A large body of evidence suggests that the basis of word reading process is essentially linguistic (Bishop & Snowling, Reference Bishop and Snowling2004; Catts et al., Reference Catts, Adlof, Hogan and Weismer2005; Pennington & Lefly, Reference Pennington and Lefly2001; Snowling, Reference Snowling2008; Snowling et al., Reference Snowling, Gallagher and Frith2003, Reference Snowling, Nash, Gooch, Hayiou-Thomas, Hulme, Language and Team2019; Torppa et al., Reference Torppa, Lyytinen, Erskine, Eklund and Lyytinen2010). Phonological awareness (PA), the recognition that spoken words can be segmented into small units (Ziegler & Goswami, Reference Ziegler and Goswami2005), is one of the most studied linguistic domains in the context of reading. According to the phonological sensitivity approach (as phrased by Dickinson et al., Reference Dickinson, McCabe, Anastasopoulos, Peisner-Feinberg and Poe2003), phonological skills explain a large portion of the variance in reading achievements (Snowling, Reference Snowling1998). Other language skills were also associated with word reading. These skills include i) semantic knowledge: word meanings, vocabulary size, and word definition, ii) morphological awareness: the ability to understand, use, and manipulate the smallest meaningful units such as root words, prefixes, and suffixes, iii) syntactic awareness which comprises a set of rules that mediate between word structures (e.g., word order, the rigidity of this order in the specific language, and the types of words), and iv) pragmatics that determine the manner subjects use their language in a communicative context and discourse (Toppelberg & Shapiro, Reference Toppelberg and Shapiro2000). The contribution of these language skills to basic reading skills has been reported across different tasks and languages (e.g., in English: Scarborough, Reference Scarborough1990; in Finnish: Torppa et al., Reference Torppa, Lyytinen, Erskine, Eklund and Lyytinen2010; in Arabic: Abu-Rabia, Reference Abu-Rabia2007; El Akiki & Content, Reference El Akiki and Content2020; Schiff & Saiegh-Haddad, Reference Schiff and Saiegh-Haddad2018; Taha & Taha, Reference Taha and Taha2019; Tibi & Kirby, Reference Tibi and Kirby2017, Reference Tibi and Kirby2019; and in Hebrew: Schiff & Lotem, Reference Schiff and Lotem2011).
This dissociation between phonological and language skills in the reading research inspired the central aim of the current study. Accordingly, four linguistic profiles have been established among Arabic-speaking kindergarteners assessing their reading achievement 1 year later at first grade. Below is an introduction of the association between developmental language disorders (DLDs) and dyslexia, followed by two theoretical frameworks regarding the language-reading relation (the two-dimensional model and the comprehensive language approach [CLA]), and summing up the introduction by a brief review of the Arabic language and the research questions and hypotheses.
DLD and dyslexia
The association between low linguistic skills and reading difficulties has led many researchers to examine the reading achievements of children with DLD and the linguistic levels of children with dyslexia (Bishop & Snowling, Reference Bishop and Snowling2004; Catts et al., Reference Catts, Adlof, Hogan and Weismer2005). For example, previous studies showed that young adults (at ages 19–24 years) who were identified as DLD in childhood performed significantly worse than controls adults with no DLD background on learning measures, including reading (Snowling et al., Reference Snowling, Bishop and Stothard2000; Whitehouse et al., Reference Whitehouse, Line, Watt and Bishop2009; Young et al., Reference Young, Beitchman, Johnson, Douglas, Atkinson, Escobar and Wilson2002), spelling, and calculation (Young et al., Reference Young, Beitchman, Johnson, Douglas, Atkinson, Escobar and Wilson2002). According to Young et al (Reference Young, Beitchman, Johnson, Douglas, Atkinson, Escobar and Wilson2002), 36.8% of the DLD group (diagnosed at the age of 5 years) met the criterion for reading disability in young adulthood. Higher percentages also met other learning disabilities’ criteria (spelling and arithmetic). In another study, out of 102 English-speaking children aged 5–9 years diagnosed with language disorders, 51% were also reading-impaired, and out of 110 reading-impaired children, 55% also exhibited oral language impairment (McArthur et al., Reference McArthur, Hogben, Edwards, Heath and Mengler2000). Furthermore, English-speaking children with DLD showed lower phonological scores compared to children at family risk of dyslexia, who in turn gained lower scores than typically developing children when followed longitudinally at ages 3–4 years and 4–5 years (Nash et al., Reference Nash, Hulme, Gooch and Snowling2013). However, only the children with DLD showed significantly lower scores in non-phonological broader linguistic skills compared to children with typical development and those at family risk of dyslexia.
A similar picture emerges when looking at the relation between reading and language disorders from the perspective of reading. Almost one-third of the children at family risk of dyslexia were also diagnosed as DLD at the age of 3.5 years (Nash et al., Reference Nash, Hulme, Gooch and Snowling2013). In the same vein, an earlier study had shown that English-speaking children (from dyslexic families) with reading impairment in second grade had significantly exhibited low syntactic scores at the age of 30 months beyond phonological and lexical measures (Scarborough, Reference Scarborough1990). Looking at a broader age range (3:09 years, 6 years, and 8 years), 60% of children from a high family risk of dyslexia whose reading difficulties were verified displayed deficits in grammatical skills and vocabulary beyond their phonological deficits (Snowling et al., Reference Snowling, Gallagher and Frith2003). The relatively good reading skills of the remaining 40% of the children with a high family risk of dyslexia were explained by earlier preserved vocabulary and expressive language. The relatively good oral language skills in the early ages were considered as a semantic compensation and protective factor for reading (Snowling, Reference Snowling2008). In a retrospective study on Finnish, an orthographically highly transparent language, the analysis revealed significantly poor performance in receptive vocabulary and sentence length at the age of 2–2.5 years among reading disabled children in second grade compared to typical readers (Torppa et al., Reference Torppa, Lyytinen, Erskine, Eklund and Lyytinen2010). Reading-disabled children also showed poor performance in inflectional morphology, picture naming, phonological sensitivity, rapid naming (objects), and letter naming at the age of 3.5 years. The linguistic gap between reading disabled and typical readers was also found at the age of 5–5.5 years. Path analysis showed that inflectional morphology, and phonological processing, with letter naming and rapid naming, were the direct early predictors for later reading accuracy and fluency, whereas expressive and receptive language were indirectly related to reading.
Theoretical frameworks
Different models were proposed to describe and explain the relationship between language and reading (Bishop & Snowling, Reference Bishop and Snowling2004; Catts et al., Reference Catts, Adlof, Hogan and Weismer2005; Kamhi & Catts, Reference Kamhi and Catts1986). The Two-Dimensional Model (Bishop & Snowling, Reference Bishop and Snowling2004) proposes that DLD and dyslexia are distinct disorders but share close behavioral similarities in phonological processing and word reading. Given the phonological shared deficit in DLD as well as reading deficits, children with low language skills were found to display a high risk for reading difficulties (McArthur et al., Reference McArthur, Hogben, Edwards, Heath and Mengler2000; Nash et al., Reference Nash, Hulme, Gooch and Snowling2013; Scarborough, Reference Scarborough1990; Snowling et al., Reference Snowling, Bishop and Stothard2000; Snowling et al., Reference Snowling, Gallagher and Frith2003). According to this model, the distinction between these disorders relates to the centrality of language difficulties in DLD compared to dyslexia. This model proposed four linguistic profiles and predictions to explain the variance in reading: i) children with typical phonological and non-phonological skills are expected to show no impairment in reading skills; ii) children with low phonological skills (and typical non-phonological skills) are expected to manifest dyslexia; iii) children with low phonological and non-phonological skills are expected to establish the DLD group; and iv) children with low non-phonological skills (and typical phonological skills) are expected to show poor comprehension skills with intact decoding skills. The profiles and hypotheses derived from this model, which mainly attribute reading difficulties of children with dyslexia and DLD to low phonological processing skills, were largely examined in English (Catts et al., Reference Catts, Adlof, Hogan and Weismer2005; Nash et al., Reference Nash, Hulme, Gooch and Snowling2013; Nation et al., Reference Nation, Cocksey, Taylor and Bishop2010) relative to other languages, for example, Dutch (De Groot et al., Reference De Groot, den Bos, der Meulen and Minnaert2015) and Greek (Talli et al., Reference Talli, Sprenger-Charolles and Stavrakaki2016).
The other view, the CLA, posits that the various linguistic (phonological and non-phonological) skills develop concomitantly and with relation to each other. Their overall interactions explain the significant variance in literacy and reading (Dickinson et al., Reference Dickinson, McCabe, Anastasopoulos, Peisner-Feinberg and Poe2003; Dickinson & McCabe, Reference Dickinson and McCabe2001). The results of different studies coincide with the premises of this approach showing a significant role for language skills in literacy and reading acquisition (Abu-Rabia, Reference Abu-Rabia2007; Asadi et al., Reference Asadi, Khateb and Shany2016; Hansen, Reference Hansen, Saiegh-Haddad and Joshi2014; McKague et al., Reference McKague, Pratt and Johnston2001; Mokhtari & Thompson, Reference Mokhtari and Thompson2006; Nagy et al., Reference Nagy, Berninger and Abbott2006; Ramus, Reference Ramus2003; Saiegh-Haddad & Geva, Reference Saiegh-Haddad and Geva2008; Snowling et al., Reference Snowling, Hayiou-Thomas, Nash and Hulme2020).
To the best of our knowledge, no study in the Arabic clinical and educational contexts has investigated the reading performance of children with low linguistic skills based on the differentiation between phonological and language skills in kindergarten. This investigation would enhance understanding of the relations between different linguistic skills and reading and provide practical implications for early reading intervention. The motivation for the present study also derived from the linguistic and orthographic idiosyncrasies of Arabic as well as its diglossic nature. These particular features emphasize the need to examine in Arabic the reading and language-based theoretical frameworks that were originally derived from English data (Share, Reference Share2008).
The Arabic language
Arabic is the sixth most spoken language in the world with nearly 300 million speakers (Eberhard et al., Reference Eberhard, Simons and Fennig2019). It is written from right to left in an abjad or consonantal writing system (Daniels, Reference Daniels1992, Reference Daniels2018) and consists of two sets of graphic signs: horizontally arrayed letters and vertically arrayed extra-linear diacritic-like signs. This writing system consists of 29 letters. Twenty-eight of the 29 letters denote consonants (except /ʔalif/). Three letters /ʔalif/, /ya:ʔ/, and /wa:w/, function as matres lectionis and represent both consonants and long vowels. As vowels, they are used to represent the three Arabic long vowels /i:/, /u:/, /a:/ (Saiegh-Haddad, Reference Saiegh-Haddad2013). In addition, extra-lineal diacritic-like signs are used extensively in Arabic and appear primarily above but at times below the letters. There are two classes of diacritizations: phonemic and morpho-syntactic (Saiegh-Haddad, Reference Saiegh-Haddad2018). The phonemic diacritization system consists of five major signs, three of which consistently map the three short vowels /a, u, i/ (respectively 
), one that denotes null vowelization (
), and one that denotes consonant germination/lengthening (
). In contrast, the morpho-syntactic diacritics consist of the three short vowels that can also appear word-finally along with other three extra-lineal signs, called nunation /tanwi:n/ (see, for details, Saiegh-Haddad & Henkin-Roitfarb, Reference Saiegh-Haddad, Henkin-Roitfarb, Saiegh-Haddad and Joshi2014).
Arabic writing system shares another essential feature which employs two versions of the same orthography differing in the amount of phonological information they supply: the phonologically transparent version is diacriticized (short vowelized) and mainly used in printed materials in the initial years of learning to read, generally up to the fourth grade, as well as in poetic texts and the Holy Scriptures (Saiegh-Haddad & Henkin-Roitfarb, Reference Saiegh-Haddad, Henkin-Roitfarb, Saiegh-Haddad and Joshi2014). All necessary phonological information is available in this script, and readers rely more on grapheme-to-phoneme conversion rules. The second version of the orthography, the default for Arabic readers, is the non-diacriticized (non-short vowelized) script, which relies on letters with no diacritic-like signs (for more details, refer to Bar-On et al., Reference Bar-On, Shalhoub-Awwad and Ibraheem2018). This non-diacriticized script illustrates two significant challenges: identifying the word encoded in the written string and resolving the homographic word. Hence, children can rely on the internal morphological structure of the words (Abu-Rabia, Reference Abu-Rabia2007; Saiegh-Haddad, Reference Saiegh-Haddad2013, Reference Saiegh-Haddad2018; Saiegh-Haddad & Geva, Reference Saiegh-Haddad and Geva2008), context cues, and prior knowledge (Abu-Rabia, Reference Abu-Rabia1996), utilizing parsing preference of familiar syntactic structures, sentence context (Hermena et al., Reference Hermena, Drieghe, Hellmuth and Liversedge2015), and frequencies of the word forms (Grosvald et al., Reference Grosvald, Al-Alami, Idrissi, Stockwell, O’Leary, XU and Zhou2019) based on prior reading experience or probability for ambiguity (the number of words with the same orthographic form).
Reading in Arabic is also influenced by the diglossia (Ferguson, Reference Ferguson1959) defined by the existence of two varieties of Arabic used under different conditions. While Standard Arabic (StA) is used in formal contexts and for reading and writing, Spoken Arabic (SpA) is used in everyday life and acquired naturally. These varieties show differences at all linguistic levels: phonology, lexicon, morphology, and syntax. The differences in phonological representations across the two varieties are thought to impact initial reading negatively (Saiegh-Haddad, Reference Saiegh-Haddad2003). This is because not all phonological units are explicitly accessible before reading acquisition starts. The absence of some literary phonological representations from the spoken lexicon [e.g., the StA phoneme /ث/ (θ) is often pronounced as /ت/ (t) in the SpA] illustrates the language availability problem when children are required to use the standard phonological representations in reading (Saiegh-Haddad & Henkin-Roitfarb, Reference Saiegh-Haddad, Henkin-Roitfarb, Saiegh-Haddad and Joshi2014). Another phonological distance relates to the architecture of the syllable. Saiegh-Haddad and Spolsky (Reference Saiegh-Haddad, Spolsky, Saiegh-Haddad and Joshi2014) analyzed the spoken corpus of 5-year-old children who speak the central Palestinian dialect and the lexical basis of first- and second-grade textbooks (which represents the StA corpus). Results established that the predominant SpA syllable structures were CVC (51.8%), followed by CCVC (26.8%). However, in the StA, the most common syllable structures were CVCC (46%) and CVC (42%). This phonological distance complicates the development of PA and likely poses challenges for learning to read in Arabic. This complex effect of the diglossic constraint on reading even when the transparent script is used may enhance readers to rely on morphology while reading words. The nonlinear derivational morphological system with a majority of words comprised of a consonantal root (that provides the core meaning) and a word pattern (a fixed prosodic template that specifies the word’s categorical meaning and some of the phonological characteristics of the surface form: vocalic, syllabic, and prosodic form) (Saiegh-Haddad, Reference Saiegh-Haddad2018) tunes readers from the initial phases of reading to extract morphological cues to access meaning. The role of morphological awareness in reading or spelling in Semitic languages (Arabic and Hebrew) was widely documented (Abu-Rabia, Reference Abu-Rabia2007; Mahfoudhi et al., Reference Mahfoudhi, Elbeheri, Al-Rashidi and Everatt2010; Saiegh-Haddad, Reference Saiegh-Haddad2013; Saiegh-Haddad & Geva, Reference Saiegh-Haddad and Geva2008; Schiff & Ravid, Reference Schiff and Ravid2007; Taha & Saiegh-Haddad, Reference Taha and Saiegh-Haddad2016; Taha & Taha, Reference Taha and Taha2019). The contribution of morphological awareness in Arabic was found to be significant for both reading accuracy and fluency in first and fourth grades (Asadi et al., Reference Asadi, Khateb, Ibrahim and Taha2017), and also for reading accuracy of both real and pseudowords in third grade (Tibi & Kirby, Reference Tibi and Kirby2017). In later stages (fourth and sixth grades), morphological awareness was a significant predictor for reading comprehension but not for word reading (Layes et al., Reference Layes, Lalonde and Rebaï2017).
Besides diglossia, Arabic orthography introduces a set of additional challenges. The presence of short vowels above or below the letters, and coping with the specific visual-orthographic features of this writing system together with diverse writing rules, all add in perceptual load and slow down word processing (Eviatar & Ibrahim, Reference Eviatar and Ibrahim2014; Hansen, Reference Hansen, Saiegh-Haddad and Joshi2014; Ibrahim et al., Reference Ibrahim, Eviatar and Aharon-Peretz2002). For instance, letters’ connectedness leads to visual changes of the basic form of the majority of letters according to their position and orthographic neighbors. This ligaturing has been shown to developmentally affect word recognition during the initial stages of reading acquisition, with non-connected words being faster and more accurate to read than connected ones (Khateb et al., Reference Khateb, Khateb-Abdelgani, Taha and Ibrahim2014). Later, when readers become more proficient (more familiar with the connected forms, which statistically become considered more frequent), reading connected words becomes faster and more accurate than non-connected words (Khateb et al., Reference Khateb, Taha, Elias and Ibrahim2013). Following these features, it is not surprising that low levels of decoding accuracy in the first elementary grades were reported when reading the transparent script (Abu-Ahmad et al., Reference Abu-Ahmad, Ibrahim, Share, Saiegh-Haddad and Joshi2014; Hende, Reference Hende2012).
It follows from the above literature that despite using the so-defined transparent orthographic script (with consistent spelling-to-sound mappings), reading acquisition in Arabic is a challenging process from the initial phases of elementary school. It implies that other linguistic factors beyond PA may be involved in reading. While recent research has shown that PA continues to play a consistent and important role in decoding the Arabic script (Abu-Rabia et al., Reference Abu-Rabia, Share and Mansour2003; Abu-Ahmad et al., Reference Abu-Ahmad, Ibrahim, Share, Saiegh-Haddad and Joshi2014; Asaad & Eviatar, Reference Asaad and Eviatar2014; Mannai & Everatt, Reference Mannai and Everatt2005; Saiegh-Haddad & Geva, Reference Saiegh-Haddad and Geva2008; Saiegh-Haddad & Taha, Reference Saiegh-Haddad and Taha2017; Taibah & Haynes, Reference Taibah and Haynes2011) even in the highest elementary school (sixth) grade (Asadi et al., Reference Asadi, Khateb and Shany2016), the contribution of language skills for the initial reading in Arabic were less studied. Although morphological awareness was considered to be an important factor in reading as it was mentioned above, the contribution of vocabulary to decoding and word reading was not significant in first grade (Asadi et al., Reference Asadi, Khateb, Ibrahim and Taha2017) and third grade where word, nonword, and paragraph reading were measured (Batnini & Uno, Reference Batnini and Uno2015). Moreover, while syntactic awareness has explained 11% of the variance in Arabic word reading in second grade (Abu-Ahmad et al., Reference Abu-Ahmad, Ibrahim, Share, Saiegh-Haddad and Joshi2014), according to another study, its contribution was not significant to reading measures across the elementary grades (Asadi et al., Reference Asadi, Khateb, Ibrahim and Taha2017).
The current study
The goal of the present study is to shed new light on the contribution of phonological and language skills to the reading process in Arabic and to examine the compatibility of the theoretical models, the two-dimensional model, and the CLA, derived from English data to the case of Arabic. The study also aimed at estimating the correlation between the linguistic profiles and the proportion of reading difficulties. For these purposes, four distinct profiles of children were established based on different linguistic tasks in kindergarten: i) typical phonology and typical language, ii) low phonology, iii) low language (typical phonology), iv) low phonology and low language. All profiles were followed 1 year later at first grade to assess their reading achievements. Selecting this developmental stage using a longitudinal design allows following children with different linguistic profiles before the start of formal reading instruction (kindergarten) through establishing critical milestones in reading acquisition before reaching complex reading levels for comprehension (first grade). To approach the contribution of phonological and language skills to reading in Arabic, the following questions and hypotheses were specified:
-
1. Do children with low phonological skills, children with low language skills (intact phonological skills), and children with a double deficit (in kindergarten) differ significantly from typical children across decoding and word reading tasks (measured in first grade)?
Lower reading levels among two groups of children: those with low phonological skills and those with the double deficit, compared to children with typical language development and those with low language skills, will support the fundamental relationship between phonological skills and reading as hypothesized by the two-dimensional model. Yet, lower reading levels among three groups of children: those with low phonological skills, those with the double deficit, and those with low language skills, compared to children with typical language development, will accord with the CLA.
-
2. Do children with low phonological skills show significantly lower decoding and word reading performance than children with low language skills?
A positive answer will support the unique and central role of PA in reading. A negative answer, that is, no significant differences in reading across these profiles, will indicate that both phonological and language skills are crucial for reading, corroborating the CLA hypothesis.
-
3. Do children with the double deficit show significantly lower decoding and word reading performance than children with low phonological skills or low language skills?
Significantly lower reading levels among children with double deficit, compared only to children with low language skills, but not to children with low phonological skills will add support to the critical and unique role of phonological skills for reading. However, significantly lower reading levels among children in the double-deficit group, compared to children with low language skills and children with low phonological skills, will provide a support to an additive effect of language.
-
4. From a prospective view what is the prevalence of reading difficulties (measured in first grade) across different linguistic profiles (measured in kindergarten)?
Method
Participants and procedure
A total of 1158 children from 73 kindergartens from the north district of Israel participated in this longitudinal two-phase study (kindergarten and first grade). Participants were all monolingual Arabic-speaking children. Of these, 561 were boys and 591 were girls; for the remaining six participants, gender value was not documented (M age in months = 68.71; SD = 3.4 in kindergarten). From this sample, four groups differing in linguistic proficiency were constituted in kindergarten and followed in first grade. Inclusion criteria for group selection are detailed below in the procedure section. The kindergartens included in the study were ranked low-middle socioeconomic status according to a welfare index ascribed to each kindergarten. The children differed in their Arabic-spoken dialects, roughly classified as rural, urban, and Bedouin. The similar dialects of the research assistants were considered in the assigning process to match their dialects to the dialects of the children as much as possible. This study was approved by the chief scientist of the Ministry of Education (file no’ 9667) and the Research Ethics Committee of the Faculty of Education at the University of Haifa (approval no’ 043/18). Parents provided informed written consent for their child to participate.
All children carried out different linguistic tasks in kindergarten (T1). Of the total sample, 956 children were followed in first grade (T2) to assess their reading skills. The linguistic tasks in T1 were individually administered in a quiet room in the children’s kindergarten in four separate sessions through January–May 2019. The reading tasks were carried out in two separate sessions through January–March 2020 in a quiet room in the children’s school. In both phases, the order of the tasks within each session was counterbalanced across participants, but the order of the items per task was kept intact. The tasks were administered by research assistants who were graduate students from learning disabilities departments or holders of other relevant academic degrees. In both phases, all the research assistants participated in training sessions of several days to ensure they understood the instructions and administration procedures. The missing data for children in T2 (N=202) constitute a dropout percentage equal to 17.44%. These children did not participate in T2 due to the spread of the coronavirus pandemic and the total shutdown of the education system in Israel in March 2020.
Measures
All the linguistic measures reported below were carried out in the spoken Arabic variety. Intraclass correlations reported below were based on an external sample (N = 40) of children from different kindergartens representing various spoken dialects. Examples of the task’s items shown in supplementary materials (Appendix 1–14) can be retrieved by the following OSF’s link: https://osf.io/fz6j4/.
Kindergarten’s phonological measures
-
CV isolation. Two subtests of this task were developed for the purpose of the study (Jabbour-Danial et al., Reference Jabbour-Danial, Abu-Ahmad, Mansour-Adwan, Joubran-Awwadia, Yassin and Shalhoub-Awwad2018). One subtest included CV (consonant-vowel) isolation from two-syllable words composed of two structures: /CV.CVC/ and /CVC.CVC/. The children were asked to repeat the target word and then to isolate the initial syllabic or sub-syllabic unit (CV): [i.e., /qu:l dulfi:n/ (‘say dolphin’), /bkilmet dulfi:n mnismaʕ bilʔawwal/ (‘in the word dolphin we hear at first’) _____ (correct responses: du/d/demi phoneme ʔed)/]. One example and four training items providing feedback were presented before the task started. This subtest included 12 items (maximum score = 12), intraclass correlation = .80, and Cronbach’s alpha reliability = .79. The second subtest included CV isolation from one-syllable words of CVC structure. The same procedure was administered (maximum score = 10), intraclass correlation = .91, and Cronbach’s alpha reliability = .80. A composite score of these two versions was calculated (r = .54, p < .001) with total alpha reliability = .81. Examples of the task’s items are shown in Appendix 1.
-
Final sound isolation. This task was developed for the purpose of the study (Abu-Ahmad et al., Reference Abu-Ahmad, Jabbour-Danial, Mansour-Adwan, Joubran-Awwadia and Shalhoub-Awwad2018a) and was administered twice at kindergarten. A composite score for both administrations was calculated, r = .56, p < .001. The children were asked to repeat the target word and then required to isolate the final consonant. All items were CVC words [i.e., /qu:l dob/ (say bear), /bkilmet dob mnismaʕ bilʔaxer/ (in the word bear we hear at the end) _____ (correct responses: b/demi phoneme ʔeb)]. One example and four training items providing feedback were presented before the task started (maximum score in each task = 12). Intraclass correlation = .94, .91 and Cronbach’s alpha reliability = .85, .83 for the first and second tests, respectively, and total alpha reliability = .87. Examples of the task’s items are shown in Appendix 2.
-
First sound isolation. Two versions of this task were developed for the purpose of the study (Abu-Ahmad et al., Reference Abu-Ahmad, Jabbour-Danial, Mansour-Adwan, Joubran-Awwadia, Yasin and Shalhoub-Awwad2018b), and a composite score was calculated, r = .48, p < .001. One task comprised words with CCVC syllabic structure, and the second task included words with CVC syllabic structure. The children were asked to repeat the target word and then required to isolate the initial consonant [i.e./qu:l mra:y/ (say mirror), /bkilmet mra:y mnismaʕ bilʔawwal/ (in the word mirror we hear at first) _____ (correct responses: m/demi phoneme ʔem)/]. One example and four training items providing feedback were presented before the task started. Each subtest included 12 items (maximum score = 12), intraclass correlation = .84, .93, and Cronbach’s alpha reliability = .76, .82 for the first and second subtests, respectively, and total alpha reliability = .84. Examples of the task’s items are shown in Appendix 3.
Kindergarten’s language measures
-
Vocabulary. To assess expressive vocabulary, a picture naming task was adapted to Arabic based on the TAVOR test (Tavor, Reference Tavor2008) and administered twice in kindergarten. A composite score of both trials was calculated, r = .41, p < .001. Children were shown pictures of objects (such as a bridge), actions (such as “to knock”) and adjectives (such as “angry”), and they were asked to name what they saw in each picture. Score 1 was given for each correct answer (maximum score = 11 and 12 for the first and second tests, respectively). Intraclass correlation = .91, .81 and Cronbach’s alpha reliability = .55, .69 for the first and second tests, respectively, and total alpha reliability = .73. Examples of the task’s items are shown in Appendix 4.
-
Noun-pluralization. In this task developed for the purpose of the study (Joubran-Awwadia et al., Reference Joubran-Awwadia, Mansour-Adwan and Shalhoub-Awwad2018a), children were required to provide orally the spoken plural form (feminine sound plural, masculine sound plural, and broken plural) or the dual form of 15 nouns. The examiner presented to the child a picture of a single item (e.g., /tuffa:ħa/ ‘apple’), and a picture of four items for the plural form (e.g., /ʔarbaʕ tuffa:ħa:t/ ‘four apples’) and says while pointing to each picture respectively: “Here, there is /tuffa:ħa waħdi/ ‘one apple’, and here there are four…” (expected answer: /tuffa:ħa:t/ ‘apples’). This task included three feminine sound plural items (noun+ suffix a:t, e.g., /tuffa:ħa:t/ ‘apples’), four broken plural items (e.g., /kya:s/ ‘bags’), three masculine sound plural items (noun+ suffix i:n, e.g., /mharʒi:n/ ‘clowns’) and five dual items (noun+ suffix e:n, e.g., /ʕasˁfu:re:n/ ‘two birds’) (maximum score = 15), intraclass correlation = .74., and Cronbach’s alpha reliability = .83. Examples of the task’s items are shown in Appendix 5.
-
Verb-derivation. In this task developed for the purpose of the study (Mansour-Adwan et al., Reference Mansour-Adwan, Joubran-Awwadia and Shalhoub-Awwad2018a), 12 sentences were presented orally to the child and s/he was required to complete the sentence with the correct derived form of the verb (e.g., /eddahha:n/ ‘The painter’.…; the expected answer is /bidhan/ ‘paints’). Two examples and two training sentences were presented before the task started (maximum score = 12), intraclass correlation = .73, and Cronbach’s alpha reliability = .70. Examples of the task’s items are shown in Appendix 6.
-
Sentence repetition. This task was developed for the purpose of the study to assess sensitivity to grammatical structures (Mansour-Adwan et al., Reference Mansour-Adwan, Cohen-Mimran and shalhoub-Awwad2018b). It was administered twice in kindergarten, and a composite score of both versions was calculated (r = .38, p < .001). Children were orally presented with different sentences at a rate of one word per second and were asked to repeat them verbatim, one sentence for each trial. In the first test, the sentences were presented orally by sound recording using E-prime software. Due to some technical problems, the research assistants orally presented sentences in the second test in real time. One practice item was given before the scored items were presented. The sentences consisted of five to eight words with various grammatical structures. The grammatical structures included compound sentences, a sentence with three successive verbs, relative clauses, sentences containing embedded clauses, and subordinating conjunctions; cause-relation clauses, conditional clauses, and direct speech. Responses were recorded, and a binary scoring system was implemented, that is, score one was given for each item if the child repeated all speech parts of the original sentence (consistent articulation errors were not considered for the scoring process). Any deviation from the original sentence rendered the entire sentence incorrect, with a score of 0. Dialectical articulation and phonological differences were not considered as errors (e.g., /ʕalaʃa:n/ instead of /ʕaʃa:n/ ‘because’). Research assistants were recommended to record the children’s productions and listen to their recordings for the scoring process (maximum score = 8 and 13 for the first and second test). Intraclass correlation = .90 for the first test and Cronbach’s alpha reliability = .72, .80 for the first and second tasks, respectively, and total alpha reliability = .83. Examples of the task’s items are shown in Appendix 7.
-
Receptive syntax. This task, based on the Clinical Evaluation of Language Fundamentals, CELF (Semel et al., Reference Semel, Wiig and Secord2000), was adapted to Arabic for the purpose of this study (Mansour-Adwan et al., Reference Mansour-Adwan, Cohen-Mimran and shalhoub-Awwad2018c). The children heard a sentence and were asked to point to the appropriate picture out of three possibilities. The target sentences were syntactically complex (conjunctions, relative clauses, adjectives, negative elements, and time clauses). This task included 12 items yielding thus a maximum score = 12, with intraclass correlation = .60 and Cronbach’s alpha reliability = .57. Examples of the task’s items are shown in Appendix 8.
-
Verb inflection using pseudo-root (ʃ.l.z). This task, originally developed by Shalev-Leifer et al. (Reference Shalev-Laifer, Reznik-Nevet and Share2016), was adapted to Arabic for the study (Joubran-Awwadia et al., Reference Joubran-Awwadia, Mansour-Adwan and Shalhoub-Awwad2018b). In this task, all pseudowords were verbs inflected from the pseudo-root (ʃ.l.z). This task includes 12 sentences which were orally presented to the child, and s/he was required to complete a parallel sentence with a morphological agreement for tense (past to present/present to past), number, gender, and person [e.g., /mba:reħ ʔana: ʃalazet (yesterday I ʃalazet), mba:reħ ʔintu:/ (yesterday you) _____ /ʃalaztu:/]. Maximum score = 12, Intraclass correlation = .70, and Cronbach’s alpha reliability = .83. Examples of the task’s items are shown in Appendix 9.
-
Derivation of resultative adjectives from verbs. In this task, also developed for the purpose of the study (Mansour-Adwan et al., Reference Mansour-Adwan, Joubran-Awwadia and Shalhoub-Awwad2018d), 12 sentences were presented orally along with a pair of pictures containing an event alongside an associative result. The child was required to complete the sentence with the correct derivation of the noun from the spoken verb, hence, the resultative adjective [i.e., /qassamu ttuffaħa/ (they cut the apple), /sˁaret ʔittuffaħa/ (the apple became) _____ /mqassame/maqsu:me/ (cut)]. One example and two training items were presented before the task started. Maximum score = 12, Intraclass correlation = .77, and Cronbach’s alpha reliability = .77. Examples of the task’s items are shown in Appendix 10.
First-grade reading fluency measures
In all reading tasks, time limit was fixed to maximum 3 min, and the scores were calculated by the number of correct items per minute. The values of Cronbach’s alpha reliability for reading tasks are presented in Table 2.
-
CV units. This task was developed for the purpose of this study (Abu-Ahmad et al., Reference Abu-Ahmad, Jabbour-Danial, Mansour-Adwan, Joubran-Awwadia, Mutlak-dahoud and Shalhoub-Awwad2019). It included 18 CV units, half of them composed of consonant and long vowel (i.e., /da:/) and the remaining items composed of consonant and short vowel (i.e., /ma/). Two long vowels were used in this task: /a:/ and /u:/. No items included the long vowel /i:/ because children were not yet formally exposed to this vowel by their formal Arabic instruction books during the testing period of the school year. The same consideration of familiarity was implemented for the short vowels including only the two short vowels /fatħa/ﹷ/ (for /a/) and /d ʕamma /ﹹ/ (for /u/) but not /kasra/
/ (for /i/). For letters, familiarity (based on the curriculum) and a frequency consideration were also taken into account (Boudelaa et al., Reference Boudelaa, Perea and Carreiras2020). The children were instructed to read aloud these units in a clear voice, as fast and accurate as possible, while paying attention to diacritics. The intraclass correlation = .81. Examples of the task’s items are shown in Appendix 11. -
Frequent words. This task, developed for this study (Jabbour-Danial et al., Reference Jabbour-Danial, Mansour-Adwan, Abu-Ahmad, Joubran-Awwadia, Mutlak-Dahoud and Shalhoub-Awwad2019), relied on four reading instruction books to determine the most common words and syllable structures. Based on the curriculum of the first instruction semester, 25 words (21 nouns, 4 verbs) were composed for this task: 5 words had CVC structure [e.g., /da:r/ (‘house’)], 4 words had CV.CV.CV structure [e.g., /rasama/ (‘drew’), 12 words had CV.CVC structure [e.g., /raza:n/ (‘Razan, a given name’)], and 4 words had CV.CV structure [e.g., /fa:di:/ (‘Fadi:, a given name’)]. The lexical status of the words: 76% identical, 16% cognates, and 8% unique for StA. Children were instructed to read aloud these words in a clear voice, as fast and accurate as possible, and to pay attention to diacritics. The intraclass correlation = .82. Examples of the task’s items are shown in Appendix 12.
-
Complex words. This task was developed for this study (Mansour-Adwan et al., Reference Mansour-Adwan, Jabbour-Danial, Joubran-Awwadia, Abu-Ahmad, Mutlak-Dahoud and Shalhoub-Awwad2019) to assess reading skills from a developmental point of view. It consisted of different words rated at different levels of syllabic and morphological complexity, fitting the curricular sequence of the first and second gradesFootnote 1 . The task included 50 words: 34 nouns and 16 verbs. These words consisted of different levels of syllabic structure complexity for nouns [e.g., CV.CVC /ʕinab/ (‘grapes’) and CVC.CV.CVC /burtuqa:l/ (‘orange’)] for simple and complex structures, respectively. This consideration was also drawn for verbs [e.g., CV.CV /ʒa:ʔa/ (‘came’) and CV.CV.CV /ʃariba/ (‘drank’)]. Furthermore, different morphological complexity levels were also considered for nouns [e.g., CV.CVC /xaru:f/ (‘sheep’) and CV.CVC /θi:ra:n/ (for the broken plural of the word ‘bulls’)] and for verbs [e.g., CV.CV.CV /ʃariba/ (‘drank’- pattern 1) and CV.CV.CV /sa:ʕada/ (‘helped’-pattern 3)]. The lexical status of the words: 56% cognates, 30% identical, and 14% unique for StA. Children were instructed to read aloud these words in a clear voice, as fast and accurate as possible, and to pay attention to diacritics. The intraclass correlation = .83. Examples of the task’s items are shown in Appendix 13.
-
Pseudowords. In this task (Joubran-Awwadia et al., Reference Joubran-Awwadia, Mansour-Adwan, Jabbour-Danial, Abu-Ahmad, Mutlak-Dahoud and Shalhoub-Awwad2019), all frequent words (derived from the second task) were modified in such a way that letters in words were reversed in their order or substituted with other letters to transform the real words into pseudowords. No change was made to the syllabic structure of the words. The same instructions as the previous tasks were presented to children but emphasized that these words have no meaning. The score was also calculated in the same manner. The intraclass correlation = .85. Examples of the task’s items are shown in Appendix 14.
/ (for /i/). For letters, familiarity (based on the curriculum) and a frequency consideration were also taken into account (Boudelaa et al.,
Descriptive statistics for the different measures collected in kindergarten from the general sample
Note. Derivation of RA = derivation of resultative adjectives from verbs.
∫.l.z=Verb inflection using pseudo-root.
Descriptive statistics for reading measures and nonverbal ability (T2) based on the general sample
Note. The means in all reading measures are represented by the number of correct items per minute.
First-grade cognitive measure
-
General nonverbal ability. A colored, shortened version of Raven’s Progressive Matrices (Raven et al., Reference Raven, Raven and Court1998) was administered to assess children’s nonverbal ability. Participants were asked to select the missing element of a presented pattern in 18 trials of increasing difficulty: six items were selected from each set (A, B, and AB). Maximum score = 18.
Data analysis and groups selection
The participants’ scores in all linguistic tasks administered in kindergarten (hereafter T1) were subjected to a factor analysis. All these tasks were loaded into two distinct factors: the first factor represented language skills and included vocabulary, noun-pluralization, verb derivation, derivation of resultative adjectives from verbs, receptive syntax, sentence repetition, and verb inflection using pseudo-roots with the following loading values: .73, .71, .68, .68, .66, .53, and .53, respectively. The second factor represented PA skills and included CV isolation, first sound isolation and final sound isolation tasks with the following loading values: .78, .74, and .68, respectively. Following this analysis, two composite scores were calculated separately for phonological and language skills and were utilized to constitute four linguistic profile groups in kindergarten based on standard cutoffsFootnote 2 used in the literature:
-
1. Group of typical linguistic skills and typical PA skills (hereafter TLTPh): This group comprised children whose composite scores for phonological and language skills were both in between the 35th and 65th percentilesFootnote 3 (N = 135).
-
2. Group of low PA skills (hereafter LPh). The children in this group scored below the 25th percentile in the PA composite score (N = 120) and above the 35th percentile in the language composite score (i.e., typical achievement in linguistic tasks).
-
3. Group of low linguistic skills (hereafter LL). Children in this group gained a composite score below the 25th percentile in the language domain (N = 111) and above the 35th percentile in the PA composite score (i.e., typical phonological skills).
-
4. Group of low linguistic and low phonological skills (here after LLLPh). This group comprised children whose composite scores in both domains were below the 25th percentile (N = 139).
Similarly, two reading profiles were constituted in the first grade after computing a composite score based on all four reading tasks: CV units, frequent words, complex words, and pseudowords (see Table 3 for correlations):
-
1. A group of typical readers (hereafter TR) included children whose reading composite score ranged between the 35th and 65th percentile.
-
2. A group with low reading skills (hereafter LR) comprised children who gained a composite reading score below the 25th percentile.
Correlations between phonological, language, and reading tasks
**p < .01. Phonological (1) and language (2) tasks refer to composite scores for measures collected in kindergarten. All reading measures (3 to 6) collected in first grade refer to correct words per 1 minute. a = kindergarten. b = first grade.
Statistical analysis
Descriptive statistics are separately reported for linguistic and reading measures. Correlations between the reading tasks at first grade and the phonological and language tasks measured at kindergarten were also computed. To examine the differences between different linguistic profiles in reading measures beyond nonverbal ability, a multivariate analysis of covariance (MANCOVA) was conducted. Furthermore, a cross-tabulation procedure was used to describe the relationship between the two categorical variables (linguistic and reading profiles).
Results
The descriptive statistics for the different PA and language measures that were collected in T1 (kindergarten) are presented in Table 1. The mean scores for the phonological tasks range from ∼29% to ∼71%, and the linguistic scores range from ∼49% to 82%. In the phonological domain, lower scores were obtained for the final sound isolation task, while the highest scores were obtained for the first sound isolation. In the language domain, lower scores were obtained for verb inflection using pseudo-roots, and the highest scores were obtained for verb derivation. Descriptive statistics for reading tasks are presented in Table 2. It shows that the lowest scores were obtained for reading complex words, while the highest scores were obtained for reading CV units. Table 3 shows the correlation between the reading measures collected in first grade and phonological and language composite scores computed from kindergarten measures. This analysis showed high correlations between the different reading measures.
Interestingly, and as could also be expected, the highest correlation was found between reading pseudowords and complex words, confirming that the higher the child’s decoding ability, the better his/her performance in reading complex words. We hypothesize that both tasks required a high degree of decipherability in this early reading stage. In addition, this analysis showed that, although both phonological and language composite scores positively correlated with reading tasks, correlations were found to be higher with the phonological (that included CV isolation, first, and final sound isolation) than with the language measures, attesting of the important well-established link between reading and phonology already during this very early stage of literacy acquisition.
Table 4 presents the descriptive statistics of the four reading measures as a function of the kindergarten linguistic profile (labeled as A to D). A MANCOVA analysis was conducted to examine whether the different linguistic profiles in kindergarten differed statistically in their reading fluency assessed by the different reading measures. This analysis showed a significant effect for group, F(12, 918.37) = 2.96, p < .001, Wilks’Λ = .90, Partial η2 = .03) indicating a statistically significant adjusted mean difference between the different linguistic profiles in all reading measures after controlling for nonverbal ability (assessed by RAVEN, used as a covariate). Pairwise comparisons using Bonferroni tests indicated that the LLLPh (group D) performed significantly lower than all the other groups (TLTPh, LPh, LL) in reading CV units and pseudowords. In reading frequent words and complex words, the LLLPh group exhibited significantly lower scores than TLTPh and LL groups, still marginally significant difference was observed in reading complex words (p = .06) between LLLPh (D) and LPh (B).
Descriptive statistics (means and standard deviations), one-way ANOVA on reading measures as a function of kindergarten profiles, and pairwise comparisons
Note. Reading measures (administered in first grade) refer to number of correct words in 1 minute. TLTPh = typical language and typical phonological skills. LPh = low phonological skills. LL = low language skills. LLLPh = low language and phonological skills. PWC = pairwise comparisons (Bonferroni-corrected).
*p < .05. **p < .01. ***p < .001.
In the next step, we examined the relationship and the statistical overlap between the linguistic profiles constituted in T1 and the two reading profiles (TR and LR) constituted in T2. Chi-Square testFootnote 4 reveals a significant relationship between linguistic profiles in kindergarten and reading profiles in first grade (χ2 (3) = 102.86, p < .001). To understand this relationship, we used cross-tabulation analysis (see Table 5), which indicated that while the majority of children with TLTPh (85.5%) and the minority of children with LLLPh (37%) had typical reading skills in first grade, only small group of children with TLTPh (14.5%) but the majority of children with LLLPh (63%) had low reading skills. Similar trends were observed for the LPh and LL profiles: one-third of the children in each group showed low reading skills, whereas two-thirds showed typical reading skills. Risk ratio analysis indicated that a child with LLLPh is 4.93 times more likely to be LR than a control child from the TLTPh group. A child with LL is 2.69 times more likely to be LR than a control child from the TLTPh group, and finally, a child with LPh is 1.81 times more likely to be LR than a control child from the TLTPh group. Overall, a higher prevalence of reading difficulties among the LLLPh group was found, almost twice the prevalence of reading difficulties among other groups (LL and LPh).
The cross-tabulation of linguistic profiles in kindergarten and reading profiles in first grade
Note. TR = typical reading skills. LR = low reading skills.
Discussion
This study investigated the prospective effects of early phonological and language skills on reading. More specifically, it aimed to reveal the reading achievement (assessed at first grade) of children assigned to different linguistic profiles (TLTPh, LPh, LL, and LLLPh) at kindergarten. The profiles’ constitution was based on the differentiation between phonological and language skills inspired by the two-dimensional model (Bishop & Snowling, Reference Bishop and Snowling2004). Exploring the prevalence of reading difficulties across these profiles was an additional study interest. The overall findings of this study are as follows: 1) the LLLPh was the only group to differ significantly from TLTPh group across reading tasks; 2) children with LPh did not significantly differ in reading achievements from children with LL despite their lower raw scores; 3) children with LLLPh gained significantly lower scores than children with LL in all reading tasks and compared to children with LPh in half of the tasks; and 4) prevalence estimates indicated that most children with LLLPh (63%) and about a third of the children with LL and LPh showed reading difficulties. These results bring new data in Arabic to bear on the role of language skills for decoding and word reading beside the widely accepted part of phonological skills. These findings will be discussed further below.
Differences in reading measures across linguistic profiles
In addition to the higher positive correlations between PA skills and reading measures supporting previous findings in Arabic (Asadi et al., Reference Asadi, Khateb, Ibrahim and Taha2017; Asadi & Khateb, Reference Asadi and Khateb2017; Tibi & Kirby, Reference Tibi and Kirby2018) and across languages (Ziegler et al., Reference Ziegler, Bertrand, Tóth, Csépe, Reis, Faísca, Saine, Lyytinen, Vaessen and Blomert2010), somewhat milder but still significant correlations were found between language composite measure and reading measures. These correlations support the profiles’ comparisons across reading tasks. It had been found that children with LLLPh obtained the lowest scores on all reading assignments. Of particular interest from a clinical perspective, statistically significant differences were found between LLLPh group and all other groups in two out of four reading tasks: CV units and pseudowords. The potential ability of the CV reading task to differentiate between the LLLPh group and all other groups highlights the underlying role this unit has in Semitic languages (Saiegh-Haddad, Reference Saiegh-Haddad2003; Share & Blum, Reference Share and Blum2005; Tadmor-Troyansky, Reference Tadmor-Troyansky2019) as well as many other languages such as Spanish and Italian (Goswami, Reference Goswami, Brunswick, McDougall and de Mornay Davies2010). The findings that only children with LLLPh differed significantly from typical children after controlling for nonverbal ability, and that the LPh group did not significantly differ from the TLTPh/LL groups in any reading tasks, emphasize that both phonological and other linguistic skills are critical for reading acquisition among Arabic-speaking children. Also, the finding that children with LLLPh gained significantly lower reading scores than children with LL (in all reading tasks) and children with LPh (in half of the tasks) imply that lower reading scores among children with LLLPh cannot be attributed exclusively to phonological skills. Alternatively, these results support multiple risk factors for reading difficulties, including both phonological and language skills.
The significantly low reading achievements of children with LLLPh imply that children with low scores in only one domain (e.g., PA or language) would not gain reading scores worse than the low linguistic combined profile. The contribution of the language skills even when the diacriticized script is used (and which could be precisely deciphered by grapheme-to-phoneme correspondence rules) supports the notion that the information that is extracted from the writing system involves complex relations of orthography, phonology, morphology, and meaning (Frost, Reference Frost2012), suggesting that “….the actual computation of an orthographic code in a given language is determined on-line by the transparency of mapping of graphemes into phonemes on the one hand, and by morphological and semantic considerations on the other hand, given the language properties in which reading occurs” (Frost, Reference Frost2012, p. 23).
The current results also support the hypotheses derived from the CLA, suggesting a critical role for various oral language skills in reading achievements (Dickinson et al., Reference Dickinson, McCabe, Anastasopoulos, Peisner-Feinberg and Poe2003; Dickinson & McCabe, Reference Dickinson and McCabe2001). In Arabic, despite the little support for the unique contribution of vocabulary to reading (Asadi et al., Reference Asadi, Khateb, Ibrahim and Taha2017; Batnini & Uno, Reference Batnini and Uno2015), a unique contribution for syntactic (Abu-Ahmad et al., Reference Abu-Ahmad, Ibrahim, Share, Saiegh-Haddad and Joshi2014) and morphological awareness skills for word reading across different ages was documented (Abu-Rabia, Reference Abu-Rabia2007; El Akiki & Content, Reference El Akiki and Content2020; Schiff & Saiegh-Haddad, Reference Schiff and Saiegh-Haddad2018; Taha & Taha, Reference Taha and Taha2019; Tibi & Kirby, Reference Tibi and Kirby2017, Reference Tibi and Kirby2019; Wattad & Abu Rabia, Reference Wattad and Abu Rabia2020). We agree that overemphasizing the phonological processing skills in many studies to the extent that other linguistic skills are underestimated is methodologically risky (Bishop, Reference Bishop1991; Bishop & Adams, Reference Bishop and Adams1990; Storch & Whitehurst, Reference Storch and Whitehurst2002). Other language skills are important for top-down strategies that likely to contribute to word recognition and orthographic learning (Share, Reference Share1995). The prominence of root consonants and word pattern morphemes in Arabic seems to enhance deciphering printed words at a very early stage of reading establishment (Bar-On et al., Reference Bar-On, Shalhoub-Awwad and Ibraheem2018; Shalhoub-Awwad & Leikin, Reference Shalhoub-Awwad and Leikin2016). Although the current study did not examine the effect of diglossia on the contribution of phonological and language skills to reading, it would be interesting to test in the future if the role of these linguistic skills differs across reading words with different lexical statuses (identical words, cognates, and unique StA words).
The prevalence estimates of reading difficulties across linguistic profiles
The results of the current study indicated that low reading level was observed in about two-thirds (63%) of the children with LLLPh and one-third with LPh and LL. From a practical perspective, these results highlight the importance of early identification of children with low linguistic abilities before they met academic failures in reading and writing experiences. Very recently, it was proposed that, because of the hidden nature of DLD that leads to under-identification of these children, an active involvement of speech language pathologists (SLPs) and collaboration between them and teachers in educational contexts would help to accurately flag children with potential language difficulties (McGregor, Reference McGregor2020). From a theoretical standpoint, the significantly low reading achievements of children with LLLPh (that may have resembled children with DLD) and the manifestation of low reading levels in the majority of them contradict the comorbidity model (Catts et al., Reference Catts, Adlof, Hogan and Weismer2005) that predicts a relatively low overlap between dyslexia and DLD, since phonological deficits and word reading problems were not assumed to characterize the latter group. In this respect, the present findings align better with the higher overlap estimates predicted by the two-dimensional model (Bishop & Snowling, Reference Bishop and Snowling2004). However, the conservative interpretation of this model’s prediction is that children with DLD would always have dyslexia with no existence of children with only DLD (Ramus et al., Reference Ramus, Marshall, Rosen and Van Der Lely2013). The current results do not support the perfect overlap between DLD and poor reading, since 37% of children with LLLPh had intact reading skills. Since our low scores on language and reading were determined by the cutoff criteria of 25th percentile, children with relatively mild linguistic “deficit” might be part of the PL and PLPPh profiles. We suggest that the ascription of such children might increase the probability of intact reading among these profiles. Furthermore, the phase at which the reading abilities were assessed (in the middle of first grade) might not be sensitive enough to differentiate between good and poor readers since typical readers still acquiring reading milestone. Also, considering the variability of methodological methods, tasks’ repertoire, and sampling across studies, this conservative assumption becomes practically undefendable.
Adopting a data-driven approach that quantitatively proposes an estimated prevalence of reading difficulties among children with LLLPh would be more suitable because it informs an already existing large body of studies. More particularly in English, prevalence estimates of 36.8% (Young et al., Reference Young, Beitchman, Johnson, Douglas, Atkinson, Escobar and Wilson2002), 48% (Snowling et al., Reference Snowling, Hayiou-Thomas, Nash and Hulme2020), 43% (Snowling et al., Reference Snowling, Nash, Gooch, Hayiou-Thomas, Hulme, Language and Team2019), and even 51% (McArthur et al., Reference McArthur, Hogben, Edwards, Heath and Mengler2000) were reported. Relative to other reports, the relatively high prevalence of low reading levels observed here among children with the LLLPh can be attributed to the binding of PA and language skills in the inclusion criteria for this group. Indeed, in contrast with the binding implemented in the current study, the weight given to phonological skills in defining DLD differs across studies. In addition, given the fact that no discrepancy criterion between IQ and achievement scores was included in the current study may also affect (enlarge) the observed overlap. The controversy regarding this criterion is still up to date (for more details, refer to Adlof & Hogan, Reference Adlof and Hogan2018; Catts et al., Reference Catts, Fey, Zhang and Tomblin1999; Tannock, Reference Tannock2013).
However, it should be noted that despite the differences in prevalence estimates between studies, the reported values still align well with the frequently co-occurrence wide range of 17%–71% found across studies. These differences could be attributed to sampling differences (clinically referred samples vs. epidemiological studies), differences in time point of diagnosis (parallel or consecutively) (Adlof & Hogan, Reference Adlof and Hogan2018), or other methodological issues such as inclusion criteria and definitions. The risk ratio to having low reading skills among children with language difficulties found in our study is not exceptional (LLLPh is 4.93 times more likely to display reading difficulties than TLTPh) compared to previous findings. A previous study had reported a risk ratio of 4.6 for reading disabilities 14 years after DLD was identified (Young et al., Reference Young, Beitchman, Johnson, Douglas, Atkinson, Escobar and Wilson2002). It appears clear from the results reported here that low linguistic skills in kindergarten place a child at substantial risk of later reading difficulties and academic vulnerability.
Limitations
The results of the current study represent, to our knowledge, the first evidence of assessing reading skills in Arabic-speaking children with different linguistic profiles. Any conclusion in this context must be formulated with some caution. First, because the phonological tasks used here were limited to isolation tasks that might not be the best reliable combination to represent the PA construct at this assessment time (kindergarten: before formal reading) (Mansour-Adwan et al., Reference Mansour-Adwan, Asadi and Khateb2020). Hence, future studies should assess the overlap between LPh and LR groups based on different phonological tasks (e.g., syllable/phoneme deletion tasks or blending tasks). Second, the current measures were drawn upon only one component of phonological processing, with no reference to verbal short-term memory, working memory, automatized rapid naming, or other skills that might be crucial for decoding written words and for differentiating between children with DLD and children with dyslexia (Bishop et al., Reference Bishop, McDonald, Bird and Hayiou-Thomas2009; Ramus et al., Reference Ramus, Marshall, Rosen and Van Der Lely2013). It had been found that rapid naming has differentiated children with DLD from children with DLD + dyslexia and was more related to reading than to language (Bishop et al., Reference Bishop, McDonald, Bird and Hayiou-Thomas2009). Third, the fact that the frequency of items unique for StA was not completely balanced across reading tasks (8% and 14% for the frequent and complex words, respectively) might unintentionally influenced the difficulty level of these tasks beyond the syllabic and morphological complexity of the words. Fourth, we suggest treating the current linguistic and reading categorizations carefully, since we know little at this phase about the stability of these groups over time. As a previous study has shown, measuring reading status stability between second and eighth grade revealed that only less than half of the children met the criteria for persistent reading disorders, besides significant proportions of children who were late-emerge reading disordered or others who had resolved their reading disorders (Torppa et al., Reference Torppa, Eklund, van Bergen and Lyytinen2015). The stability of the phonological deficit is vital because resolved deficiency of this domain among children with DLD was previously reported alongside their relatively good decoding skills (Bishop et al., Reference Bishop, McDonald, Bird and Hayiou-Thomas2009; Snowling et al., Reference Snowling, Nash, Gooch, Hayiou-Thomas, Hulme, Language and Team2019, Reference Snowling, Hayiou-Thomas, Nash and Hulme2020). Finally, the unique contribution for each of the oral language skills to reading cannot be inferred due to the nature of the specific methodology of a multivariate language-based achievement. In this regard, previous studies imply that the correlations between unitary linguistic domains and reading would be much weaker relative to a composite measure of oral language (Catts et al., Reference Catts, Fey, Zhang and Tomblin1999). For a synthetic language with rich and dense morphology, such as Arabic (Tibi & Kirby, Reference Tibi and Kirby2019; Tibi & Kirby, Reference Tibi and Kirby2017, Wattad & Abu Rabia, Reference Wattad and Abu Rabia2020), morphology might be neglected and undermined when grouping language skills together. To disentangle the effect of morphology from vocabulary and syntax, we suggest examining the probability of reading difficulties among children with low morphological skills, children with low non-morphological skills, children with low syntactic skills, and children with low non-syntactic skills.
Conclusions
This longitudinal study showed that establishing linguistic profiles in kindergarten based on the distinction between phonological and language skills allowed the identification of children with LLLPh that appeared to be the most at risk for reading difficulties 1 year later. The low performance of this “double-deficit” group relative to “single-deficit” groups (LL, LPh) provided evidence in support of the CLA, indicating that language skills, as well as phonological skills, significantly affect reading skills in first grade. We propose that children with low phonological and language skills are undoubtedly inferior in reading, which may increase their academic vulnerability. The high prevalence of reading difficulties among this group of children confirms that early language skills and reading are tightly related and that early language interventions are essential.
Supplementary material
For supplementary material accompanying this paper visit https://doi.org/10.1017/S014271642300019X
Replication package
The data and code required to replicate all analyses in this article are available at https://osf.io/fz6j4/
Acknowledgments
This work was supported by the Israeli Science Foundation (Grant no’ 2695/19) and by the Edmond J. Safra Brain Research Center for the Study of Learning Disabilities.
Open access
