10 ROAR Dyslexia Screening and Subtyping
Beyond measures of reading skills, many state dyslexia screening initiatives require additional predictors of dyslexia. Dyslexia is defined in terms of reading skills (Maggie Snowling and Hulme 2024; O’Brien and Yeatman 2021; Catts et al. 2024; Lyon, Shaywitz, and Shaywitz 2003; Fletcher et al. 2006; Elliott and Grigorenko 2024), but there are two reasons why additional predictors are useful:
- Improved prediction accuracy: Even though dyslexia is defined in terms of reading skills, that does not mean dyslexia begins in school. There are a collection of (potentially but not necessarily causal) factors that can be measured prior to formal reading instruction that predict future risk for dyslexia. There are also skills that can be measured at the earliest stages of reading instruction that are useful for prediction and early risk assessment. Thus, from a practical standpoint, including early predictors of dyslexia in a screener improves sensitivity and specificity (irrespective of the causal relationship between the predictors and reading development).
- Dyslexia subtyping and individualized intervention: Dyslexia research has embraced multifactorial models where dyslexia is considered a probabilistic outcome that emerges from the combined influence of a collection of risk factors and protective factors (O’Brien and Yeatman 2021; Catts et al. 2024, 2024; B. F. Pennington 2006; Compton 2021; Zuk et al. 2021; van Bergen, van der Leij, and de Jong 2014; Wolf and Bowers 1999). Even though phonological awareness (PA) is still considered one of the most important mechanisms underlying dyslexia (Margaret Snowling 1998; M. J. Snowling, Hulme, and Nation 2020; Wagner and Torgesen 1987; Stanovich 1998), the field of dyslexia research has broadly reached consensus that a collection of mechanisms beyond difficulties with PA confer risk for dyslexia. This new understanding of dyslexia embraces heterogeneity: not every student’s struggles are the same and knowing the root of a student’s struggles could help plan the most efficacious intervention. However, currently, the notion of personalized intervention for particular dyslexia subtypes is more of an aspiration than a reality. One of the challenges is that the measures that are useful for phenotyping are not necessarily the most effective intervention targets. For example measures of Rapid Automatized Naming (RAN) and Rapid Visual Processing are prime examples: even though a wealth of research has established that these measures are useful for prediction, and can identify different profiles of struggling readers, it has not been established how intervention should differ based on these profiles.
We organize the ROAR Dyslexia Screening and Subtyping battery into screening measures that target:
- Foundational Reading Skills that indicate dyslexia Section 10.1 and are established intervention targets
- Dyslexia Prediction and Subtyping which includes measures of various mechanisms that are hypothesized to be causally related to dyslexia, improve screening sensitivity and specificity, but aren’t established intervention targets Section 10.2.
ROAR has been validated as a dyslexia screener beginning in the Spring of kindergarten. Many of the individual measures are valid earlier (as young as 4 years of age (Gijbels, Burkhardt, and Ma 2024)), but validation of ROAR dyslexia screening measures in the Fall of kindergarten is ongoing.
10.1 Dyslexia screening based on foundational reading skills
People with dyslexia struggle learning how to read. The most direct way to screen for difficulties consistent with dyslexia is to use measures of foundational reading skills, which is why these skills are included in the ROAR Foundational Reading Skills Suite. Children with dyslexia are delayed relative to their peers in the development of: (1) Phonological Awareness (PA; see Chapter 7), (2) Letter Sound Knowledge (see Chapter 8), (3) real word and pseudo word reading (see Chapter 5), and (4) reading speed, efficiency or fluency (see Chapter 6). Difficulties with all these skills can persist through adulthood without proper identification and intervention. The goal of screening based on foundational reading skills is to a) identify challenges early in in elementary school and b) intervene while the developing brain is optimally plastic and interventions are most effective (Gaab and Petscher 2022; Blachman 2013; Torgesen 2004, 1998; Lovett et al. 2017). For example, Lovett et al. (2017) has shown that intervening early is more efficient (larger effects per hour of intervention) compared to waiting until later in elementary school. To quote Torgesen (1998), the goal of early screening is to “catch them before they fall”.
10.2 Dyslexia prediction and subtyping
Wolf and Bowers (1999) and colleagues first introduced the “double deficit hypothesis” as a response to the “core phonological deficit hypothesis” and demonstrated that a) some children with typical PA skills still struggle learning to read and b) many of these struggling readers have early challenges with Rapid Automatized Naming (RAN) (Denckla and Cutting 1999; Denckla and Rudel 1976; Wolf and Bowers 1999; Compton, DeFries, and Olson 2001; Wolf et al. 2002). Additionally, children who struggle with PA and RAN tend to have even larger challenges with reading than those who only struggle on one skill. RAN is now required by most dyslexia screening legislation. However, RAN is not a useful intervention target per se: whereas a child who struggles with PA will benefit from training targeting PA (Bradley L and Bryant P E 1983), a child struggling with RAN does not simply need to practice RAN. Each of the measures in the ROAR Foundational Reading Skills battery is a core component of reading development and a useful intervention target. Measures in the Dyslexia Prediction and Subtyping battery are predictive of reading development and often help to understand the mechanisms underlying a student’s struggles, but are not proven intervention targets. For example, RAN (see Chapter 11) is highly predictive of future reading development, and also indicates a different type of struggle than PA (i.e., automatization or connectivity between visual and verbal processing), but is not a skill that needs direct instruction. Rapid Visual Processing (see Section 10.2.1) is another measure that has mounting evidence of a causal relationship to reading development and has utility as a screener but is not a skill that should be directly taught (Ramamurthy, White, and Yeatman 2024a; Ramamurthy et al. 2024; Lobier and Valdois 2015; Marie Line Bosse, Tainturier, and Valdois 2007).
10.2.1 Rapid Visual Processing
10.2.1.1 Theoretical background
There is a long-standing controversy about the visual factors associated with dyslexia. While phonological difficulties have been widely recognized as a core feature of dyslexia, debate has persisted over visual processing theories, suggesting a more complex, multifaceted etiology (R. Pennington and Pennington 2011; Vellutino, Fletcher, and Snowling 2004; O’Brien and Yeatman 2021). The field has yet to reach consensus because existing theory on visual processing differences in dyslexia is often built upon small samples, using experimental measures without established psychometric properties, that are deployed across different age ranges often without prior work demonstrating the validity of the measure in each developmental window. Validated measures of visual processing that predict future reading development could contribute in overcoming the challenges faced by other screening measures since visual development is language-agnostic and not directly taught in preschool.
Rapid Visual Processing (RVP) refers to the ability to rapidly encode and recall multiple visual elements simultaneously in a brief glimpse (Sperling 1960, 1983). In this task, a string of letters or symbols is briefly flashed at the center of the screen and then the participant is cued to report the identity of a single randomly chosen element. Of all the measures of sensory processing that have been studied in relation to word reading difficulties, the rapid visual processing task has the strongest evidence for identifying a subgroup of struggling readers who are not captured by conventional measures of phonological awareness. Previous studies have demonstrated that this task:
- Consistently correlates with reading ability (Marie-Line Bosse and Valdois 2009; Ramamurthy, White, and Yeatman 2024b).
- Differs in children with dyslexia (Lobier, Zoubrinetzky, and Valdois 2012; Ramamurthy, White, and Yeatman 2024b).
- Cannot be explained as a consequence of dyslexia (Lobier and Valdois 2015).
- Might be a useful intervention target (Valdois et al. 2014; Zhao et al. 2019; Zoubrinetzky et al. 2019).
- Identifies a subset of poor readers that have high phonological awareness despite their reading difficulties (Saksida et al. 2016; Valdois, Reilhac, and Ginestet 2020).
- Recent evidence suggests that the difficulties with Rapid Visual Processing in children with dyslexia are consistent across languages including French, English, Dutch and Chinese (Huang, Liu, and Zhao 2021; Lobier and Valdois 2015) making this an ecologically relevant measure related to reading across different languages.
By including both letter and symbol stimuli, the task allows for the assessment of rapid visual processing skills both within and independent of language experience. The measure’s language-agnostic nature makes it particularly valuable for early screening, as visual processing issues precede reading difficulties. This approach, of including visual measures that are linked to reading, aligns with emerging multifactorial models of dyslexia, contributing to a more comprehensive understanding of the various cognitive factors influencing reading development (O’Brien and Yeatman 2021; Catts et al. 2024; B. F. Pennington 2006; Compton 2021; Zuk et al. 2021; van Bergen, van der Leij, and de Jong 2014; Wolf and Bowers 1999).
10.2.1.2 Structure of the task
The Rapid Visual Processing Task was developed in close collaboration with the UCSF Multitudes project. Task design, data collection and analysis was shared across the two projects.
ROAR-RVP has 2 versions that measure the same construct of rapid visual processing: Rapid Visual Processing with Letters (RVPL) and Rapid Visual Processing with Symbols (RVPS). RVPL measures the ability to rapidly locate and identify letters in 2-, 4-, and 6-letter strings. RVPS assesses the ability to rapidly locate and identify non-namable visual symbols, making it language-agnostic. These tasks are considered promising tools for early identifying of struggling readers not captured by conventional phonological awareness measures.
- Letters (RVPL)
- Symbols (RVPS)
ROAR-RVP employs a six-alternative forced choice task presented as an engaging underwater adventure game. There are two versions, each with two difficulty blocks (2- and 4- element strings. 6-element string can be added for older students). Participants help a lost dolphin (RVPL) or whale (RVPS) find friends and treasures. The task sequence involves fixating on a central point where 2-4 elements (letters or pseudo-letters) briefly appear (240ms). A post-cue then indicates the target position, and participants select the correct element from six choices. Each version begins with 6 longer-duration practice items. Narrated instructions, encouragement animations, and feedback sounds guide participants throughout. The game is designed for K-2 children but can be extended to older students and adults.
Figure 10.1 shows the rapid visual processing task with letters (RVPL) a string of letters is briefly flashed at the center of the screen (240ms) and then the participant is cued to report the identity of a single randomly chosen letter. The participant’s task is to report the identity of the target element that was at the prompted single location, by tapping on one letter from a set of 6 letter choices provided. This task was optimized to provide highly reliable metrics of rapid visual processing in children as young as five years of age (see Section 22.2)
10.2.1.3 Are visual measures biased to social factors like one’s socio-economic status and primary language?
While different cognitive and language processes have been increasingly reported to be influenced by socioeconomic status and childhood experience Schwab and Lew-Williams (2016), in theory, visual processing measures should not be influenced by language, socioeconomic status or early childhood experience. This is because most sensory processes have a developmental trajectory that precedes Stein (2022) formal reading instructions, making them valuable screening measures for identifying potential future reading challenges.
We first examined whether the visual measures we developed exhibit bias related to students’ primary language or socio-economic status as indexed by eligibility for free and reduced-price meals (see Figure 10.2). This investigation addresses a fundamental challenge in early screening: many measures show bias across various social factors. Consequently, even when measures demonstrate good predictive power for reading challenges, it often remains unclear whether this predictive ability stems from the measure assessing a construct fundamental to reading skill or simply from capturing the same differences in social factors that are known to influence reading outcomes. Our findings show that ROAR-RVP were not affected by English language proficiency or socioeconomic status, which shows promise for addressing a major challenge in the field.
