the children with hearing loss. For them Russian is a foreign language, probably that is why they do not know so many words as their peers.

4.4 Raven's Progressive Matrices

A nonverbal intelligence was measured by Raven's Colored Progressive Matrices (Raven, 2004). The task is to identify the missing part that makes a given pattern complete. In Table 3 there are results for typically developing hearing children between the age of 4.5 and 11 years old.

According to this, our participants who completed this test had normal nonverbal intelligence. Our first participant got 32 points, while the average value for her age 26 right answers (from 36) in a range from 18 to 34. Second participant is 9 years old girl. She got 35 points in this test her results are even higher than the average value for her age (average = 29, range 20 - 35).



5. Discussion

We expected that children with hearing loss may have both advantages and disadvantages in reading. As they are less familiar with Russian language than typically developing hearing children or children with dyslexia, we expected them to have lower accuracy level. However, our hypothesis was not confirmed. Children with hearing loss demonstrated high comprehension level. Their results were even numerically better than for the group of children with dyslexia (87% vs. 83%). Nevertheless, accuracy level for children with hearing loss was slightly different from the results of control group of typically developing children (who demonstrated high accuracy of 94%).

At the same time, we did not expect children with hearing loss to have shorter total reading times as they are reading in a foreign language. On the contrary, as we expected, children with hearing loss had fewer fixations and shorter reading times than hearing children and children with dyslexia. It seems that children with hearing loss have more developed peripheral vision which gives them such advantages compared to the typically developing hearing children. The advantage might not necessarily stem from the peripheral vision, but from the effective visual information processing in general and precise oculomotor control. Anecdotally, during data collection we noticed that children with hearing loss deal better with the calibration process compared to children with dyslexia or typically developing children. Children with hearing loss usually completed the calibration process from the first try, while typically developing children and children with dyslexia are more likely to have recalibrations. Hearing impaired children easily fixated their eye on the small dot used for calibration. Moreover, children with hearing loss almost did not need any recalibrations during the whole experiment (only after short breaks when they moved their head from the chin rest). This characteristic is more likely to stem from better oculomotor control. It seems that children with hearing loss are better in controlling their eye movements.

Children with hearing loss additionally completed an online vocabulary test and a nonverbal intelligence test, Raven's Colored Progressive Matrices.

The results of vocabulary test supported our assumption about the low proficiency level in spoken Russian language in hearing impaired children. All participants demonstrated significantly lower results in this test than their peers without hearing loss. The average number of words for a typically developing child at the age of 8 should be around 23000 - 24000 words. While both our 8-years-old participants had from 13000 to 16000 words in their vocabulary. For 9-years-old typically developing children the average number of words is around 26000. Our 9-years-old participant had in her vocabulary only 14000 words. A 10-years-old girl - out last participant is estimated to have a vocabulary of 19000 words, when the average number of words for typically developing children in her is 28000 words.

...