Early cognitive and linguistic profiles of different types of 7‐ to 8‐year‐old readers
Abstract
enThe aim of this study was to investigate the early characteristics of four profiles of readers established in second grade (7–8 years of age): good readers, specific poor decoders, specific poor comprehenders and general poor readers. These profiles were compared retrospectively on a range of measures administered 2 years earlier, in kindergarten. These measures were based on factors known to be predictors of either decoding skills or comprehension performance. The results showed that children experiencing comprehension difficulties in second grade have early deficits in vocabulary, inferencing, working memory and morphology, while the children with decoding difficulties have limited letter knowledge and rapid naming performance. Phonological skills and verbal short‐term memory appeared to be deficient in all three profiles of struggling readers. These results are discussed in terms of the early identification of at‐risk children in kindergarten and the remedial programmes that could be provided to them at an early stage.
Highlights
aaWhat is already known about this topic?
- Reading implies two main components: reading and comprehension.
- Each component of reading is underpinned by different predictors.
- Different profiles of struggling readers with specific deficit to one component could be identified.
What this paper adds?
- Profiles of struggling readers are characterised by different deficits in linguistic and cognitive skills.
- These deficits are observed early on, before the formal instruction of learning to read.
Implications for practice and/or policy:
- Different remediation programmes should be administered to different types of struggling readers.
- These programmes should foster the deficits characterising each profile.
- These programmes could be proposed early on to prevent future difficulties in reading.
Recent results of an epidemiological survey conducted in the Organisation for Economic Co‐operation and Development (OECD) countries show that a significant number of children still have reading difficulties at the end of compulsory education (OECD, 2014). In France, for example, the proportion of these struggling readers at the end of the ninth grade represents around 20% of the population. This proportion does not seem to decrease over time (see Programme for International Student Assessment's results in the past few years). Researchers have generally acknowledged that the difficulties experienced by older students start to develop early on, sometimes well before formal instruction in learning to read. The reading difficulties these children will experience at a later age could therefore be detected at an early stage by examining their reading‐related linguistic skills and cognitive abilities. These difficulties might also affect different aspects of the reading ability itself. It therefore seems crucial to examine the precise nature of the difficulties these children may present. The objective of the present study was thus to distinguish between different profiles of poor readers at the beginning of learning to read and to compare these profiles on a large number of cognitive and linguistic skills measured in kindergarten. Few studies have adopted a longitudinal approach to address this issue (but see Catts, Hogan & Fey, 2003; Elwér, Keenan, Olson, Byrne & Samuelsson, 2013), and few have been conducted in languages more transparent than English even though this factor could lead to important differences (Florit & Cain, 2011). Finally, the particular focus of the present research also lies in its emphasis on the factors known to be specifically related to text comprehension beyond those known to promote the development of decoding skills.
Different profiles of poor readers
According to the Simple View of Reading (Hoover & Gough, 1990), reading is the product of two main components: word reading and language comprehension. The identification of written words is a component that is specific to reading, whereas the comprehension component appears to be more general and not specific to one particular modality (Gernsbacher, Varner & Faust, 1990; Kendeou, Bohn‐Gettler, White & van den Broek, 2008). According to this model, word decoding and comprehension are highly inter‐related. As a result, if any of these components is deficient in any way (i.e., if the child cannot decode a word or is not able to correctly understand an utterance), then reading will not be fully effective. Nonetheless, research has also demonstrated that each of these components can be specifically impaired (e.g., Aaron, 1991; Aaron, Joshi, Gooden & Bentum, 2008). Hence, several profiles of readers can be distinguished. In most readers, good reading comprehension is coupled with good decoding abilities. Indeed, automatic decoding processes free up attention and working memory resources, which are therefore available for comprehension (Laberge & Samuels, 1974; Perfetti & Hart, 2001). However, slow or inefficient word reading will lead to the consumption of cognitive resources and therefore hinder the understanding of what is being read. Children with such a profile might exhibit reading comprehension difficulties solely because of their inadequately automated decoding skills. This pattern of performance is typical of dyslexic individuals (e.g., Lyon, Shaywitz & Shaywitz, 2003). However, other types of struggling readers might present deficits specific to only one aspect of reading. On the one hand, some poor readers, labelled as poor decoders, have difficulties in decoding written words but no deficit in listening comprehension. As stated earlier, they may present difficulties in understanding written texts, but these difficulties do not seem to arise when they are given unlimited reading time (Walczyk, Marsiglia, Bryan & Naquin, 2001). On the other hand, researchers have isolated another profile of children who possess adequate decoding skills but exhibit a persistent deficit in comprehension for both written and oral language (e.g., Cain & Oakhill, 2006). These children may be labelled as specific less‐skilled comprehenders. Finally, some children might experience difficulties in both domains: identifying written words and understanding language (Aaron, 1991; Aaron, Joshi & Williams, 1999). The prevalence of these different profiles varies greatly depending on the native language, age and characteristics of the sample, but struggling readers in general represent around 10% and up to 15% of the school age population (Torppa et al., 2007). The distribution of these profiles also varies. In their study with English‐speaking second graders, Catts et al. (2003) found more children with specific decoding difficulties (35.5%) or general reading difficulties (35.7%) than with specific comprehension difficulties (15.4%).
Early precursors of reading
Longitudinal studies have highlighted several early‐emerging factors in kindergarten that might make it possible to predict later reading levels in elementary school. In this regard, authors have generally observed that some of these factors are more specifically related to one component of reading (decoding or comprehension) than the other. For instance, as regards decoding abilities, there is a relative consensus about the involvement of three main factors in predicting the accurate development of word reading skills. First, the acquisition of the alphabetic principle depends mainly on the ability to represent and manipulate the meaningless and abstract units of one's native language (phonemes, syllables, etc.). Difficulties in phonological processing in young children therefore generally lead to difficulties in acquiring efficient word reading abilities (e.g., Ecalle & Magnan, 2007; Wagner, Torgesen & Rashotte, 1994; Wimmer, Mayringer & Landerl, 2000). Poor decoders therefore usually present poorer phonological performance (Bowey, Cain & Ryan, 1992; Gillon & Dodd, 1994; Jimenez, 1997). Second, numerous studies have demonstrated that greater knowledge about letters at an early age predicts future word decoding abilities (e.g., Ecalle, Magnan & Biot‐Chevrier, 2008; Foulin, 2005; Muter, Hulme, Snowling & Taylor, 1998; Scarborough, 1998). In alphabetic systems, knowledge about the names of letters could indeed help subjects to access the corresponding sounds and thus establish the first grapheme‐to‐phoneme correspondences (e.g., Badian, 1995; Catts, Fey & Tomblin, 2001; Puolakanaho et al., 2007; Schatschneider, Fletcher, Francis, Carlson & Foorman, 2004). Finally, relatively strong predictive associations have also been observed between word reading and the speed of access to lexical representations (rapid automatised naming [RAN]; de Jong & van der Leij, 1999; Wolf & Bowers, 1999). In sum, better phonological skills, greater letter knowledge and faster access to lexical representations in kindergarten constitute the basis for an effective acquisition of decoding processes. Thus, children with a deficit in word identification (i.e., poor decoders or general poor readers) should exhibit impairments in these three abilities at a younger age, before formal instruction in learning to read. However, such impairments have not been systematically identified in previous research (Catts et al., 2003).
While the predictors of decoding abilities have been extensively documented, few studies have specifically investigated the precursors of reading comprehension performance within a longitudinal framework starting in kindergarten (but see Cain & Oakhill, 2006, or Oakhill & Cain, 2012, for longitudinal follow‐ups in the elementary school years). However, research in the field of comprehension development or the characteristics of less‐skilled comprehenders has demonstrated that several factors support reading comprehension (for reviews, see Oakhill & Cain, 2003; Rapp, van den Broek, McMaster, Kendeou & Espin, 2007). Reading comprehension is indeed a complex task that draws on many different skills and processes at (1) word level (e.g., word identification, vocabulary and morphology); (2) sentence level (e.g., syntax); and (3) text level (e.g., inferencing). One might therefore expect children with a comprehension deficit in the elementary school years (i.e., poor comprehenders and general poor readers) to present impaired performance in these abilities in kindergarten. To our knowledge, little research has directly examined all these comprehension‐related precursors in one single study within a longitudinal framework. Nonetheless, Elwér et al. (2013) have reported that less‐skilled comprehenders tend to exhibit grammatical limitations in kindergarten, but not necessarily early deficits in vocabulary, as compared with specific poor decoders. Further studies are therefore needed to explore the specific early profiles these children might present in kindergarten.
Finally, even if the two components of the Simple View of Reading (Hoover & Gough, 1990) model seem to be underpinned by specific predictors, some factors might be involved in an undifferentiated way in both components. Memory capacities, for instance, could be called upon both to identify written words (involvement of the phonological loop in working memory; Swanson & Berninger, 1995) and when there is a need to connect different statements in the text or to update the mental representation as the text is being read (Fletcher, Lyon, Fuchs & Barnes, 2007; Gathercole, Alloway, Willis & Adams, 2006). In this latter case, the central executive component of working memory seems to be the memory system that is most strongly solicited (Swanson, Howard & Saez, 2006). Consequently, children with poor reading skills (whichever component of reading is deficient) could have lower memory capacities (Savage, Lavers & Pillay, 2007). However, specific poor decoders might be more impaired in tasks involving the phonological loop and the passive storage of information (e.g., classic word span tasks), whereas poor comprehenders could be impaired in tasks that rely mainly on the central executive system and require the transformation of the stored information (e.g., backward span, n‐back; Carretti, Borella, Cornoldi & De Beni, 2009; Goff, Pratt & Ong, 2005; Swanson, 1999). To address this issue, the present study made use of two different measures to assess children's working memory abilities: one to evaluate their phonological loop capabilities (passive storage of verbal information) and the other to assess the central executive system (updating processing).
Objectives and hypotheses
In sum, each component of reading seems to be foreshadowed by different linguistic and cognitive abilities. As a result, children experiencing difficulties in one aspect of reading only could exhibit particular deficits in these underlying factors. As all of these factors can be assessed via the oral modality, these particular profiles could therefore be detected at an early age, before formal instruction in learning to read. The aim of the present study was to investigate this issue using a retrospective comparison between different profiles of poor readers identified in second grade based on several linguistic and cognitive reading‐related skills evaluated in kindergarten. Such an approach goes beyond previous results from classical correlational studies as it permits the differentiated examination of the role of several predictors of the two components of reading. In addition, while few studies have directly investigated this question within a longitudinal framework, the results reported in the current literature are not always clear‐cut with regard to the particular patterns of performance different profiles of readers might present at an early age (Catts et al., 2003; Elwér et al., 2013). We expected that comprehension‐related measures (e.g., vocabulary, morphology and syntax) administered in kindergarten would be impaired in children with specific comprehension difficulties in second grade. Conversely, these children might perform as well as good readers in tasks related to decoding (e.g., phonological measures, rapid automated naming and letter knowledge). For these latter measures, we expected the poor decoders would typically exhibit early difficulties. As regards memory tasks, both groups of poor comprehenders and poor decoders would present deficient scores. Finally, we expected children detected as general poor readers in second grade to exhibit the difficulties experienced by both the poor comprehenders and the poor decoders in all types of tasks.
Method
Population
The original sample at the beginning of the study consisted of 131 children in kindergarten from six schools (mean age = 5.6 years old). Two years later, in second grade, complete data were only available for 106 children (mean age = 7.2 years old) because of some children having moved away between kindergarten and elementary school or having been absent during one or more testing sessions. All these children were native French speakers, and none of them had any cognitive impairments, behavioural problems or diagnosed language difficulties. Among these remaining 106 second graders, we sought to determine different profiles of readers based on the children's performance on a word reading measure and a listening comprehension task (see below for a description of these two tasks). An analysis of deviance was conducted by calculating z scores for all the children on these two measures. The cut‐off criterion of a z score lower than −1.3 was considered deviant from the norm and corresponded to the 10th percentile. A task was not considered successfully completed if the associated z score was lower than −1. The differences between z scores in the listening comprehension and the word reading task were also calculated for the groups of poor decoders and poor comprehenders to ensure they indeed presented a specific deficit in only one reading component. All the children in these groups presented a difference greater than 0.50 (mean difference for the two groups = 1.5). In sum, four profiles of children were identified in our data set: 79 good readers (word reading+/comprehension+), 11 specific less‐skilled comprehenders (word reading+/comprehension−), 10 specific poor decoders (word reading−/comprehension+) and 6 general poor readers (word reading−/comprehension−). To assume equal variances across the groups, a smaller subset of the 79 good readers (n = 15) was then randomly selected to form the final control group for subsequent statistical analyses.
Material
Measures used in second grade
Word reading
Word reading was assessed using a task extracted from the French Screening Battery for Dyslexia (ODEDYS; Jacquier‐Roux, Valdoix & Zorman, 2005). In this task, the children had to read aloud three types of items in a limited amount of time: 20 regular words, 20 irregular words and 20 phonologically plausible nonwords. The numbers of correctly read items and the time taken to read each type of word were recorded. A composite score was then calculated by dividing the number of correctly read words by the time taken to read them.
Listening comprehension
To assess listening comprehension, we used a short narrative (169 words) in which the word frequency was controlled for (Lété, Sprenger‐Charolles & Colé, 2004). This task was first pretested in a large population (N = 158) of children in second grade. The experimenter read the text aloud and then asked 12 multiple‐choice questions: four questions referred to the explicit information present in the text (literal comprehension), four questions required the generation of coherence inferences and four questions required the production of knowledge‐based inferences. The order of presentation of the questions was randomised. For each question, three possible responses were presented. The dependent variable was the total number of correct responses. Reliability for this task, assessed by calculating the Cronbach alpha over items, was adequate at α = .71.
Reading comprehension
The same task as in the listening comprehension condition was used here. In the reading condition, the children had to read the text silently and then answer the 12 questions by circling the correct answer. They were not able to consult the text when responding. The time taken to complete the task was not controlled for. The dependent variable was the total number of correct responses (maximum = 12). Reliability for the reading comprehension task was adequate at α = .71.
Measures in kindergarten
Listening comprehension
A short narrative (11 sentences, 136 words; Brigaudiot, 2000) was read aloud to each participant by the experimenter without any pictorial support. Comprehension was then evaluated using 20 yes/no questions. Ten questions referred to information that was explicitly stated in the text (literal questions) while the other 10 questions required the production of an inference (inferential comprehension). The order of presentation of the questions was randomised. The dependent variable was the number of correct responses. The maximum score was 10 for each type of questions (literal and inferential). Reliability for this task was adequate at α = .72.
Vocabulary
Receptive vocabulary was evaluated using a standardised test for assessing oral language in children (ELO, Oral Language Assessment; Khomsi, 2001). In this task, the children had to correctly identify a named picture from four possible black‐and‐white drawings. The total number of correct answers constituted the total raw score. The test consisted of 20 items including one first item by way of example. The maximum score was therefore 19.
Morphological knowledge
This task (Sanchez, Ecalle & Magnan, 2012) involved the judgment of morphological relationships. The children had to decide whether two aurally presented words did (or did not) belong to the same morphological family. The test consisted of 16 pairs of words, eight of which had a morphological relationship (e.g., danse/danseur [dance/dancer]) and eight of which only had a phonological relationship (e.g., rat/râteau [rat/rake]). After giving two examples, the order of presentation of the pairs was randomised. The dependent variable was the total number of correct responses (maximum = 16). Reliability was adequate at α = .75.
Syntactic knowledge
In this task, the children had first to judge the grammaticality of 12 sentences. Six of these sentences were correct, and six were agrammatical. Two types of grammatical errors were presented: three sentences had a complete word order violation (e.g., sur table la est couteau le [on table the is knife the]), and three sentences had a word order violation but without nominal syntagm disconnection (e.g., à la fraise mange Marie une glace [a strawberry ice‐cream Marie is eating]). After judging the grammaticality of the sentences, the children then had to correct the sentences they considered to be incorrect. The dependent variable was a composite score of the children's performance on these two tasks. The maximum score was 18. Reliability was adequate at α = .68.
Central executive system of working memory
An experimental n‐back task was used to assess the central executive system of working memory (Potocki, Ecalle & Magnan, 2013). The children had to decide whether a picture presented on a screen (at a rate of one picture every 2 seconds) was identical to the picture presented one or two steps previously. The pictures corresponded to concrete everyday objects. In order to diminish the semantic load of the task, the frequency of the words depicted by the pictures was controlled for (Lété et al., 2004). The task consisted of four trials of five items each. The dependent variable was the total number of correct responses (maximum = 20). Reliability for this task was adequate at α = .66.
Verbal short‐term memory
Verbal short‐term memory was assessed using a nonword repetition task extracted from the French ODEDYS (Jacquier‐Roux et al., 2005). In this task, the children had to repeat 20 nonwords of increasing length that were first pronounced orally by the experimenter. The dependent variable was the number of correctly recalled words.
Phonological skills
The THaPho (Ecalle, 2007) standardised test was used to assess the children's phonological skills. In this test, the children had to name the sound that was common to two words presented orally. In a first condition, the two common sounds were syllables (e.g., cochon/bouchon) while in a second condition, they were phonemes (e.g., mur/manche). These sounds could appear either at the beginning or at the end of the word. Each condition consisted of six trials. One score was computed for each condition. It represented the total number of correctly extracted sounds.
Letter knowledge
Letters were written in capital letters, and each was presented on an individual cardboard sheet (15 × 10 cm). The order of letter presentation was randomised. The children had to give the name of each letter. One point was awarded for each correct response. The dependent variable was the total number of correct responses. The maximum score was 26.
Rapid serial naming
A traditional RAN paradigm was used (Denkla & Rudel, 1976). In this task, which was extracted from the standardised ODEDYS (Jacquier‐Roux et al., 2005), the children had to name, as fast as they could, a series of 25 pictures depicting everyday objects. Before starting the task, the experimenter made sure the objects and pictures were correctly recognised by the children. The total response time taken to complete the tasks was recorded. Shorter times therefore indicated better performance.
Nonverbal intelligence
Children's nonverbal reasoning abilities were controlled for using a task extracted from the Evaluation des Compétences Scolaires (Assessment of School Skills; Khomsi, 1997) standardised test. This test is based on the same principle as the Raven Progressive Matrices: using logical reasoning, the children had to identify the missing element in a drawing. The task consisted of 15 items, and the dependent variable was the number of correct responses.
Results
Descriptive statistics are presented in Table 1. Beyond the tasks allowing their identification in second grade, each profile of readers seemed to exhibit a particular pattern of performance in the different tasks administered in kindergarten. For instance, whereas the poor decoders presented low scores in letter knowledge and RAN, the poor comprehenders had more difficulty with the vocabulary and morphological tasks. However, the four profiles of readers seemed to exhibit relatively equivalent scores in the syntactic knowledge task. The phoneme extraction task also appeared very difficult for all the children and especially for the three groups of struggling readers.
| Good readers, N = 15 | Poor comprehenders, N = 11 | Poor decoders, N = 10 | Poor readers, N = 6 | |
|---|---|---|---|---|
| Decoding (G2) | 0.69 (0.07) | 0.57 (0.20) | 0.14 (0.07) | 0.13 (0.07) |
| Listening comprehension (G2) | 9.9 (0.83) | 5.8 (0.45) | 9.2 (1.9) | 5.7 (0.82) |
| Reading comprehension (G2) | 9.6 (1.3) | 6.5 (1.4) | 6.4 (3.3) | 4 (2.1) |
| Nonverbal reasoning (KG) | 5.3 (1.4) | 4.6 (1.7) | 4.3 (2.4) | 4.3 (1.9) |
| Literal comprehension (KG) | 9.6 (0.50) | 8.4 (0.9) | 8.9 (1.7) | 7.8 (1.8) |
| Inferential comprehension (KG) | 5.8 (1.1) | 4.3 (0.88) | 5.1 (0.74) | 4 (0.63) |
| Vocabulary (KG) | 13.3 (1.8) | 10.5 (1.7) | 12.5 (1.6) | 10.2 (1.5) |
| Morphology (KG) | 11.8 (2.1) | 9.6 (2.2) | 12.1 (1.9) | 9 (3.2) |
| Syntax (KG) | 11.9 (2.9) | 11.8 (2.5) | 11.5 (3.9) | 10.5 (4.4) |
| Letter knowledge (KG) | 23 (2.5) | 22.7 (3.3) | 18.4 (7.4) | 17.3 (5.7) |
| RAN (KG) | 35.8 (10.1) | 32.2 (4.7) | 44.3 (10.4) | 43.3 (5.4) |
| Phoneme extraction (KG) | 1.5 (1.6) | 0.5 (1.2) | 0.1 (0.3) | 0.2 (0.4) |
| Syllable extraction (KG) | 2.5 (1.5) | 1.9 (1.4) | 1.4 (1.3) | 1.8 (2.1) |
| Working memory (KG) | 13 (1.9) | 10.5 (2) | 13.1 (2.3) | 10 (4.3) |
| Short‐term memory (KG) | 16.7 (2.9) | 13 (3.7) | 14 (3.8) | 14.6 (2.5) |
Given the small sample sizes in each group, nonparametric Kruskal–Wallis tests were used to compare the scores of the different groups. First, the nonverbal reasoning performance of the three profiles of struggling readers was analysed to make sure their reading difficulties could not be due to more general cognitive limitations. This did not seem to be the case as no significant differences between these three groups and the group of good readers were observed on this measure.
The four groups were then compared on the reading comprehension task administered in second grade. Here, we observed a significant effect of profile (χ2 = 19.28, df = 3, p < .001) but no significant differences between the three groups of struggling readers, each of them performing significantly more poorly than the good readers.
Then, we examined the differences in the early linguistic and cognitive skills of the four profiles of readers on the measures administered in kindergarten.
Comprehension measures
As regards the predictors of comprehension performance, the analyses revealed a significant effect of group in the listening comprehension task, for both the literal question (χ2 = 9.1, df = 3, p = .02) and the inferential question (χ2 = 14.67, df = 3, p = .002). For both types of questions, Dunn's multiple‐comparison test revealed that the poor comprehenders and general poor readers obtained significantly lower scores than the good readers. In addition, for the inferential questions, the scores of these two groups were also lower than those obtained by the poor decoders. No differences were observed between poor decoders and good readers on these comprehension tasks. We also observed a significant effect of profile on the vocabulary measure (χ2 = 13.25, df = 3, p = .004) and on the morphological knowledge task (χ2 = 10.25, df = 3, p = .02). In these tasks, the performance of the poor comprehenders and general poor readers was lower than that of both the good readers and poor decoders (Figure 1). No significant effect of group was found in the syntactic knowledge tasks.
Word reading measures
As far as word reading skills were concerned, the analysis revealed a significant effect of group for the letter knowledge task (χ2 = 8.37, df = 3, p < .05), with the poor decoders and general poor readers having significantly less knowledge about letters in kindergarten than the good readers and poor comprehenders (Figure 2). Similarly, we observed a significant effect of group for the RAN task (χ2 = 16.58, df = 3, p < .001). In this task, the poor comprehenders performed as well as the good readers (shorter response times). In contrast, both the poor decoders and general poor readers presented longer response times (Figure 2). These two groups did not significantly differ from each other.
A significant group effect was also found in the phoneme extraction task (χ2 = 8.15, df = 3, p < .05). In this task, the performance of the three groups of struggling readers – poor comprehenders (mean score = 1.9; SD = 1.4), poor decoders (mean = 1.4; SD = 1.3) and general poor readers (mean score = 1.8; SD = 2.1) – was below that of the good readers (mean score = 2.5; SD = 1.5). The group effect was not significant for the syllable extraction task.
Memory tasks
Finally, regarding the memory measures, the analyses revealed a significant effect of group for both the working memory (χ2 = 9.16, df = 3, p < .05) and short‐term memory measures (χ2 = 8.15, df = 3, p < .05). For the latter, the scores of the three groups of poor readers were lower than those obtained by the good readers (Figure 3). In the working memory task, however, the poor decoders did not differ from the norm, whereas the performance of the poor comprehenders and general poor readers was lower than that of the other two groups of poor decoders and good readers (Figure 3).
Discussion
The Simple View of Reading model (Gough & Tunmer, 1986) defines reading ability as the product of two main components: decoding and comprehension skills. Based on this model, different subtypes of readers can be identified: good readers (good in both word reading and listening comprehension), specific poor decoders (deficit in word reading only), specific poor comprehenders (deficit in listening comprehension only) and general poor readers (deficits in both word reading and listening comprehension). Most previous studies of these reader subtypes have been based on a single point in time, and there has rarely been any longitudinal follow‐up (but see Catts et al., 2003; Elwér et al., 2013; Torppa et al., 2007). In addition, the difficulties experienced by struggling readers of elementary school age are generally foreshadowed by early deficits in linguistic and cognitive reading‐related factors. The main objective of this study was therefore to retrospectively compare these four profiles of readers detected in second grade on a range of measures administered 2 years previously, in kindergarten. We sought to determine various early linguistic and cognitive profiles that could lead to different types of reading disabilities later on. Such an approach also furthers our knowledge about different precursors of reading by examining their role in both decoding and listening comprehension in a differentiated way.
Our results confirm that each component of the Simple View of Reading model (i.e., decoding and comprehension) is primarily underpinned by specific predictors. Indeed, we found that children with impairments in word reading skills (poor decoders and general poor readers) had limited letter knowledge and RAN abilities in kindergarten, whereas children with specific comprehension difficulties did not present such limitations. Conversely, children who presented particular difficulties in language comprehension (poor comprehenders and general poor readers) showed early deficits in inferencing, vocabulary, morphology and working memory (central executive) capacities, whereas specific poor decoders did not. These results therefore confirm previous research into the development of reading by highlighting the involvement of several linguistic and cognitive factors in this ability and the need to differentiate between the two main components of reading (Bianco et al., 2010; Oakhill, Cain & Bryant, 2003).
However, we were unable to show that some factors typically acknowledged as being involved either in decoding or in comprehension do indeed make it possible to distinguish between the different profiles of poor readers. First of all, syntactic knowledge did not appear to be as severely impaired in the two profiles of children with comprehension difficulties as might have been expected from previous studies (Bentin, Deutsch & Liberman, 1990; Gaux & Gombert, 1999; Nation & Snowling, 2000). Yet this result is not that surprising given that the role of syntax in comprehension is still the subject of debate in the current literature. Cain and Oakhill (2007; see also Oakhill et al., 2003), for instance, have suggested that syntactic abilities could play a role in comprehension, but only later in development. The absence of deficits in the poor comprehenders could also be explained by the specific task used to assess syntactic knowledge (grammatical judgment and correction), which may call less upon sentence comprehension skills than other tasks typically do (e.g., Test for the Reception of Grammar; Bishop, 2003). Second, phonological abilities did not appear to be a specifically linked predictor to the development of decoding skills. Instead, all the profiles of poor readers, and even those with a specific deficit in comprehension, performed worse in the phonemic awareness task than the good readers. This result, however, has already been observed in the literature (e.g., Nation, Cocksey, Taylor & Bishop, 2010; Sparks, 2001). For example, Catts et al. (2003) reported that the poor comprehenders in their sample performed as poorly as children with word recognition difficulties in a phonological awareness and rapid naming task. Some authors have suggested that this finding may be explained by the poor comprehenders' difficulties in understanding the task demands and may not, in fact, be a reflection of a true phonological deficit (Nation, 1999). Another explanation could take the form of the resource limitations that all types of struggling readers might experience, especially when required to store verbal material (Carretti et al., 2009; Savage et al., 2007; Seigneuric, Ehrlich, Oakhill & Yuill, 2000). Hence, some developmental studies have found evidence that phonological skills at a younger age could be significant predictors of language comprehension performance at an older age (Vellutino, Tunmer, Jaccard & Chen, 2007). Indeed, comprehension might depend to some extent on the subject's capacity to use speech‐coding mechanisms to encode and retain linguistic units in order to assemble them into meaningful propositions. This could explain why the poor comprehenders exhibited difficulties in our phonological task in which they had to encode and store two words in order to find their common sounds.
Similarly, while most of the reading‐related precursors seem to be preferentially involved in one component of reading rather than the other, other factors could foreshadow both components. This is the case for short‐term memory capacities, which appeared to be impaired in all three groups of struggling readers. This result confirms the theoretical view that different systems of working memory are preferentially involved in one specific reading activity (Cain, Oakhill & Bryant, 2004; Stothard & Hulme, 1992; Swanson & Berninger, 1995; Swanson et al., 2006). This is especially true for the central executive system of working memory, which appeared to be specifically impaired in the children with comprehension difficulties (e.g., Cain et al., 2004; Yuill, Oakhill & Parkin, 1989), whereas the children with specific word reading impairments did not present such difficulties. However, the passive storage of verbal information based on the phonological loop of working memory seems to be involved in both decoding and comprehension. Indeed, in this task, all the groups of struggling readers performed worse than the good readers.
Finally, it is also worth noting that the three subgroups of struggling readers did not significantly differ from each other on the reading comprehension task. This result highlights the need to add a listening comprehension task to the screening battery used to assess young readers' reading skills. Indeed, the observed reading comprehension difficulties in these three groups did not arise from the same factors. While for the poor decoders, difficulties in understanding reading texts might have stemmed from insufficiently automated word decoding skills (Perfetti, 1985), the same difficulties in the poor comprehenders might have come from a core deficit in language comprehension (thus observed in both reading and listening situations).
Conclusion
To conclude, the present study has important implications, in particular with regard to the early identification of at‐risk children and interventions that could potentially be used to help them. Our results suggest that there are different profiles of struggling readers during the early stages of learning to read and that these children experience different types of reading difficulties. Each of these difficulties is also foreshadowed by different linguistic and cognitive impairments before formal instruction in learning to read. Consequently, children who are likely to present difficulties in acquiring the alphabetic code and decoding skills exhibit early limitations in phonemic awareness, letter knowledge and rapid naming. In contrast, the performance of children who are likely to become poor comprehenders tends to be impaired at the level of inferencing, vocabulary, morphology and working memory (central executive system). Interventions should therefore target these specific skills and be proposed in a differentiated way to each profile of at‐risk children. A study conducted by Bianco et al. (2012; see also Bianco et al., 2010) tested the effects of two types of training: one phonological training programme and one comprehension training programme. Both were presented in preschool and in kindergarten. The authors observed that these programmes led to improved performance the next year in first grade. These improvements were specific rather than general. Indeed, the children trained with the phonological programme exhibited better decoding abilities but not better comprehension, whereas the children who received the comprehension programme showed improved performance in comprehension but not in decoding. These results suggest that it is possible to train each component of reading at an early age by focusing on its specific underlying factors (Lonigan, Purpura, Wilson, Walker & Clancy‐Menchetti, 2013). Such interventions could be introduced before formal instruction in learning to read in an attempt to prevent severe reading disabilities from developing in the future.
Biographies
Anna Potocki is an assistant professor of Developmental Psychology at the University of Poitiers and works in the Center of Research on Cognition and Learning (CeRCA, UMR 7295). Her main research interests are related to the development of children's and adolescents' reading and comprehension skills.
Jean Ecalle is a Professor of Cognitive and Developmental Psychology at the EMC Laboratory of the Lyon 2 University. He is currently pursuing research on the field of reading development and reading difficulties. He is also interested in constructing new tools for assessing and remediating reading abilities in normal and pathologic children.
Annie Magnan is a Professor of Cognitive and Developmental Psychology at the University of Lyon 2. She is also a director of the team research devoted to the study of language at the EMC Laboratory and a member of the Institut Universitaire de France. Her main research interests are reading development and language or reading difficulties (dyslexia, specific language impairment and deafness).
Number of times cited: 1
- Nina Kleinsz, Anna Potocki, Jean Ecalle and Annie Magnan, Profiles of French poor readers: Underlying difficulties and effects of computerized training programs, Learning and Individual Differences, 57, (45), (2017).
