Language development, reading and word learning in Autism Spectrum Disorder (ASD): a review on eye tracking studies

Abstract

Objectives: The purpose of this article is to review research that has been conducted over the past five years on language development, reading skills and word learning with the use of the eye tracking machine as regards the population with autism spectrum disorder (ASD) in comparison to typically developed population.

Materials and methods: A combination of relevant terms from Google Scholar, Research Gate and PsychINFO databases was used and as a result 24 studies emerged. The total number of studies that met the inclusion criteria was 21.

Results: Studies focusing on the language development of people with ASD have presented significant findings regarding vocabulary processing and the preferential focus on specific stimuli (images, audio) over a wide range of ages that in some cases have not been investigated until then. Furthermore, studies on reading have found that adults with ASD showed a strong preference for images and symbols over texts, longer reading time and performances similar to the typical population in vocabulary processing. Studies on word learning demonstrated that adults with ASD are able to rely on gaze cues in order to learn a new word and they have the ability to use syntactic bootstrapping. For preschool and early school-aged children with ASD the results showed that they are capable of cross-situational learning.

Conclusions: This review provides information on the effectiveness of the eye tracking method as a tool that can contribute to the identification of deficits in language processing on the part of individuals with ASD from early childhood to adulthood, and more specifically as regards the domains of language development, reading and word learning.

Introduction

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder characterized by impairment in social interaction, communication and repetitive and restricted behaviors and interests (Volkmar et al. 2005). The development of people with ASD presents significant differences to that of typically developing people, in terms of various aspects, including language. Language is commonly impaired or absent in people with ASD, and if present, it tends to be used for instrumental, rather than social, purposes, with its content being repetitive and egocentric (Boucher 2003). In addition, reading skills have been reported to vary among different ASD groups, and while their reading accuracy remains at high level, their reading comprehension skills are low. At word level, the ASD population adopts an immature, mimetic approach to learning words and fails to associate the reference intention with the words (Parish-Morris et al. 2007).

More specifically, language development in children with Autism Spectrum Disorder (ASD) is complex and deficient, and language delay is not only a common feature, but also, often, the first recognized symptom (Boucher 2012). Researchers have identified significant discrepancies during the speech process of children with ASD. In the majority of cases, a significant percentage of these discrepancies are encountered as abnormalities in the use of language (e.g. phonics, stereotypes, pronoun reversal, neologisms, unusual intonation, etc.), although syntax, vocabulary and articulation are generally believed to follow developmental standards (Boucher 2012). Significant deviations have been found mainly in speech as a lot of children with ASD develop limited speech and show deficits in pragmatic skills. Comprehension and use of non-verbal communication signals are also reduced, regardless of whether language is present or not. Thus, comprehension and use of facial expressions, gestures, ironic statements and participation in funny word games always develop in an abnormal way in contrast to vocabulary and grammar (Boucher 2012).

In addition, many studies confirm the presence of reading difficulties in children with ASD, attributing their appearance to the difficulties that these children present in verbal communication and expression (Catts and Kahmi 2005, Newman et al. 2007). Readers with ASD tend to show a discrepancy between comparatively good word reading and poor reading comprehension skills (Brock and Caruana 2014, Ganz and Flores 2009, Huemer and Mann 2010, Jones et al. 2009). In addition, Gabig (2010) attributes the inadequate reading ability of children with ASD to the deficits they present in the field of phonology. It has also been argued that many children with ASD have an early interest in letters and reading, thus they recognize words that are impossible to understand at such an early age (Newman et al. 2007), an ability referred in literature as "hyperlexia" (Newman et al. 2007). The hallmarks of hyperlexia are advanced word recognition in children who otherwise have significant cognitive, linguistic, or social handicaps; a compulsive preoccupation with reading, letters, or writing, and a significant discrepancy between strong word recognition and weak comprehension of what has been read (Tager-Flusberg et al. 2005). However, it has been argued that only a subset of children with ASD show a remarkable decoding ability and only 5 to 10% of children with autism show hyperlexia (Grigorenko et al. 2003). Therefore, despite the fact that this rate is much higher than the one which occurs in typical development, the researchers concluded that hyperlexia is not synonymous to autism (Grigorenko et al. 2003).

Several researchers have explored the word learning process in children with ASD, in terms of their ability to determine the referents of new labels (Luyster and Lord 2009, Norbury et al. 2010). McDuffie et al. (2006) reported that the introduction of labels increases attention to novel objects for young children with ASD. However, these children appear to be less proficient at maintaining and following attention, indicating that they may only successfully map words if an adult follows the child’s attention. In addition, Swensen et al. (2007), using an intermodal preferential looking paradigm with children with ASD, reported that children with ASD demonstrate the noun bias, in that – like typically developing children – they interpret a new word as referring to an object rather than an action. Furthermore, Preissler and Carey (2005) argued that the ASD population was much more successful in labelling correctly when the labelling process is child-focused instead of adult-focused. However, Parish-Morriset al. (2007), suggested that in word learning situations, where the label was directed towards a “boring” object, or when children were required to deduce the referent of a label based on their partner’s social intent, children with autism were not consistently successful.

The eye tracking method is an electronic system that allows the registration of eye movement, using a computer, more specifically a camera, which detects the focus of the eye and the direction of the user's gaze. Typically developed adults show a very specific pattern of gaze when viewing faces, fixating mainly on the eyes, but also on the nose and mouth, the so-called ‘core features’. The emergence of the eye-tracking method as an objective and accessible tool for examining perceptual characteristics has facilitated research on abnormal visual attention and the oculomotor patterns it involves. The eye tracking method as a tool has several advantages for people with ASD. First, eye-tracking can be implemented to infants (Jones et al. 2009, Jones and Klin 2013), thus it can provide important information during a critical period in the development of ASD and it can be used for early identification of the disorder and leading to early intervention (Chawarska et al. 2012, Jones and Klin 2013). Furthermore, eye-tracking provides quantitative data that can be used potentially as biomarkers that indicate atypical visual attention and possible information processing deficits (Guillon et al. 2014). Eyetracking method provides information about eye movements which are typically divided into fixations and saccades. Fixations are the points at which gaze pauses at a certain position, and saccades are rapid movements to a new position. In the context of language and communication, there is much interest in the extent to which children with ASD fixate the eye and mouth regions of the human face, as these are purveyors of social cues important to language. Furthermore, looking at the timing of fixations, we can grasp a number of cognitive processes, such as cognitive intent, interest and attention, salience, and cognitive load (Aslin 2007). Also, first fixations can indicate the viewer's interest in specific words which he probably knows and understands more easily. The studies that have applied these methods to people with ASD illustrate the potential of such methods to yield new insights into qualitative differences in foundational skills (visual and social attention) and in the underlying mechanisms involved in processing language as it unfolds.

In their study about language development, Hosozawa et al. (2012) examined the gaze patterns of children while watching excerpts from movies or children’s shows and reported atypical gaze patterns for children with ASD, accompanied by decreased interest and involvement in social stimuli in a scene. These findings are consistent with previous research that suggested poor social communication, reduced language development and deficits in comprehension for this population (Sasson 2006·Shic et al. 2011).

In another study about reading, Sansosti et al. (2013) made a research on children with ASD and children with typical development while they were reading short texts, with the aim of examining their reading comprehension skills. They measured the average reading time and the correct answers to questions and recorded the eye movements of the participants during the reading process. Eye movement data revealed that children with ASD spent more time correcting text, making overall corrections and making more regressions while reading as opposed to typically developing controls.

In another study about word learning, Akechi et al. (2011) examined the eye fixations of ASD children while learning new words and reported difficulties on their part in using social cues such as gaze, which could help them in the process of word learning. In the same line, Norbury et al. (2010) recorded the eye movements of children with ASD and children of typical development while completing a word learning task in a social context. According to their findings, typically developing children showed consolidation of semantic and phonological information, while the population with ASD did not, indicating that the two populations follow different mechanisms in learning new words.

Therefore, on the basis of the above, the aim of the present study is to review recent studies that were conducted in the last 5 years, with the use of the eye tracking method and focused on a) language development b) reading skills and c) word learning of the ASD population in comparison to typical population.

Materials and methods

Selection of methods

Studies included in this review were selected based on the following criteria:

1. The studies were published in peer reviewed journals and written in the English language.

2. The research included experimental-quantitative methodologies.

3. The method of eye tracking and more specifically Tobii or Eyelink, was used.

4. The studies were published within the last five years (i.e. 2015–2020).

5. Participants had an official diagnosis of ASD.

Search and data description methodologies

During the first search, the time range was not restricted to studies conducted from 2015, but this yielded a large number of studies which were difficult to group and some which some were not related to the aim of the study. Therefore, and since the eye tracking methodology has been extensively used in the ASD population mostly to detect gaze behavior while looking at facial expressions, we decided to restrict the time span from 2015 to 2020. By doing so, the most recent findings on the areas of interest which are compatible to the aim of the study are presented below. The search process was conducted electronically by searching in three databases: (a) Google Scholar (b) ResearchGate (c) PsychINFO. In the search, we used the following search terms: 'reading in autism with eye tracking method', 'eye tracking reading progress-autism', 'reading ability-eye tracking -autism', 'reading behavior in autism', 'eye movement in autism-reading ability', 'language processing in autism', 'language development in ASD-eye tracking', 'language abilities in ASD', word learning in autism with eye tracking', 'word meaning in autism spectrum disorder-eye tracking'. As a result of the search process, 24 relevant studies emerged. Each of them was examined by the lead author to determine whether they met the criteria for inclusion in the present study. Three of the studies were not included as they used a camera in the methodology and did not use any of the known systems of measuring the eye movements (e.g. tobii, eyelink, eyetech etc.) or were drafts of researches that would be based on the eyetracking method in the future. The total number of studies that met the inclusion criteria was 21. The studies were divided according to their aim and the methodology they used. In the language development section the studies that were included focus on some aspects of receptive language. In the reading section, the studies focus on reading ability and used similar tasks. In the third category, studies focus on learning new words, the preference for specific parts of speech (e.g. verbs), as well as the ability to use existing syntactic knowledge to learn new words. Specifically, from the 21 studies (a) 6 concerned language development, (b) 12 reading skills and (c) 3 word learning. Table 1 presents a summary of the studies included in this review.

Table 1.

Studies on reading, language development and word learning of the ASD population. 

Author	Year	Participants		Mean age (m)		Gender		Research area
		ASD	TD	ASD	TD	Male	Female
Yaneva, Temnikova and Mitkov	2015	20	20	30.75	30.81	22	18	Reading
Au-Yueng, Kaakinen, Liversedge and Benson	2015	18	19	32.4	23.8	29	8	Reading
PlesaSkwerer, Jordan, Brukilacchio and Tager-Flsberg	2015	19	–	13.1	–	15	4	Language development
Hahn, Snedeker and Rabagliati	2015	40	40	7.8	7.5	69	11	Language development
Aldaqre, Paulus and Sodian	2015	15	15	36.9	32.5	15	15	Word learning
Yaneva, Temnikova and Mitkov	2016	20	20	30.75	30.81	22	18	Reading
Bavin, Prendergast, Kidd, Baker and Dissanayake	2016	48	56	6.68	6.62	62	42	Language development
Štajner, Yaneva, Mitkov and Ponzetto	2017	31	32	35.33	31.66	40	23	Reading
Davidson and Weismer	2017	23	23	11.07	11.07	32	14	Reading
Micai, Joseph, Vulchanova and Saldaña	2017	22	22	12.6	13	34	10	Reading
Howard, Liversedge and Benson	2017	19	18	31.37	28.33	30	7	Reading
Au-Yeung, Kaakinen, Liversedge and Benson	2018	15	16	32.2	24.6	N/M*	N/M*	Reading
Thompson, Plavnick and Skibbe	2018	10	5	6.5	5	ASD = 8	ASD = 2	Language development
Zhou, Zhan and Ma	2018	26	49	5.7	5.1	N/M*	N/M*	Language development
Righi, Tenenbaum, McCormick, Blossom, Amso and Sheinkopf	2018	45	32	5.12	3.04	57	20	Language development
Horvath McDermott, Reilly and Arunachalam	2018	32	–	3.3	–	N/M*	N/M*	Word learning
Micai, Vulchanova and Saldaña	2019	20	22	13.3	13.5	29	13	Reading
Fong, Ma, Pang, Tang and Law	2019	20	20	5.35	5.15	27	13	Reading
Ferguson, Black and Williams	2019	25	25	36.19	37.56	34	16	Reading
Venker	2019	18	21	6.33	4.83	33	6	Word learning
Barzy, Filik, Williams and Ferguson	2020	25	24	34.4	34	34	15	Reading

ASD: Autism Spectrum Disorder, TD: typically developing, N/M*: no mentioned in the studies.

Results

Results of the studies on language development in ASD

The studies which were included in this systematic review, examining language development on the part of the ASD population with the use of the eye tracking method, revealed valuable information on the subjects’ gaze behavior during the processing of sound stimuli, their focus on specific information and the behavior of the eyes after a long-time engagement with a particular stimulus.

In their study, Plesa-Skwerer et al. (2016) used several methods, the eye tracking method among others, to assess language comprehension in minimally verbal children and adolescents with ASD; they made the assumption that even with uniformly minimal expressive skills, receptive language may vary in this population, and that conventional standardized assessments often fail to capture the linguistic potential of this specific population. More specifically, they used a standardized direct assessment of receptive vocabulary, caregiver report measures, including scores on the Vineland-II Communication domain, a vocabulary questionnaire, an eye-tracking task of word comprehension and a computerized assessment, using touch screen in order to directly measure word comprehension with the same stimuli which were used in the eye-tracking experiment. The sample of the study consisted of 19 children and adolescents with a diagnosis of ASD (mean age = 13.1 years) and very limited expressive language. To assess autism diagnostic status the ADOS assessment was used (Lord et al. 1999). Participants were assessed in language (receptive lexical knowledge, vocabulary, receptive communication) and cogniton (nonverbal cognitive ability) and were matched according to their performance. In the eyetracking task, the participants were asked to look at two pictures on a computer screen, hear a word related to the pictures a few seconds later, and focus on the picture to which the word they had heard referred. The target words included nouns, verbs and adjectives. Specifically, they investigated the search time of the target picture after the onset of the auditory stimulus and explored the differences between the picture search time, when it appeared as a target word, and when it appeared as a second choice in another target word recognition task. The results showed that participants looked proportionally longer at the target pictures in comparison to the second choice pictures, suggesting that eye movements provide evidence of word comprehension for minimally verbal children and adolescents with ASD. ​The results of this multi-method approach revealed significant heterogeneity in language reception skills among the ASD population compared to the typical population.

Furthermore, Hahn et al. (2015) used the eye tracking method in order to examine language comprehension skills in children with ASD and more specifically to explore whether young, highly verbal children with ASD are impaired in using context. The participants were 40 high-functioning children with ASD and 40 typically developing children, aged 6-9 years. Measures that were used for participants with ASD were the Autism Diagnostic Interview-Revised (ADI-R, Rutter et al. 2003) and ADOS (Lord et al. 1999). In addition, both groups of participants were assessed prior to their participation in the research through language and cognitive tests and matched according to their skills. Both groups were asked to listen to sentences in which the last word was missing and they had to complete the sentence by choosing the correct word from 4 different pictures they were presented with. Both ambiguous and unambiguous target words were used. The study found that young children with ASD performed similarly to the typical population. In addition, it was examined whether differences in eye movements over time showed different degrees of indirect semantic initiation among conditions (ambiguous or unambiguous) and groups. The findings suggested that there was a reliable ambiguity with context-based interaction in both groups. Finally, it was investigated the way in which children’s ambiguity resolution ability varied over age. The findings revealed a reliable interaction between ambiguity, context, and age, showing that older children are better at using context to resolve ambiguity in both groups.

In addition, Bavin et al. (2016) aimed at investigating the incremental semantic and syntactic processing in interpreting noun modification, and specifically whether the eye movements of young children aged 5–9 years with ASD were similar to those of children with TD over the course of a sentence. The participants of the study were 56 children with typical development and 48 children with a prior diagnosis of ASD, with a mean age of 6.8 years. Diagnoses of ASD were confirmed with the ADOS-G (Lord et al. 1999), and other measures on language and cognitive skills were used in order to determine the sample’s participation in the study. In the visual display for each sentence, four items of the same category of words (e.g. verbs) were provided and each display contained a picture of the target object and three other items of the same category as the target (one of these three was a possible target until the additional information was presented and the other two were distractors). The four pictures were placed in the four corners of a monitor, and the arrangement varied so that the target and the distractors appeared in different locations at different items. Nine sentences with different structures and different visual stimuli were used. For each item, the audio stimuli started 1 s after the four pictures appeared on the monitor, to allow participants to examine them briefly before listening to the sentence. Researchers encouraged the participants to look at a specific target word each time (e.g. ‘Look at the blue square with dots’), and examined the looking time and the fixation time concerning the target. The results showed that both groups initially had access to the two possible objects that matched the description, that is before listening to the modifying phrase that reduced the possibility of these two objects to one, and looked more at these two elements than at the two distractions. In other words, the children in both groups were able to process the audio information into the noun phrase and link it to the two appropriate pictures. However, children with ASD were slower to focus on the two matching objects.

Furthermore, Zhou et al. (2019), studied the ability of children to use verb information to predict the upcoming language contribution. The sample consisted of 26 five-year-old children with ASD, 25 typically developing children of the same age and 24 typically developing four-year-old children. Diagnoses of ASD were confirmed through the ADOS (Lord et al. 1999) and participants were matched accordingly as regards Mean Length of Utterance (MLU) and verbal IQ. The 5-year-olds with ASD were matched with the TD 4-year-olds on both MLU levels and IQ scores. The 5-year-olds with ASD had significantly lower MLU level and IQ score compared to their age-matched TD peers. They constructed 16 target objects with 1 picture and 2 sentences each, with a verb showing strong bias, and a neutral verb. There were 3 objects in each picture, 1 object matching the situation and the others being neutral. The researchers encouraged participants to look at the picture that best matched the oral sentence they heard during the process. They divided each picture into 4 areas of interest and examined the percentage of fixations in a specific area at a specific time location. The results showed that 5-year-olds with ASD, like TD participants, presented verb-based anticipatory eye movements during real-time sentence comprehension. No difference was observed between the ASD and TD groups in the time course of their ocular patterns, indicating that preschool children with ASD are able to use verb information efficiently and quickly predict future language input.

In another study, Righi et al. (2018) examined audio-visual speech processing in children with ASD, as compared to TD controls, and recorded the relation between audio-visual speech processing and language skills to understand factors that can lead to language deficits. The sample consisted of 77 participants, 45 with ASD (mean age; 5.12 years, 36 male, 9 female), and 32 typically developing participants (mean age;3.04 years, 21 male, 11 female). Diagnoses of ASD were confirmed through the ADOS (Lord et al. 1999) and all participants underwent cognitive and language testing. Two identical video recordings were presented on a computer screen and one audio track, matching either one or both recordings, was put. The recordings were 14 sec long each, and showed a woman speaking about topics falling into the interests of young children. The stimuli were presented in four conditions that differed in length of delay of the asynchronous recording: a) both recordings were synchronous with the audio track b) in three conditions, one of the recordings preceded the audio track by 0.3 s, 0.6 s or 1 full second. Τhe researchers asked the participants to look at the screen and examined the areas of interest and the time participants spent looking at each area. More specifically, they divided the stimuli into two areas of interest, one focusing on the mouth and the other on the eyes. The results showed that TD participants were successful in discriminating between the synchronous and asynchronous recordings and showed a preference (more fixations) for the ones that were synchronous, while the ASD participants did not demonstrate any preference for the synchronous over the asynchronous recordings, at any of the delays. Comparisons between the groups showed a tendency among ASD participants to look less at the synchronous recordings, as opposed to TD participants.

Another study (Thompson et al. 2019), using the eye-tracking method, examined gaze behaviors of children with ASD and typically developing children when presented with an e-book. The participants were 10 minimally verbal children with ASD and 5 typically developing children, with a mean age of 5 years. The studies used an e-book consisting of 12 pages where each page contained 10-12 words and a picture related to the topic it dealt with. The researchers created three conditions: a) silent presentation b) audible reading c) text that was both read aloud and highlighted to the participant, while each page was presented to children for 8 s. For the purposes of this study, areas of interest were created for the entire slide, the text, and the picture on each page of the storybook. This allowed the determination of the total fixation duration for the parts of the slide with text and the parts with a picture. It also allowed the comparison with the total fixation duration for white spaces (i.e. total e-book page minus text and picture). Fixation durations were measured and reported in seconds. Subsequently, they examined the instance of sequential paired dots, counted as left-to-right gaze, which was assessed for all participants across each condition. The findings of the study showed that children with ASD did not focus much on the pictures and print embedded within the e-book. More specifically, children with ASD attended the e-book for approximately half of the time it was displayed and only half of their attending time was allocated to salient stimuli within the e-book. In addition, it was found that some children with ASD demonstrated slightly higher levels of attending to print when the text was both read aloud and highlighted, compared to the text being solely presented or only read aloud. Text highlighting also seemed to match looking from left to right and make it easier for readers with ASD.

Results of the studies on reading skills in ASD

The use of the eye-tracking method to investigate reading skills of the population with ASD revealed important information about the gaze fixations on different stimuli, the focus on a specific subject, the strategies the participants used to understand written text and the time required by this population to complete the processing of the text. The studies that were included in this section focused on measuring similar variables such as the first pass reading, the regression path reading, the re-reading time, the regressions in/out and the skipping. Specifically, the first pass reading is defined as the summed duration of the fixations on a region until readers moved their eyes to fixate on another region. The regression path reading time is defined as the sum of all fixations from the very first - entering a region from the left side- until the participant made a fixation to the right of this region. The re-reading time is defined as the total durations in a region after having left that region to the right. Regressions refer to the probability of making a leftward eye movement towards the target word having already “left” that word to the right. Regressions-out refer to the probability of making a leftward eye movement away from the target word before leaving the word to the right. Skipping is defined as the proportion of trials in which a region was skipped.

More specifically, Yaneva et al. (2015) studied the differences between autistic and non-autistic people with regard to the time they spent looking at pictures and text paragraphs. They examined the focus of the gaze of thirty-seven people of an average age of 30 (20 with ASD, 20 typical development) while reading. Researchers did not use language or cognitive measures in order to decide on the participants. Criteria for participation in the survey were spoken language of participants, age, diagnosis regarding ASD participants and absence of learning disabilities or developmental delay. They assumed that children with ASD focus better and are more effective in decoding text when it contains pictures. Three evaluation tools were used; texts with photographs (20 photographs in total), texts with symbols (19 symbols in total) and plain texts (with no images). Images and text paragraphs were defined as areas of interest (AOIs) and a number of gaze-based metrics were set based on how many times and for how long participants looked at these areas. Τhe participants were asked to read the texts and answer a multiple-choice question, to be tested on the comprehension of the text they had just read. Researchers used the following metrics with eyetracking method a) the average time viewed by all participants was measured in seconds, including the durations of all fixations and all revisits, b) the average number of fixations and c) the average number of revisits, d) the reading time score as the mean reading time per text in each group which was then divided by the number of words in the text. Two more measurements were used in order to examine the perceived level of difficulty and the text presentation preference. The first measurement was obtained through Likert scale, 1 to 5 items after each text through which the participants would rate the perceived level of difficulty. The second measurement was gathered through the following survey question: “In your everyday life, do you prefer reading texts with: a) photographs, illustrating the main ideas b) symbols, illustrating the main ideas, c) plain texts without any images or d) It makes no difference to me”. The results showed a significant difference in the amount of time spent looking at pictures and text paragraphs between the two populations. More specifically, the participants with ASD seemed to focus more on the pictures or symbols than on the text, in contrast to the neurotypical participants, for whom it mostly made no difference whether or not there were pictures accompanying the text.

In addition, Au-Yueng et al. (2015) recorded the eye movements of 19 typically developing participants (13 males, 6 females), aged 18–35 and of 18 participants with ASD (16 males, 2 females), aged 20–52. Participants with ASD had a previous diagnosis and all participants underwent intelligence testing. The aim of the study was to gain some insight into how individuals with ASD process irony, and to compare this with what we know about neurotypical processing of irony. They hypothesized that the population with ASD would show increased interruption in the processing of ironic statements. The participants were asked to read the experimental stimuli that consisted of 36 short texts, each made up of three sentences, with two versions of each text, one ironic and one non-ironic. They used three different eye movement measures; a) first-pass reading time, b) regression path reading and c) total reading time. Researchers used the first-pass reading times in order to examine an indication of early stages of linguistic processing; the regression path reading time reflected early processing difficulty as well as time spent reinspecting the text in an effort to recover from any initial difficulty; and total reading time provided a measure of overall processing difficulty associated with a portion of the sentence. The eye movement data showed that for both participant groups, total reading times were longer for the critical region containing the ironic statement compared to the non-ironic condition. The participants with ASD, however, spent overall more time than TD controls re-reading both types of texts, the ironic and the non-ironic version, a finding which indicates that individuals with ASD either take longer to construct a coherent discourse representation of the text, or that they take longer to make the decision that their representation of the text is reasonably based on their knowledge of the world.

Furthermore, the study of Yaneva et al. (2016) presented a corpus of text data and its corresponding gaze fixations obtained from autistic and non-autistic readers. The participants of the study were 20 adults (7 female, 13 male) with an official diagnosis of autism and 20 TD adults (11 female and 9 male) with a mean age of 30.75 and 30.81 respectively. Researchers did not use language or cognitive measures. Criteria for participation were a confirmed diagnosis of ASD participants, age, years spent in education, spoken language and not having comorbid conditions affecting reading. Τhe participants were asked to read 9 texts and answer 3 comprehension questions with four possible answers each. The excerpts used in the experiment came from educational, newspaper, and informative texts. The data was elicited through random reading comprehension testing combined with eye-tracking recording. The researchers defined each content word from the texts as an area of interest and for each area of interest three gaze-based measures were calculated: a) total fixation time, b) number of fixations of gaze on a given area of interest, and c) re-reading time. The results showed that long words caused significantly longer fixations on the part of the autistic participants compared to the TD ones. In addition, significant differences were found between the two populations in the number of corrections and reviews, something that according to the researchers was in line with previous research (e.g. Sansosti et al. 2013).

In another study, Howard et al. (2017) used the eye tracking method to examine lexical identification and syntactic and semantic processing in adults with ASD and TD. The participants were 19 adults with a formal diagnosis of ASD, aged 18-52 years, and 18 adults of typical development, aged 20-52 years. Each ASD participant was measured with ADOS-4 (Lord et al. 2012) and all participants were matched through intelligence tests, expressive language ability and reading ability. In order to examine the lexical identification they used 34 pairs of sentences that included low frequency (e.g. zombie) and high frequency (e.g. people) target keywords and participants were asked to read the sentences. In the middle of the process the participants had to answer a simple comprehension question about what they had just read and respond with a “Yes/No”. In order to examine the syntactic and semantic processing, 44 sentences were devised which included a prepositional phrase that could either be attached to a verb (high) (e.g. Charlie demolished the dilapidated house with a huge crane last year.) or to the noun phrase that immediately preceded the preposition (low) (e.g. Charlie demolished the dilapidated house with a huge fence last year). The sentences were divided into regions, a) the pre-target region, which consisted of a determiner and an adjective that immediately followed the preposition b) the target, which consisted of the disambiguating noun and c) the post target region, which included the one or two words that followed the target. The following eye movement measures were calculated: a) skipping, b) first fixation durations, c) single fixation durations, d) gaze durations, e) go past times (the time from when a word is first fixated until the eyes leave this region to the right, including all time spent re-reading previous regions of the sentence) and f) the proportion of regressions. In the pre-target and target regions no effect of attachment was found for any of the measures. In the post target region no differences were found for skipping, first fixations, single fixations or gaze durations. There were differences in go past times, with both groups taking longer to progress to the right, past the post target region, when the sentences were low attached, in comparison to high attached sentences. In addition, there was a difference between groups, with ASD participants having longer go past times overall for this region. For total times, there was a main effect of attachment and group, with all participants spending more time fixating on this region for low attached sentences and the ASD group spending more time in this region overall. Both groups made significantly more regressions out of the post target region when the sentences were low attached, finding that did not differ between or interact within groups. There were no differences between groups or sentence types in the proportion of regressions made in this region.

Furthermore,  Štajner et al. (2017), studied the differences in word processing of 31 participants with autism and 32 participants without autism, aged 31-38 years. The criteria of participation in the study were the following: native spoken language, the similar number of years spent in formal education, not having diagnoses that affected reading and not having any developmental delay. Their purpose was to find out what words might be considered challenging for both groups. Τhe participants were presented with 20 texts through the eye tracking system, which they had to read and then answer comprehension questions. They used three types of texts, educational, newspaper, and general informative articles; each text was self-contained and coherent. The average number of words per text was 156 (min = 74 words and max = 242 words) and each word in the texts was defined as an area of interest to be analyzed. The output contained three gaze based measures for the total number of words, for each participant, separately: total fixation time, number of fixations on an area of interest and the re-reading time. The findings showed that even though there were no differences between the two groups as far as the level of comprehension of the texts was concerned, the analysis of the gaze showed that readers with autism produced significantly more fixations and revisits, as well as longer viewing times per word. The researchers explained these results as a possible alternative reading strategy on the part of ASD participants, and argued that it may not be that ASD readers spend more time and make more fixations because reading itself is challenging, but rather, because they simply read more cautiously.

In another study, Davidson and Ellis Weismer (2017) viewed the contributions of different supporting skills (e.g. individual differences in age/development, word reading, oral language) as a collective set of skills necessary for context integration with the aim of examining individual differences in reading comprehension of 23 children with ASD and 23 typically developing participants, aged 8–14 years. Participants with ASD have already received a diagnosis of ASD so as to enter the study. Participants were measured with standardized tests so that researchers evaluate cognition (nonverbal), oral language (vocabulary and morphosynactic comprehension) and reading (comprehension, word recognition and decoding). Τhe participants completed a word or sentence reading process where they had to understand the content and choose between two pictures, the one that best completed the meaning of the word or of the text, which they were presented with randomly, using the eye tracking method. Eye fixation measures were obtained from two areas of interest in each sentence or word: the context and the target word. They investigated the first pass reading in order to measure the time spent reading the context, including both first-pass eye movements and any looks back to the context after reading the target word and the regressions so as to measure comprehension, or the participant’s awareness. The results indicated that children with ASD, similar to their TD peers, integrated the context to access the correct homonym meanings while reading. However, after reading the sentences, when participants were asked to select the best picture that completed the meaning of the text, both groups experienced difficulty in semantic ability between the two meanings. However, this difficulty hindered sentence representation for children with ASD to a much greater extent in comparison to their typical peers.

The objective of the study of Au-Yeung et al. (2018) was to investigate whether there were time course differences in the way people with ASD detect and process anomalies in understanding the text, compared to a TD control group. The sample consisted of 15 participants with ASD (age range 20-52 years) and 16 typically developing (age range 18-35years). Participants with ASD had a previous diagnosis and all participants were matched on verbal intelligence. In an eye movement experiment they used 24 excerpts from a text consisting of 5-7 sentences, which included anomalies in passage level and sentence level. Τhe participants read the texts carefully and answered comprehension questions. Each text was auto-segmented into single regions of interest. They investigated the first pass reading time and regression path reading time. The results showed that readers with ASD had difficulty in understanding the anomalous words, in comparison with TD readers (longer regression path times on the anomalous vs. non-anomalous word). In addition, it was found that ASD readers engaged in significantly more re-reading of the passages than did TD readers (longer reading time in the regression path reading, both with regard to the sentence and passage anomalies).

Furthermore, the aim of the study of Micai et al. (2017) was to explore inference generation skills and reading strategies of children and adolescents with autism. The sample included 22 participants with ASD (mean age 12.6 years), and 22 typically developing participants (mean age 13 years). Participants with ASD had been diagnosed through the ADOS (Lord et al. 2000). The ASD and control sample were statistically matched on chronological age, nonverbal IQ, grammatical comprehension, vocabulary size, reading speed, and comprehension accuracy. Five narrative stories were created and the participants were asked to read each paragraph silently and to answer questions by choosing, by means of a key press, one of the three possible responses which were displayed at the bottom of each paragraph. They assessed the average of fixations during reading the entire paragraphs and answering the question. Next, eight local eye-movement measures were explored in relation to predefined target words: first fixation duration, single fixation duration on a word, gaze duration, regression path reading time, total fixation time, re-reading time, regressions-out and regressions-in. The results showed that the ASD group was as accurate as the control group in generating inferences when answering questions about the short texts, and no differences were found between the two groups in the paragraph reading and responding times. However, the ASD group displayed longer gaze latencies on a target word necessary to produce an inference. They also showed more regressions into the word that supported the inference, as compared to the control group, after reading the question, irrespective of whether an inference was required or not. Therefore, the ASD group achieved an equivalent level of inferential comprehension but showed subtle differences in the reading comprehension strategies they used, as compared to the control group.

Another study by Micai et al. (2019) aimed to explore the ability of individuals with ASD to detect errors when asked to do so, and more specifically, if instructions would have an impact on their ability to detect semantic and orthographic errors. In addition, the study aimed to explore if reading behavior changed, depending on different instructions and error types. The participants with ASD (n = 20, mean age 13.3 years) and the typically developing (n = 22, mean age 13.5 years) with a mean age of 13 years, read texts consisting of sentences that were spelled correctly and sentences that contained semantic and spelling mistakes. Participants with ASD had been diagnosed through the ADOS (Lord et al. 2000). Both groups measured in non-verbal intelligence, vocabulary, grammatical comprehension, reading speed, number of decoding errors (hesitations, rectifications and repetitions). One hundred and forty-four sentences in Spanish, divided into eight stories (18 per story), were created. Each story consisted of six sentences with semantic errors, six sentences with spelling mistakes and six correct sentences. The study measured the time taken for the participants to read the sentence and press the space bar in order to report if an error was present. Specifically, the total number of fixations during that time, the total fixation time, and the regression path reading time were measured. Furthermore, the researchers explored eye-movement measures in relation to the errors contained in the sentences; specifically they examined and measured: a) ​ the fixation duration on the error regardless of whether it was the only fixation or the first of multiple fixations, b) the single fixation duration, identified as the duration of the initial fixation on the error when only one fixation was performed on that error during first pass, c) the gaze duration on the error prior to moving to another word, and the go-past time, as the sum of all temporally continuous fixations including fixations after a regressive eye movement to the left of the region, until the fixation point progresses to the error to the right, d) the total fixation time on the error, e) examined the re-reading time, as the total fixation duration in a region after having left the error to the right, f) regressions-out, referring to the probability of performing a leftward eye movement out of the error before leaving the word to the right, g) regressions-into, referring to the probability of returning to the error after leaving it, and last, h) the omission of not reading the error altogether. The results, on the reading measures concerning both the entire sentences containing the semantic or orthographic errors, and the entire correct sentences, showed that the sentences which contained semantic errors received more and shorter fixations compared to the sentences containing orthographic errors. The eye movements related to error detection indicated no overall significant difference in the reading behavior between the ASD and control groups.

Subsequently, Fong et al. (2019) investigated the immediate effects of coloured overlays on reading performance, using eye tracking in preschool children with ASD and TD children. Specifically, the sample was 20 children with ASD and 20 typically developing children, aged between 4-6 years. The inclusion criteria for the ASD group were clinical diagnosis of ASD, chronological age, intelligence (IQ > 70), ability to read the top line of characters on the computer screen with both eyes open and then with each eye at about 50 cm away with or without wearing corrective lenses and ability to recognize the single-digit numbers (1–9). For TD participants’ the criteria of inclusion were the chronological age and not having any language or cognitive disability. The reading test consisted of three rows of eight single-digit numbers presented on a computer screen. The participants were required to read aloud the numbers one by one, line by line, from left to right as they appeared in random order. They had to read the test paper once with and once without their chosen coloured overlay. The researchers set the span from the third to fifth line as an area of interest to examine participants’ reading performance during the reading test. Four parameters were used in this study: total fixation duration, first pass reading, regression path reading time and number of fixations. The findings showed that colored overlays had no significant, immediate effect on improving the ocular performance and reading speed of children with ASD or TD, although individual improvements were detected in some children with ASD. Therefore, the conclusion reached was that the use of colored overlays may not be useful for improving reading and ocular performance in children with ASD especially when used only on one single occasion, though repeatedly may have different results.

Furthermore, Barzy et al. (2020) used the eye-tracking method to investigate the processing of emotional responses to ironic vs. literal criticism in autistic adults. The sample consisted of 49 participants, specifically, 25 autistic adults (mean age 34.40 years) and 24 typically developing adults (mean age 33.04 years). Autistic participants had a previous formal diagnosis and all participants were matched on gender, age and intelligence. Both groups were instructed to carefully read each scenario for comprehension, and then click, using a mouse, when they had finished reading, and proceed either to the next scenario or a comprehension question (25% of trials). The scenarios included a victim and a main character. Each sentence was separated into three regions for analysis (pre-critical region, critical region and post-critical region). The emotional response was always the critical region, the word directly preceding it was always the pre-critical region, and the word/phrase that was presented directly after, was the post-critical region. Pre-critical and post-critical regions were identical across conditions, and the critical region was equated in terms of length across conditions. Five measures of reading behavior were extracted from the eye movements: first-pass reading time, first-pass regressions out, regression path reading time, second-pass reading time, and skipping rate. The results showed that in the pre-critical word region there was a significant effect in first-pass reading times, in first-pass regressions out, and in regression path reading times, revealing that participants with ASD had longer reading times and made more regressions out when the target sentence referred to the victim’s, rather than the main character’s perspective, as compared to the TD group. As regards the critical region, there was a significant effect in regression path reading time, as participants in the autistic group had longer reading times compared to the TD group. There was also a significant effect in regression path reading time, reflecting longer reading times in the literal criticism condition, compared to the ironic criticism condition. In addition, results in this region showed that participants with ASD made more regressions out from the critical region, when the victim found the comment amusing, compared to when the main character intended the comment to be amusing. This pattern suggested that autistic participants successfully tracked the two characters’ perspectives, and were immediately sensitive to the victim’s expected emotions following the criticism, but, importantly, did not distinguish between literal and ironic criticism. In the post-critical region, there was a main effect of emotion in regression path reading times with longer reading times when the character was described as feeling amused, compared to when the character was described as feeling hurt. Comparisons revealed that TD participants made more regressions out, when the victim perceived the literal criticism as amusing, compared to when the victim perceived the ironic criticism as amusing. Autistic participants showed no difference in this regard.

In another study, Ferguson et al. (2019), examined how adults, with and without ASD, make sense of reality-violating fantasy narratives by testing real-time understanding of counterfactuals. Participants were 25 adults with ASD (mean age 36.19 years) and 25 typically developing adults with a mean age of 37.56 years. Participants in the ASD group had all received formal diagnoses and both groups were matched as regards gender, age, verbal intelligence, being native English speakers, having normal or corrected to normal vision, and not having a diagnosis of dyslexia or intellectual disability. Participants were asked to read, silently and at their own pace, narratives that consisted of 4 types of scenarios (consistent counterfactual, inconsistent counterfactual, consistent factual and inconsistent factual). The experimental passages were divided into three regions of analysis: a) pre-critical region, b) critical region, and c) post-critical region. For each region, the first fixation duration, the first-pass reading time, the first-pass regressions out, the regression path reading time and the total reading time were examined. For the first fixation duration there was no significant main effect for either of the groups, or any significant group interactions, indicating that initial processing was similar among both participant groups. For the first-pass reading time, results for the groups showed that within a factual context first-pass reading times were significantly longer for inconsistent rather than consistent, critical words however, within a counterfactual context the pattern was reversed, with a trend for longer reading times of consistent, compared to inconsistent, critical words. In the post-critical region, a significant effect of consistency revealed that first-pass reading times were longer following a consistent, rather than an inconsistent critical word. For the first-pass regressions out, results showed that in the post-critical region, a significant effect of consistency indicated that participants were more likely to make regressive eye movements when the critical word was inconsistent with the context, compared to when it was consistent. For the regression path reading time, a significant effect of consistency in the post-critical region showed that participants took longer to move past this region when the critical word was inconsistent, compared to when it was consistent with the context. Finally, analysis of total reading times revealed a significant effect of consistency in the critical region, reflecting longer reading times when the sentence contained an inconsistent critical word, compared to when it contained a consistent critical word. A significant effect of group was found on the critical region and showed that participants with ASD spent longer overall reading this critical region than TD participants did.

Results of the studies on word learning skills in ASD

Several studies, that used eye tracking technology, in the course of the process of learning new words by a population with ASD, provided significant findings with regard to gaze behavior during the presentation of a new target word, the preference for specific parts of speech (e.g. verbs), as well as the ability to use existing syntactic knowledge to learn new words.

In their research study, Aldaqre et al. (2015) compared the performance of adults, with and without autism, in a word learning task by recording their gaze behavior using an eye tracker. The sample consisted of 15 high-functioning adults with ASD, aged 19–61 years (mean age: 36.9 years) and 15 typically developing adults, aged 20–53 years (mean age: 32.5 years). Adults with ASD were diagnosed by a qualified clinical psychologist or psychiatrist, and they met the International Classification of Diseases 10th Revision (ICD-10) criteria for Asperger syndrome, autistic disorder or childhood autism. All participants were measured in non-verbal and verbal intelligence tests before participating in the survey. The stimuli included short animated films, where participants were taught a new target word each time. Three conditions were presented to each participant: one familiarization and two test conditions, each of which was presented twice. During the familiarization phase, four items (e.g. apple, car, fish, and boat) were presented to the participants, who were then asked to focus only on one of them. Next, the objects were displayed in a different order and participants were asked to select the target object card referenced in the familiarization phase and give it to the researcher. Each test condition was divided into two trials: learning trials and response trials. In the learning trial, 2 new objects appeared, while the presenter looked at one, named it, and asked the participants to look at the target object. The response trial was similar to the familiarization trial, except that the objects were the two previously presented objects during the learning trial and two additional distractors. Data were analyzed from three areas of interest: one for each of the two objects and one for the instructor’s face. The average relative probability of looking at each area of interest during the time in which the instructor looked at and labeled the target object was calculated for each participant and each condition, separately. Relative looking time at the objects was used as an implicit measure of word learning. This was calculated by dividing the looking time spent on each of the areas of interest by the total looking time on all areas of interest, and then averaged across the two repetitions of each condition for every participant. Results showed that both groups mostly chose the correct referent, but less so the ASD group, suggesting that, while the ability to learn novel words by referring to gaze does develop in ASD adults, it does not do so, as fully as it does in their typically developing peers.

Horvath (2018) investigated whether preschool children with ASD could begin to learn new words by watching the linguistic environment in which they are presented, even in the absence of visual or social context. Participants were 32 children with ASD, aged 2–4 years. Participants were measured as regards visual reception tests, receptive language and expressive language tests before participation and there were no group differences in the average score in any of the three measures. Each experimental trial consisted of a familiarization phase, a test phase, and a generalization phase. During familiarization, children were randomly assigned to listen to a novel verb in either transitive or intransitive sentences while, an unrelated silent visual animation of shapes and lines was played on the eye-tracker monitor. The test phase consisted of baseline, prompt, and response subphases: during the baseline subphase, children viewed two alternative referent events for the novel verb; each event was depicted in a separate scene, one on each side of the monitor and children were asked to find the referent scene for the novel verb (e.g. “Where’s biffing?”). In the prompt subphase, the two scenes disappeared, were replaced by a fixation star in the middle of the screen and children were asked to find the referent scene for the novel verb. In the response subphase, both scenes were played anew simultaneously and an additional prompt was heard after 4 s had elapsed (e.g. “Find biffing!”). The generalization phase immediately followed the test phase and was structured identically to the test phase, except for the fact that the test scenes differed slightly (different actors, different backgrounds). The study examined the shift of the participants' eyes in the two scenes that were presented. Gaze behavior during the first 3.5 s of the response subphase was depicted as a proportion of looking at the causative scene with all looks included in the denominator of the proportion. It was found that during the first time window, children shifted their eyes from a central fixation star to the dynamic scenes, whereas, in the second time window, children looked primarily at the dynamic scenes. Consequently, the results indicated that children with ASD were able to use the syntactic mechanism, as they could extract accommodation information from an audio stream, quickly mapping meaning into a word, and using existing syntactic knowledge to learn a new verb. In addition, it was revealed that children with ASD preferred to look at the causal scene associated with transitive verbs rather than intransitive verbs.

Subsequently, Venker (2019), investigated cross-situational word learning in children with ASD and TD children, matched as regards vocabulary knowledge. Specifically, the participant sample included 18 children with ASD (aged 4-7 years) and 21 typically developing children (aged 2-7 years). The TD and ASD groups were matched based on tests on vocabulary knowledge and nonverbal cognitive abilities. The children participated in two experimental tasks: a cross-situational word-learning task and an ostensive word-learning task. The cross-situational task consisted of three phases: familiarisation, teaching, and test. Familiarization trials were included to familiarize participants with the task and involved 2 familiar objects and 2 familiar labels; teaching trials included 2 unfamiliar objects and 2 novel labels, whereas test trials referred to either a familiar word or a novel word. In familiar test trials, children saw two of the familiar objects introduced during the familiarization phase and were asked about one of them. In novel test trials, children saw two of the novel objects that had been taught during the teaching phase and were asked about one of them. The ostensive word-learning task similarly consisted of a familiarization, teaching, and test phase. This task mirrored the cross situational learning task with one crucial difference: label-object links were presented explicitly. Familiarization trials included 1 familiar object and 1 familiar label, whereas teaching trials included 1 unfamiliar object and 1 novel label. The design of the test phase was identical to that of the cross-situational task. Data processing was done by creating areas of interest for both tasks. They examined the proportion of data contributed by each group during the analysis window of the test trials across both tasks. The results showed that children with ASD showed cross-situational learning abilities on par with TD children who had similar levels of vocabulary knowledge. In the ostensive teaching phase, children in the TD group looked at the pictures for the same period of time as children in the ASD group did. Furthermore, results showed that young children with ASD could use a type of word learning-cross-situational learning-, to determine word meanings without relying on social cues.

Discussion

The purpose of this review was to present the research that has been conducted over the last five years and used the eye tracking method to examine three areas of research; (a) language development (b) reading skills and (c) word learning in the population with ASD as compared to typically developing population. Results of this review present significant information on language processing by people with ASD.

Language development

Specifically, studies in the domain of language development present significant results for a wide range ages. Results of Plesa-Skwerer et al. (2016) contributes in the literature as the researchers examined older children and found that the severity of ASD was related to efficiency in lexical processing, which has implications for developing the ability to integrate auditory and contextual information such as Bavin et al. (2014) who conducted a similar study in young children.

The study of Hahn et al. (2015) similarly to other researchers (Brock et al. 2008, Norbury 2005) found that young children with ASD process ambiguous words in a similar manner to matched controls, showing that they are able to use context to resolve linguistic ambiguity. These results contradict the predictions of Weak Central Coherence theory which claims that people with ASD present a domain-general difficulty as regards integrating information into a broader context (Happe and Frith 2006, Henderson et al. 2011, Joliffe and Baron-Cohen 1999). Researchers attributed this difference in performance variety based on task. Their paradigm provided an implicit measure of semantic activation with very low task demands, so individuals with weak structural language would probably perform poorly on this task. Furthermore, difference may be attributed to the choice of population as the researchers selected high verbal sample. Maybe individuals with weak structural language had lower scores.

The results of Bavin et al. (2016) showed that both groups initially accessed the two possible items that matched the description (that is, the target and competitor) before hearing a modifying phrase that restricted these two items to one, and they looked more at these two items than at the two distractors. That is, children in both groups, were able to process the auditory information in the noun phrase and link it to the two appropriate pictures. However, the children with ASD were slower in focusing on the two matching items. Contrary to that, in previous research using a contrastive context, 5–6-year-olds with TD has shown sensitivity to the presence of two possible referents before settling on the target (Nadig et al. 2003). In the same manner and with the same age range, the study of Zhou et al. (2019) investigates whether Mandarin-speaking preschool children with ASD are able to use the verb information predictively, that is to anticipate the upcoming linguistic input. They found that 5-year-olds with ASD are able to use the selectional information of verbs as effectively and rapidly as their TD peers to predict the upcoming linguistic input in contrast previous studies (Boucher 2012, Eigsti et al. 2007). They seem to suggest that the language comprehension abilities of children with ASD might be severely impaired. Furthermore, the 5-year-olds with ASD exhibited fewer looks to the target area compared to the TD 5-year-olds. This difference in the overall fixation proportions between the two groups is presumably due to their difference in the cognitive control of visual attention, as it has been previously documented that the cognitive control of visual attention is impaired in ASD (Happe et al. 2006, Sasson et al. 2008).

Righi et al. (2018) similar to a previous study of Patten et al. (2014) found that typically developing children showed a reliable preference for synchronous audio and video stimuli in comparison to audiovisual asynchronies in contrast to children with ASD who did not display any reliable preferences for the synchronous or asynchronous videos at any delay tested. These results provide further evidence for atypical processing of multimodal speech in children with ASD as it has been posited that impaired multimodal processing is a potential precursor to the language deficits observed in ASD (Stevenson et al. 2014).

Thompson et al. (2019) focused on assessing the attending behavior of children with ASD when presented with e-book content. Children with ASD did not focus much of their attention on the pictures and print embedded within the e-book, more specifically, children with ASD attended to the e-book approximately half of the time it was displayed. The study contributes to the literature on book reading with children with ASD by providing objective data regarding the extent to which a sample of children with ASD attend to e-books, with and without strategies to increase attendance to print. Results agree with the opinion that despite the potential advantages that e-books offer, evaluating the level of attention and interaction with e-books among children with ASD can be extremely difficult as they may not reliably report or otherwise demonstrate their understanding or interest (Mineo et al. 2009).

Reading

Most studies focusing on the reading ability of people with ASD included adult participants and presented a wealth of significant findings on the gaze behavior and performance of people with ASD compared or not to the typical population. Research on the performance of children with ASD also presents important findings.

Specifically, Yaneva et al. (2015) based on the opinion that children with autism are considered to have greater difficulties decoding vague representations, compared to typically developing individuals, due to their impaired ability to generalize, grasp context, or reason about the intention of the author (Hartley and Allen 2014), examined the differences between autistic and non-autistic adults with regard to the time they spent looking at pictures and text paragraphs. They founded that adults with ASD seemed to focus more on the pictures or symbols than on the text, in contrast to the neurotypical participants, for whom it mostly made no difference whether or not there were pictures accompanying the text. These results agreed with the studies which mentioned that people who have ASD experience various problems in reading texts, such as inability to understand context, process long and complex sentences and comprehend figurative language and abstract words (O’Connor and Klein 2004).

Au-Yeung et al. (2015), based on consistent with previous eye movement studies (Filik and Moxey 2010, Kaakinen et al. 2014) on ironic sentences, supported that TD participants produced longer total reading times for the critical sentences containing the utterance, and also longer total reading times for sentences that restated the contextual information that had preceded the utterance. These findings support the view that extra processing is involved when reading written irony, and there were evidence suggesting that this is due to the need to reject the more salient literal surface meaning of an utterance and infer its non-literal ironic meaning (Filik and Moxey 2010). In the study by Colich et al. (2012), in which comprehension rates were calculated separately for their ironic and sincere conditions, equivalent accuracy was found between ASD and TD groups for the two conditions. This is consistent with the current study, potentially suggesting that there could be a lack of processing difficulty associated specifically with irony.

Yaneva et al. (2016) investigated the ability of autistic adults to make bridging inferences by recording eye-tracking data while they were reading pairs of sentences, and revealed that there was a significant difference between the total fixation durations, number of fixations and number of regressions between the autistic and non-autistic participants as the one found for adolescent participants in the study of Sansosti et al. (2013).

Howard et al. (2017) examined on-line lexical processing in ASD by measuring participant’s eye movements as they read sentences that contained a target word and revealed that both TD and ASD readers showed normal word frequency effects for all target fixation measures. This finding of a normal frequency effect agreed with the current knowledge in relation to lexical processing in ASD, as it demonstrates that in addition to intact performance accuracy for isolated word identification tasks (Mayes and Calhoun 2006, Saldaña et al. 2009), the processes engaged in identifying a word during normal reading appear to be comparable between ASD and TD readers. This is in line with cognitive theories of ASD that suggest low-level bottom–up processing to be intact (Minshew and Goldstein 1998). Howard's et al. (2017) findings are inconsistent however with findings of the studies of Sansosti et al. 2013 and Yaneva et al. 2015. They reported ASD readers have longer average fixation durations. This inconsistency may be attributable to the differences between the stimuli employed or the age of participants.

In another study with adult participants, Štajner et al. (2017) collected parallel gaze data to study the differences in word processing between participants with autism and a control group of neurotypical participants in a natural reading task and found that even though both groups understood the texts at a similar level, participants with autism had significantly longer viewing times, more fixations and more revisits per word, indicative of heavier cognitive load. They explained that this result may suggest that the pattern of results observed in the ASD readers reflects a different, perhaps more cautious reading strategy rather than greater cognitive load associated with lexical processing. Finally, other than word length which is naturally highly correlated with viewing time, they mentioned that, similarly to Juhasz and Rayner (2003), the age of acquisition seems to be an important factor related to the viewing times of both groups, followed by frequency and familiarity.

In their study, Au-Yeung et al. (2018) were based on a study for typical readers by Barton and Sanford (1993) that demonstrated that the global goodness of fit of a word to a whole passage context influences anomaly detection in text reading, result that was confirmed in another study by Hannon and Daneman (2004). The study of Au-Yeung et al. (2018) showed that ASD participants spent significantly greater time re-reading all types of passages after their initial reading, in comparison to the TD group. This finding is consistent with the results from a previous study investigating irony comprehension in ASD (Au-Yeung et al. 2015) while the question of exactly why this effect emerges in ASD is clearly worthy of further empirical examination. However, results were in contrast with another study that investigated the time-course of TD individuals’ processing of text containing anomalous material (Daneman et al. 2007) revealing that detection of the anomaly was not immediate at the first encounter of the target word, but occurred later with more time spent looking back at the target words in the anomalous condition compared to the non-anomalous condition.

The study of Barzy et al. (2020), investigated how autistic adults process the emotional responses related to irony in real-time. Similar to other studies (Filik et al. 2017, Gibbs et al. 2002, Knickerbocker et al. 2015), they found that the type of criticism influenced reading of the critical word, with longer regression path reading times following literal than ironic criticism, indicating that readers found it easier to integrate an emotional response to the ironic condition. Furthermore, a study similar to other ones (Au-Yeung et al. 2015, Black et al. 2019, Ferguson et al. 2019, Howard et al. 2017, Sansosti et al. 2013) revealed group differences in overall reading time, with adults in the autistic group incurring longer regression path reading times and making more regressions out of the critical and pre-critical regions compared to the TD control group.

Ferguson et al. (2019) examined whether adults with ASD would show intact processing of impossible counterfactual events or whether difficulty would emerge among people with ASD when a more substantial change to reality was required. They found that participants detected inconsistent continuations in the early moments of processing. Both experiments they used provide strong evidence that language processing is grounded in knowledge about reality, and suggest that constraints from the real-world are prioritised for the online integration of text. Specifically, results showed similar replicated effects, evidenced in other studies, too (Ferguson and Sanford 2008, Warren et al. 2008), by showing that initial processing of a word is based on fit with real-world constraints, even when a prior fantasy context neutralised this real world violation. Further, they noted that the absence of a delay in inconsistency detection among our ASD participants supports earlier findings that have demonstrated intact global coherence and ability to integrate information during text comprehension in adults with ASD (Au-Yeung et al. 2018, Black et al. 2019, Howardet al. 2017).

The study of Davidson and Ellis Weismer (2017) examined ambiguous sentence comprehension in children (ages 8–14 years) with ASD and their typically developing peers. Both groups experienced interference when we assessed the comprehension product during the picture selection stage. This comparison of the comprehension products and processes aligns with the results of Henderson et al. (2011) who found that that their participants with ASD displayed intact processing of homonyms during the lexical access stage, but increased semantic interference during the selection stage of semantic processing. Results also showed that readers who spent more time reading the context appeared to do so consistently across items, and that there were “cautious readers” in both groups. This result is contrary to previous studies showing generally longer reading times in individuals with ASD relative to their peers (Au-Yeung et al. 2015, Howard et al. 2017, Sansosti et al. 2013). An explanation is that this study examined comprehension in school-age children while the other studies assessed comprehension processes in adults (Au-Yeung et al. 2015, Howard et al. 2017).

The study of Micai et al. (2017) aimed at exploring inference generation skills and reading strategies in a group of children and adolescents with ASD compared to a closely-matched group of control peers. Results showed that participants with ASD were as accurate as the control group in responding to questions present in the paragraph assigned both to the literal and inferential condition. The result that accuracy in the inferencing task was comparable across ASD and control participants is in agreement with some (Saldana and Frith 2007, Sansosti et al. 2013), but not all previous studies (, Loukusa et al. 2007, Norbury and Bishop 2002). The lack of differences in their study was due to the fact that the group of participants was composed of individuals with ASD, all comparable to the control group, both with respect to language skills and overall reading comprehension.

The same researchers (Micai et al. 2019) in another study aimed at exploring the ability of individuals with ASD to detect errors when asked to do so, and more specifically, if instructions would influence their ability to detect semantic and orthographic errors. In addition, it aimed at exploring if reading behavior changed with different instructions and error types. They did not find significant differences between the performance of the ASD and the control groups in semantic or orthographic error detection. Specifically, they found that the orthographic errors received longer first fixations, single fixations, and gaze duration, compared to semantic errors. Participants needed more time to inspect orthographic errors, compared to semantic errors, probably due to the visual nature of the orthographic errors. On the other hand, semantic errors were re-read for longer amount of time, received more leftward eye movements out of the error before leaving the error to the right (regressions-out) and leftward eye movements into the error having already left the error to the right (regressions-into) compared to the orthographic errors. These results are in line with previous research documenting that readers tend to make regressions back to earlier parts of the text (and longer fixations) during the processing of difficult texts containing anomalous words (Rayner et al. 2004). Also, the eye movements related to error detection indicated no overall significant difference in reading behavior between the ASD and control groups, result that is consistent with the results of Howard et al. (2017) and Micai et al. (2017), who found no differences on most eye-tracking measures.

The study of Fong et al. (2019) focused on the immediate relative effects of using coloured overlays on reading and ocular performances in preschool children with ASD and their TD counterparts. The results of this study show that there were no significant differences in the ocular performance and reading speed of the TD or ASD groups with or without coloured overlays. Given that the use of coloured overlays in preschool children with ASD has not previously been investigated, this study contributed significantly to the literature. The results are consistent with other findings that coloured overlays have no immediate effects on improving the reading speed and global reading ability of participants with reading difficulties (Ritchie et al. 2011).

Word learning

In the word learning section, surveys present results for both young children and adults. Specifically, Aldaqre et al. (2015) study contributes significantly to the literature as researchers found controversial results and used adults with ASD as a sample, whereas the previous studies mostly focused on ASD children. This study demonstrated that adults with autism are able to rely on gaze cues in word learning in contrast to previous studies which demonstrated that children with autism have difficulties in this section (Akechi et al. 2011, Preissler and Carey 2005).

In their study, Horvath et al. (2018) found that the children demonstrated an ability to use syntactic bootstrapping even though the linguistic information was presented in the absence of any social or visual cues to connect it to an event referent. Their results agree with three prior findings, first, that children with ASD can extract distributional information from an auditory stream (Eigsti and Mayo 2011, Obeid et al. 2016) second, that they can fast-map meaning to a word form segmented by statistical learning (Haebig et al. 2017); and third, that they can use syntactic bootstrapping to learn novel verbs (Naigles et al. 2011, Shulman and Guberman 2007).

The study of Venker (2019) provided the first evidence that preschool and earlyschool-aged children with ASD, like older children with ASD (McGregor et al. 2013), are capable of cross-situational learning. Specifically, the children with ASD in this study showed cross-situational learning abilities on par with TD children who, though younger, had similar levels of vocabulary knowledge. These results are significant because they confirm the availability of a type of word learning—cross-situational learning—that young children with ASD can use to determine word meanings without relying on social cues like McGregor et al. (2013) found in older children. Furthermore, these findings add to growing empirical evidence that although many children with ASD have delayed language development, they have access to word-learning abilities that are qualitatively similar to those seen in typical language development—including sensitivity to patterns in speech, use of social cues, and syntactic bootstrapping (Luyster and Lord 2009, Mayo and Eigsti 2012, McGregor et al. 2013).

Conslusions

The extended number of studies using the eye tracking method to both typical and ASD population demonstrate that it is an important means of investigating a variety of variables that have facilitated the understanding of language processing and its subfields specifically for the ASD population. As it can be documented from the research of many variables that are not only related to language skills but also to emotions, cognitive ability and game perception, this method can provide answers to as regards a complete profile with skills and deficits that is presented by the ASD population.

Disclosure statement

No potential conflict of interest was reported by the authors.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.