Focusing Our View of Dyslexia Through a Multifactorial Lens: A Commentary

Dyslexia is a language-based developmental reading and spelling disorder, with neurobiological origins, that occupies a complex multidimensional space comprising, but not limited to, issues of etiology, identification, prevention, intervention, advocacy, and policy. The papers that make up this special series of Learning Disability Quarterly focus primarily on issues of etiology, identification, and intervention. Together these papers help to place dyslexia within the broader category of children who struggle to learn to read and write and helps to orient us toward processes related to identification and intervention as it applies specifically to dyslexia and more broadly to general difficulties in reading and writing development. In this commentary a multifactorial model perspective is applied to the six papers that make up the special issue to examine the potential utility of such models to clarify issues related to children with dyslexia.

Until relatively recently, dyslexia has been portrayed as a specific deficit residing within the phonological system of the language network that results in difficulty with learning to decode and spell printed words while sparing other language functions such as semantics, syntax, and pragmatics (e.g., Schatschneider & Torgesen, 2004; Stanovich, 1988). And while there are certainly children who exhibit specific phonological processing deficits that negatively affect reading and spelling development without impacting general language functioning, it is clear that the vast majority of children who develop reading and writing difficulties have more global cognitive and linguistic issues.

Current models of dyslexia clearly indicate that single-deficit models are inadequate to account for the variability found in dyslexia (e.g., Pennington et al., 2012; Perry et al., 2019). Instead, new models tend to emphasize a more multifactorial approach to etiology with primary involvement of phonological processing deficits as well as weaknesses in other oral language skills, processing speed, and perhaps executive function (Petersen & Pennington, 2015). Multifactorial models are predicated on the idea that cognitive/linguistic risk factors are not deterministic but instead probabilistic, suggesting that not all individuals with dyslexia have the same underlying deficits that consistently lead to problems in learning to read and spell (Catts et al., 2017; Catts & Petscher, 2020; Snowling, 2008). In addition, there is growing evidence that dyslexia frequently co-occurs with other disorders such as attention-deficit hyperactivity disorder (ADHD, e.g., Boada et al., 2012; Willcutt & Pennington, 2000), math deficits (e.g., Landerl & Moll, 2010), and various language-based deficits such as specific language impairment (e.g., Bishop & Snowling, 2004; Snowling & Melby-Lervåg, 2016). Across the six papers that make up this special series, we can see a move away from single-factor accounts (i.e., reductionist approaches) toward more nuanced multifactor explanations of reading difficulties and more specifically dyslexia. These models allow us to consider how deficits in word reading, linguistic skills, and cognitive abilities (separately and in combination) influence development in children who experience difficulties learning to read and write across both transparent and opaque orthographies. In addition, multifactorial models of dyslexia have potentially important implications for early identification and intervention.

Traditionally, classification of dyslexia has focused on identifying children with severe reading and spelling deficits that are “unexplained” in relation to other cognitive abilities (Lyon et al., 2003). If strictly applied, these classification criteria result in selection of a “specific” class of reading and writing disability that emphasize underlying phonological processing deficits (e.g., phonemic awareness, rapid automatized naming, set for variability) with intact language functioning. However, multifactorial models allow influences outside the phonological domain of language to influence reading and spelling skill development. By selecting reading comprehension as the outcome measure of interest, three of the studies in this issue (Capin et al., 2020; Kim et al., 2020; Logvinenko et al., 2020) allow us to examine the contribution of word reading and broader linguistic processing to reading comprehension difficulties in various subtypes of readers based on the simple view of reading (Gough & Tunmer, 1986). Results from these studies certainly challenge the adequacy of single-deficit models of dyslexia and further highlight differential effects of orthographic structure on the relationship between word reading, linguistic skill, and reading comprehension.

In the case of Capin et al.’s (2020), the sample of fourth graders was selected based on poor reading comprehension skill permitting a search for homogeneous groups of poor readers to be identified based on the simple view of reading. Important questions here are whether a class of readers who are dyslexic would be identified, the relative size of the group, and associated cognitive/linguistic makeup. Latent profile analysis using word reading and listening comprehension measures revealed three distinct subtypes of poor comprehenders as follows: Class 1, moderate deficits in both word reading and listening comprehension of similar severity (moderate reading difficulty; 91%); Class 2, severe deficit in word reading paired with moderate listening comprehension deficit (severe word reading; 5%); and Class 3, severe deficit in listening comprehension with moderate word reading difficulties (severe listening comprehension; 4%). The two severe deficits groups, both considered “specific deficit” groups, had very similar reading comprehension skill but vastly different word reading and listening comprehension profiles indicating a dissociation between word reading and linguistic skill as hypothesized by the simple view of reading. As the authors note, results reflect both the severity and specificity hypotheses of reading skill deficits; however, less than 10% of students were classified into groups exhibiting specific reading comprehension difficulties (i.e., severe listening comprehension difficulties) and dyslexia (i.e., severe word reading difficulties). The cognitive/linguistic skill profiles of the dyslexic students tend to support a multifactorial model with this group performing lower on predictors commonly associated with phonologically based word reading deficits (phonological awareness and rapid automatized naming) along with other predictors not typically associated with a single-deficit model of dyslexia (e.g., low verbal knowledge and nonverbal reasoning). Overall, Capin et al.’s study suggests that there is a small group of poor readers that fits the general profile of dyslexia (poor word reading skill with relatively adequate linguistic competence). However, the cognitive/linguistic profile of this group of students does not strictly align with a single-deficit model of dyslexia. Importantly, the vast majority of students demonstrated commensurate difficulties in both word reading and listening comprehension suggesting the need for multicomponent interventions to address both linguistic and word reading deficits in a vast majority of students who struggle with reading comprehension in the upper elementary grades.

Using a similar study design, Kim et al. (2020) classified native Korean elementary school students based on their level of reading achievement across word recognition, reading fluency, and reading comprehension measures. Again, latent profile analysis was used to identify latent groups with similar patterns of reading achievement. Contrary to Capin et al.’s (2020) results, Kim et al.’s study identified groups based more on severity than specificity. Three latent profiles were extracted for primary-grade students, representing: Class 1 (i.e., Very Poor Word Readers, Poor Comprehenders, 12%), characterized as individuals who perform more than 1 SD below the sample mean on the word recognition and reading fluency measures and perform about 0.6 SD below the sample mean on the reading comprehension measure; Class 2 (i.e., Poor Readers, 26%) as individuals who perform between 0.5 SD and 0.7 SD below the sample mean on word recognition, reading fluency, and reading comprehension measures; and Class 3 (i.e., Average Readers, 62%) as individuals who perform above the sample mean on all three measures. Four latent profiles were extracted for intermediate-grade students, representing Class 1 (i.e., Very Significantly Poor Readers; 6%), characterized as individuals who perform more than 1 SD below the sample mean on all three measures; Class 2 (i.e., Very Poor Readers; 7%), as individuals who perform about 1 SD below the sample mean on all three measures; Class 3 (i.e., Average Word Readers, but Poor Comprehenders; 36%), as individuals who perform between 0.3 SD and 0.6 SD below the sample mean on the reading fluency and reading comprehension measures, yet have an average score on the word recognition measure; and Class 4 (i.e., Average Readers, 51%) as individuals who perform above the sample mean on all three measures.

Differences in terms of class specificity and severity between Capin et al. (2020) and Kim et al. (2020) could be due to several important differences across the studies. First, the two studies used different indexing skills to form latent profiles which could differentially affect the ability to identify classes that differ in both specificity and severity. In using word reading and listening comprehension as index variables for the classes, Capin et al. capitalized on weaker relationships between indexing measures that allowed for dissociation across classes. The decision of Kim et al. to use strictly reading measures as index variables, which are highly correlated, reduces the chance of measures demonstrating dissociations across classes. However, the more likely explanation has to do with characteristic differences in orthographies across the two studies. English is a relatively opaque orthography whereas Korean is transparent. In transparent orthographies children who are poor readers tend to develop fairly accurate word reading skill but are slow (Seymour et al., 2003; Ziegler & Goswami, 2005); whereas in opaque orthographies poor readers are more inclined to have persistent word reading accuracy difficulties (Landerl et al., 2013). Thus, in Korean most of the variance in reading comprehension is explained by differences in linguistic skill as opposed to word reading, resulting in profiles based more on severity of reading comprehension difficulties. This interpretation was supported by Logvinenko et al. (2020) who demonstrated that in Russian, another transparent language, word reading skill was not associated with individual differences in reading comprehension in a sample of elementary children Grades 2–4. Results of these studies support a model of reading development in which characteristics of the orthography affect the role of word reading in explaining reading comprehension difficulties and supports the view that the symptoms of dyslexia are moderated by the orthographic demands of learning to read a particular language (see Rueckl et al., 2015). It also seems to suggest that in English, there are likely groups of poor readers who have specific deficits (i.e., dyslexia and specific reading comprehension difficulties), but these specific groups are quite small compared with the size of a more mixed group representing difficulties with word reading and linguistic comprehension.

Turning now to issues of identification and instruction, multifactorial models of dyslexia support the notion of numerous potential pathways to word reading and spelling difficulty. However, the symptoms of dyslexia appear remarkably consistent across developmental pathways. And while there have been many attempts to distinguish meaningful subtypes of dyslexia in English-speaking populations including surface versus phonological (Castles & Coltheart, 1993), phonological versus rate (Wolf & Bowers, 1999), and IQ-achievement discrepant versus and garden variety poor readers (Stanovich, 1988), such efforts have generally identified very small classes of children without underlying phonologically based deficits varying in severity. This naturally leads to the question, what is it about English that makes learning to read difficult for children with dyslexia and why are phonological skill deficits so ubiquitous in the population? In English, the relationship between orthography and phonology is quite complex, requiring readers to grapple with both small and large grain relations between orthography and phonology (see Ziegler & Goswami, 2005). In addition, English vowel pronunciations are variant and are largely determined probabilistically by local letter context (see Kessler & Treiman, 2001), allowing a grapheme such as ea to be pronounced as /i/ in beat, /ɛ/ in head, /ɛə/ in pear, and /eɪ/ in steak.

To learn to read English efficiently, developing readers must engage in an item-based process of orthographic learning (see Nation & Castles, 2017) in which the reader binds phonology to orthography to create “amalgamated” and “modular” representations that are recognized instantly and impermeable to top-down influences (see Ehri, 2005; Stanovich, 1991). This process is highly dependent on readers having a fully represented phonological form of the word available for binding spelling to pronunciation in a way that totally distinguishes words such as trial from trail. It is important to note that there are not alternative developmental paths that lead to skill word recognition and spelling in English. As such, the absence of a phonological representation (i.e., lack of exposure to the word) or even a mild disruption to the phonological system can compromise the availability of a high-quality phonological representation and negatively affect the binding process. The lack of high-quality phonological representations ultimately reduces “grapho-phonic” knowledge (see Ehri & Saltmarsh, 1995) that is necessary to fully analyze matches between orthographic and phonological units to store high-quality word representations that include complete sublexical orthographic-phonological associations. This is consistent with the view that phonological processing deficits in children with dyslexia lead to a processing strategy that affords insufficient attention to sublexical orthographic-phonological relations (e.g., Compton, 2002; Compton et al., 2014), which results in a dependence on more global orthographic structure in dyslexia, not as a strategy but as a consequence of how orthographic-phonological mappings are learned (see Harm et al., 2003).

Findings from Steacy et al. (2020) support this general developmental model of dyslexia. In the study, children with dyslexia and typically developing children were exposed to differential mixes of words that support high- versus low-frequency vowel pronunciations. One training corpus contained a ratio of 80%–20% high- to low-frequency pronunciations (e.g., for ea; 80% ea pronounced as /i/ as in bead and 20% ea pronounced /ε/ as in dead) while the other consisted of a ratio of 20%–80%. The authors were interested in exploring the facilitative versus inhibitory effects of exposures to differential mixes of words (i.e., training corpora) that support high- versus low-frequency grapheme-phoneme correspondences (GPC) of the vowel pronunciation in words with variant vowels. Accuracy at the final exposure for a subset of 12 shared words across conditions was analyzed to explore the effects of reading skill (i.e., typically developing versus dyslexic), condition, word frequency, and vowel pronunciation (i.e., high- vs low-frequency vowel pronunciation) as predictors in the model.

Results clearly indicate that children with dyslexia differed from typically developing children in the faciliatory and inhibitory effects of reading words repeatedly that support high- versus low-frequency grapheme-phoneme correspondences. Facilitative and inhibitory effects of training, but only in the high-frequency vowel pronunciation condition and only for the typically developing children, were reported. In the children with dyslexia, no differences were observed between words that matched the training corpus and words that did not, but instead just a main effect favoring high-frequency versus low-frequency vowel pronunciation conditions and high-frequency vowel GPC pronunciation. The finding suggests that children with dyslexia, unlike their typically developing peers, fail to benefit from corpus feedback where sublexical features are the key component, supporting the hypothesis that children with dyslexia may be processing only partial information from words by placing insufficient attention on individual letters or groupings of letters and the corresponding phonological representations. Consistent with previous findings (e.g., Ehri & Saltmarsh, 1995) these results suggest that children with dyslexia need extended support to develop the sublexical orthographic to phonological connections needed to benefit from sublexical feedback across word exposures.

The Denton et al. (2020) study represents an interesting test of the utility of multifactorial models of dyslexia to inform instruction. In a previous study, Denton and colleagues (Tamm et al., 2017) evaluated the efficacy of providing the following treatments: (a) ADHD treatment alone, (b) intensive reading intervention alone, and (c) their combination on ADHD symptoms and word reading skills in Grades 2–5 children with ADHD and word reading difficulties. Results did not support an additive effect of combining ADHD and word reading treatments on either ADHD symptoms or word reading ability; instead children with ADHD and reading disability benefited from specific treatment of each disorder. Although it can certainly be reasoned that increasing on-task behavior should improve intervention efficacy the study was unable to demonstrate this relationship. In the current study, Denton et al. (2020) combined a reading and self-regulation intervention to explore whether the addition of self-regulation training would improve the reading outcomes of Grades 2–5 children diagnosed with a reading disability. This study was certainly informed by multifactorial models given the fact that many students with significant reading difficulties also tend to have impaired self-regulation. In the study, the self-regulation component consisted of instruction and activities designed to support a growth mindset, emotional self-regulation, and self-regulated strategy use, and it included training in the use of positive self-talk, goal-setting, and self-monitoring. Results of the study indicate that the combined reading and self-regulation intervention was no more effective than the typical intervention provided by schools that did not include self-regulation training. Together these studies are not supportive of the very seductive idea that multifactorial models can positively inform reading interventions for children with serious reading difficulties. With this in mind, it still seems appropriate to develop reading interventions that maximize student involvement, time on task, and positive corrective feedback to increase intervention efficacy.

Finally, multifactorial models of dyslexia would suggest the need to develop screening methods that sample from a broad range of skills that influence word reading and spelling development. However, the Fletcher et al. (2020) study indicates that a limited number of phonemic awareness, letter sound, and word reading items can be used to effectively screen for risk of developing dyslexia and general reading failure. At first glance this certainly seems at odds with a multifactorial model of dyslexia, but upon further reflection it is consistent with the idea that while there may be multiple paths leading to poor reading and spelling skill those measures that best predict individual differences in word reading and spelling skill (e.g., phonemic awareness, letter sound, and word reading) may be sufficient to adequately capture risk on a brief screening measure. The primary symptoms of dyslexia are poor word reading and spelling skill and while poor lexical retrieval skill (measured by rapid automatized naming) may contribute to development of the symptom it may not contribute additional predictive power over and above other measures for identifying children at risk of dyslexia (see Fletcher et al., 2020). In addition, it is important to note that in English a majority of children who struggle to learn to read have problems with word reading (see Capin et al., 2020) and therefore a set of predictors specifically designed to identify children at risk of developing dyslexia will additionally capture a majority children at risk of general reading difficulties (90% of the Capin et al. sample).

This was an interesting set of papers to evaluate the utility of a multifactorial framework of dyslexia as it relates to issues of etiology, identification, and intervention. Certainly, there is a growing literature supporting a model of dyslexia in which multiple cognitive/linguistic risk factors act probabilistically; suggesting different developmental pathways to dyslexia. However, there appears to be a disconnect between multifactorial developmental pathways (i.e., different etiologies) that result in significant word reading and spelling difficulties from an explanatory perspective versus what should be done instructionally for those at risk of developing dyslexia. Poor word reading skill is overwhelming due to phonologically based decoding problems that result in poor readers paying insufficient attention to sublexical orthographic-phonological relations. There is no evidence that interventions can adequately bypass the typical mechanism associated with orthographic learning, suggesting there is ultimately one path to fluent word reading and spelling skill. The key here is that orthographic learning in alphabetic orthographies requires the conscious binding of orthography to phonology to create amalgamated representations that are recognized instantly without the need for top-down influences. An instructional focus on promoting knowledge and use of sublexical orthographic-phonological relations to promote orthographic learning and word reading development treats the symptom of poor word reading and spelling by supporting the only viable pathway to efficient word reading and spelling. A similar argument can be made regarding early identification; identifying those measures that best predict the symptom may be adequate to gauge future risk. All of this is not intended to denigrate the importance of multifactorial models of dyslexia in any way, but instead to remind us all that many of the important underlying factors that lead to poor reading and spelling skill may not necessarily be informative of the solution.