Identifying and Serving Students With Learning Disabilities,Including Dyslexia,in the Context of Multitiered Supports and Response to Intervention

Historical Context and Initial Intent of RTI

Legal protections for students with LD under the Individuals with Disabilities Education Improvement Act (IDEA, 2004) include their access to a free and appropriate public education, which guides an individualized education plan, and which is delivered in the least restrictive environment. The IDEA definition of specific learning disability (SLD) is familiar to readers of JLD; it describes the nature of the disorder involving one or more psychological processes involved in language and how it manifests as academic difficulties (i.e., listening, thinking, speaking, reading, writing, spelling, or doing mathematical calculations). The definition includes specific conditions, such as dyslexia, and it excludes other conditions: learning problems stemming from sensory or motor disability, intellectual disability, or emotional disturbance and environmental, cultural, or economic disadvantages.

It has now been 16 years since the reauthorization of the IDEA, which allowed states and local education agencies to consider students’ RTI rather than just relying on IQ (intelligence quotient)–achievement discrepancy-based approaches to identify students with LD. Historically, RTI implementation was intended to address many concerns about discrepancy approaches used to identify students with LD and determine their eligibility for services expressed by researchers, policy-makers, parents, and students themselves (e.g., Bradley & Danielson, 2004; Fuchs & Fuchs, 1998; Torgesen, 2000; Vellutino et al., 1996).

One set of concerns about the discrepancy approach to identification was that students had to “wait to fail” far enough behind in academics to qualify for interventions, despite converging research evidence supporting the efficacy of earlier prevention relative to later remediation. Some students who experience word reading and spelling difficulties, struggle with mathematical word problems, or have very poor writing skills may still meet grade-level expectations and not be eligible for LD services under IDEA. Another set of concerns was that this approach yielded inaccurate identification (i.e., over- and underrepresentation), particularly given the history of systemic inequities among schools that confound identification of students with LD by excluding “disadvantages,” including poverty, particularly for vulnerable children who are racially, ethnically, linguistically, or culturally diverse (e.g., Hanushek & Rivkin, 2009; Harry & Klingner, 2006; Hosp & Reschly, 2004; Morgan et al., 2015; Skiba et al., 2008). A further set of concerns was that the assessment data from discrepancy formulas were not timely or detailed enough to guide instruction, monitor growth, and intensify interventions (cf. Vellutino et al., 1996). A common theme weaving through these concerns was whether students had received adequate academic instruction in the first place and that without the “label” students did not receive intensive intervention. Although RTI stemmed from special education legislation, the intent was to improve academic performance within both general education and special education.

This Special Series

This special series will be split across two issues. The first article provides a broad view of guidance from state education agencies (SEAs) about MTSS, RTI, and identification. Berkeley and her colleagues (Berkeley, Scanlon, Bailey, Sutton, & Sacco) update their prior JLD-published systematic review of SEA websites to explore changes in implementation guidance for RTI and MTSS (Berkeley et al., 2009). One trend they note is that following the shift in the Every Student Succeeds Act of 2015 from using the term RTI toward MTSS, many states followed suit. However, confusion has arisen because a few states use RTI and MTSS interchangeably. Encouragingly, most SEA websites communicate guidance or initiatives that support tiered systems of support. The authors also report that most states have also passed specific dyslexia legislation or early screening and intervention processes. Some of these processes are consistent with RTI, but they may involve separate practices geared toward the path to identify a child for dyslexia services (i.e., without an LD label), and, therefore, without IDEA protections. A majority of states rely on additional formal testing, including data about student response for identification for LD or dyslexia.

In the second article, Miciak and Fletcher describe the critical role of instructional response for identifying dyslexia and other LD. They delineate a series of attributes for dyslexia classification to argue for a “hybrid” method that incorporates data from school-implemented MTSS to support valid decision making and guidance about individuals’ instructional needs. Their article summarizes definitions of dyslexia, comparing and contrasting classifications models. Miciak and Fletcher also highlight challenges inherent in identifying dyslexia solely through documentation of inadequate response related to different measures and application of different cut-points and methods for defining response. After describing the etiology of dyslexia, the authors emphasize the critical importance of early identification and intervention to focus on mastering the alphabetic principle, automatic word reading, and spelling. They propose a hybrid model with three prongs: low achievement, assessment of instructional response, and examination of contextual factors and other disorders. Data collected regarding potential exclusionary attributes, such as sensory disorders or second language acquisition, must include cultural effects and language (spoken language and the language of instruction). Miciak and Fletcher argue this assessment type conducted within MTSS can reduce costs for comprehensive assessments and more directly inform intervention iteratively. They note there is little evidence for dyslexia-specific interventions, as opposed to intensive treatments within MTSS models, which may limit access to intervention for some children with other types of LD.

In the third article, Wagner and colleagues (Wagner, Zirps, Edwards, Wood, Joyner, Becker, Liu, & Beal) provide a new approach to estimating the prevalence of dyslexia. They discuss the challenge of examining a continuum of reading performance among poor readers to identify “unexpected” reading difficulty for individuals relative to better performance in language or academic areas. The authors describe evolutions in definitions of dyslexia and challenges related to (a) limited agreement among various operational definitions based on cut-points for a single construct (e.g., IQ–achievement discrepancy, inadequate RTI, and poor decoding or fluency) and (b) a lack of longitudinal stability. Like Miciak and Fletcher, Wagner et al. propose a hybrid type of identification model, which they describe as a constellation of symptoms and indicators: impaired phonological processing, genetic risk, and influences from the environment, including the quality and quantity of instruction and intervention. They selected a proxy for dyslexia, to reflect unexpected reading difficulty, which was operationalized as poorer reading comprehension relative to listening comprehension. Model-based meta-analysis was used to estimate (a) the developmental relationships between reading and listening comprehension across widely used standardized assessments and (b) the extent to which age moderates the relationship. A simulation generated distributions of the prevalence of dyslexia that varied according to severity. Their results suggest that a sample of poor readers is more likely to contain expected poor readers versus unexpected poor readers on the basis of their reading and oral language scores. Furthermore, findings suggest that individuals with low reading scores relative to language scores are not isolated to a particular region in the distribution of reading scores but instead are found throughout the distribution of reading scores. An important conclusion is the need for science to inform public policy and practice.

In the fourth article, Odegard and his colleagues (Odegard, Farris, Middleton, Oslund, & Rimrodt-Frierson) explore a policy and practice problem of lower-than-expected rates for dyslexia identification and eligibility. Their data also point to the elementary years as the time when more students are identified as having dyslexia, middle school as when the percentage tapers off, and high school as when far fewer students are identified. Odegard and colleagues explore a large data set with universal reading screening data for more than 8,000 second graders who attended 126 schools in one state to provide context and explanations for these problems of practice. By second grade, children had experienced kindergarten and first-grade English language arts instruction. The authors used screening data to identify profiles related to dyslexia (scoring below benchmark primarily in reading fluency and/or spelling), or with a mixed profile (also showing below-benchmark performance on secondary vocabulary and comprehension). They then examined whether students with these profiles, derived from the universal screener data, were associated with actual school-assigned dyslexia classification. They report that 33.76% of students exhibited characteristics of dyslexia on the screener but only 7.26% of students were classified with dyslexia and were therefore eligible for intervention. In schools with relatively higher proportions of students identified as at risk by the screener, there was a greater likelihood that individual students were “missed.” Because these students were not identified by their schools as having dyslexia, they were not eligible for intervention. The authors caution that schools were less likely to classify African American and Hispanic students as having dyslexia than Caucasian students, controlling for both literacy skills and free and reduced-price lunch status, despite data from the universal screener. This finding underscores the need to address why these vulnerable students were excluded from intervention support, which is likely to only widen their achievement gap relative to their peers. These children already face historical systemic inequalities and bias related to race and ethnicity within schools and outside of schools. As a field, we need to acknowledge these findings and advocate against intentional or unintentional bias and for increased resources to vulnerable students who attend underperforming schools.

In the fifth article, the series shifts to identification of LD, beginning in the area of math, and specifically for young Spanish-speaking students. Cho and colleagues (Cho, L. S. Fuchs, Seethaler, D. Fuchs, & Compton) argue that it is challenging to distinguish limited English proficiency from low mathematics performance. It is important to accurately identify English language learners with LD in math so these students can receive effective intervention. Cho and colleagues conducted a study to examine whether (a) first graders whose first language was Spanish would perform better on a dynamic math assessment than on traditional predictor assessments and (b) there would be differences in language of testing (Spanish vs. English) based on the students’ dominant language. They also compared the predictive validity for end-of-year mathematics assessments of both dynamic assessments (DAs) and the type of outcome (calculations vs. word problems). The study involved 17 Title I schools. In this study, the authors classified these students’ language dominance based on teachers’ reports as either Spanish-dominant (speaks mostly or only Spanish) or English-dominant (speaks mostly English or Spanish and English equally well). Students were randomly assigned to English or Spanish DAs. Classroom teachers delivered mathematics instruction in English for an average of 230 min per week. Research staff conducted assessments on static mathematics measures and domain-general (language, reasoning) measures in English, and they conducted DAs in the assigned language (Spanish vs. English). At the end of the school year, they assessed students’ calculation and word-problem solving outcomes in English. Study findings supported the validity of Spanish DAs for students who were Spanish-dominant, when predicting calculation outcomes. The English DAs showed stronger prediction for the word problems, regardless of language dominance. Important implications for our field of LD are considering language dominance to inform assessments and realizing that DAs can predict outcomes and inform intervention targets within MTSS.

The second half of this special series will be published in our next JLD issue. These remaining articles shift to RTI in reading in upper-elementary grades. In this sixth article of the series, Vaughn and her colleagues (Vaughn, Capin, Scammacca, Roberts, Cirino, & Fletcher) investigate whether individual differences in initial word reading performance predicts fourth graders’ response to reading interventions, with the aim to inform design for even more powerful interventions. They identified students as having reading comprehension difficulties using screening scores at or below 85 on a standardized reading comprehension test. These students were classified into three clusters of word reading proficiency—very low, low, and near adequate—based on their pretest performance on three standardized measures of timed and untimed word reading plus spelling. A majority of the sample (68%) was Hispanic, 22% were African American, 2% were Caucasian, and 8% were classified as “Other.” The students were randomly assigned to either a researcher-delivered small-group reading intervention that targeted vocabulary, text reading in social studies content, and decoding instruction or to a comparison condition, which varied according to what schools provided. Vaughn and colleagues found that very low word reading (e.g., on average timed word-reading standard scores of 61 and untimed word reading standard scores of 71) was an important predictor for minimal response to treatment, arguing that this “significant marker” may indicate these students need an even more robust, individualized, and specialized intervention to show meaningful progress in reading.

In the seventh article, Hendricks and Fuchs also examine RTI, through a secondary analysis of two randomized control trials involving fourth- and fifth-grade students. Their sample included students with adequate word reading (i.e., average timed word reading at a standard score of 95), which was slightly higher than the Vaughn et al. samples in the near-adequate word proficiency cluster. However, Hendricks and Fuchs’ students had relatively weaker reading comprehension (scoring at or below the 50th percentile, with the resulting average standard score of 90). Their sample was diverse: 30% were Hispanic, 44% were African American, 19% were Caucasian, and 7% were classified as “Other.” Slightly more than half (56%) of students received free or reduced-price lunch and 4% were identified as receiving special education services. The students were randomly assigned to (a) two variants of a reading comprehension, researcher-delivered small-group intervention that also focused on social studies content or (b) a comparison condition that varied according to what schools provided. Hendricks and Fuchs were also interested in understanding pretreatment predictors of response, but they examined different response classifications (growth vs. final status) and explored both standardized and researcher-developed measures of comprehension that ranged from near- and mid-transfer to far-transfer. The authors report that students were more likely to be classified as responsive across all methods and measures if they had higher pretreatment scores on expressive vocabulary, nonverbal IQ, teacher ratings of attention, and reading comprehension. In contrast to the Vaughn et al.’s studies and samples, Hendricks and Fuchs note that word reading was not predictive of poor treatment response. Students who began the study with relatively higher pretreatment scores were more likely to be classified as responsive by final status than by growth methods. Students with lower pretreatment comprehension scores were more likely to be identified as responsive using growth methods. The authors argue that without an evidence-based consistent definition of “response,” inadequate response is arbitrary, dependent on the methods and measures. They conclude—with caution for our field—that RTI on its own may not be a valid approach for identifying students with disabilities.

In the last article for this special series, Burns and colleagues (Burns, Maki, Brann, McComas, & Helman) used a quasi-experimental approach to compare reading growth among three groups: students who received interventions, typically achieving students, and students who received special education for SLD. The study involved six urban schools and focused on second- and third-grade students. A majority (81%) of students were eligible for free or reduced-price lunch programs and 65.2% were African American, 22.3% were White, 11.2% were Native American, and 1.3% were Hispanic. Students who scored at or below the 10th percentile on a computer-adaptive universal screener at the beginning of the year, and who were not already receiving special education services, were classified as the severe reading deficit group. These students were assigned to one of several Tier II small-group foundational reading interventions that included phonics or fluency; research staff were interventionists. Students who scored at or above the 50th percentile were classified as the typically achieving group; they received regular reading instruction from their classroom teacher. Students receiving special education for SLD and scoring at or below the 10th percentile were classified as the special education group; they received reading intervention from their special education teacher. The research team administered curriculum-based measures of oral reading fluency to all three groups as part of the school’s benchmark assessments. Burns and colleagues found that students in the severe deficit group who received Tier II interventions had similar reading trajectories to typically achieving peers and demonstrated significantly higher growth rates than students in the special education group. Although students in the severe deficit group improved their reading fluency, they did not show enough acceleration to catch up to their typically developing peer group. These findings suggest students with severe reading deficits with or without an SLD identification will need ongoing, intensive small-group interventions.

We opened this introduction with the historical context and initial intent of RTI and special education, and we now close with a reflection on how the field may move forward in light of the work represented in this special series. Collectively, findings from these studies demonstrate the complexity of accurately identifying students with LD and of accelerating growth for students without the label but who have severe reading and mathematics difficulties. If we consider where we are within the context of RTI/MTSS implementation science (e.g., Fixen et al., 2005), most states recommend universal screening and multitiered intervention. Thirty states presently have dyslexia-specific legislation that includes screening requirements. For test-makers, there appears to be promise for hybrid models of LD identification that incorporate data from RTI as well as other models, such as Bayesian ones, that may provide more focused information on the probability of having an LD. For test-takers and test-administrators, the added burden of additional assessment takes its toll. Local education agencies therefore may wish to prioritize existing support systems for educational interventions, such as RTI/MTSS, by which dyslexia screening and instruction may also occur. It is vital for these systems to address the needs of culturally diverse students and to consider the COVID-19 pandemic’s effect on the socioemotional well-being of vulnerable children (e.g., Ford, 2012).

The authors of the articles in this special series also point out that we do not yet have consensus for adequate versus inadequate response, but this lack of consensus has not deterred them from engaging in meaningful research to support education for all children, particularly those in vulnerable populations, that allows them to improve their academic performance and participate fully in school and in their chosen career and life pathways. As a field, we have a long road ahead of us to improve identification and interventions to support students with LD through prevention and ongoing intensive interventions.