Richard Wagner, PI
Project I: Definition, Classification, and Risk Prediction
Stephanie Al Otaiba
Colin Walsh, Consultant
This project is one of two projects that includes a significant implementation focus, as it seeks to turn multivariate models of dyslexia and specific reading comprehension disability into tools that can make predictions about functionally significant outcomes at the level of the individual. Parts of this project are designated high risk. Artificial intelligence and Bayesian inference will be the bases of two of the prediction models that will be competed along with models based on established methods such as logistic regression.
The search for additional predictors, especially neurobiological ones, will be carried out using a dynamic systems model that will be applied to studies of the co-development of brain and of language/literacy. Results of these studies will have implications for determining whether changes in brain are more likely causes of, consequences of, or mere correlates of changes in language/literacy. More specifically, dyslexia and specific reading comprehension disability (SRCD) are two important public health problems, with estimates of prevalence ranging from 3 to 20 percent for dyslexia and 8 to 10 percent for SRCD. The long-term objective of this project is to substantially increase replicable knowledge about the nature of these learning disabilities and to implement this knowledge in tools that potentially can improve the outcomes of individuals with learning disabilities and their families. Existing definitions of dyslexia or word level reading disability that prioritize a single indicator (e.g., poor decoding, inadequate response to instruction/intervention) show poor levels of agreement and longitudinal stability. However, an operational definition derived from a multivariate model of dyslexia shows substantially better performance by combining multiple indicators.
Project II: Developing Dynamic Forecasting Intervention Algorithms for Children with Severe Literacy Disabilities
Young Suk Grace Kim
One of the success stories in intervention has been the programmatic line of research that began with prospective longitudinal studies in which the effects of different amounts and kinds of instruction were measured. Parameters from these models were embedded in dynamic forecasting intervention algorithms that made predictions at the level of the individual child about how much time should be spent in which kinds of instructions to maximize development of literacy skills. Randomized control trial studies demonstrated that the recommendations worked, and this achievement was first presented in a Science article titled “algorithm-guided reading instruction” (Connor et al., 2007). Turning to children
with severe learning disabilities, there is little available science that addresses the question of
how much time should be spent in which kinds of intervention, and how much time should be given to
assistance in the form of assistive technology. This project will apply the methodology of
algorithm-guided instruction and intervention to answer this question. We view this project, which
will make predictions about recommended amounts and kinds of intervention, as synergistic with
Project 1 which will make predictions about risk and the likely value of assistive technology.
Specifically, too many children fail to achieve proficient reading and writing skills, which has
serious public health and economic consequences. This is because reading and writing difficulties
are associated with grade retention, referral to special education, dropping out of high school,
and entering the juvenile criminal justice system. Moreover, on average, the literacy proficiency
of students with learning disabilities is one standard deviation lower than that of typically
developing students, and children from low socio-economic backgrounds achieve lower literacy
proficiency than their more affluent peers. With funding from NICHD and IES, we have made important
strides in improving literacy outcomes for children through the use of dynamic forecasting
intervention (DFI) algorithms in Assessment-to-instruction (A2i) technology. A2i is a web-based
teaching support technology designed to improve teachers’ effectiveness in delivering
individualized (or personalized) literacy instruction. However, it is not clear how well the A2i
DFI algorithms predict optimal amounts and types of instruction for children whose reading and
writing skills fall at the lowest tail of the continuum, and who are least likely to be responsive
to general education and intensive interventions. Hence, the overarching aim of the proposed
research is to use what we have learned over the past 13 years of developing the DFI algorithms and
A2i and conducting randomized controlled trials to test their precision and efficacy; and to
address the learning needs of children with the most severe learning disabilities including those
with dyslexia and dysgraphia.
Project III: Understanding Risk for Early Language and Literacy Difficulties in Young Children
Leslie Rescorla, Consultant
For too many years, it was believed that reading problems could not be predicted until children were taught to read. Even today, many children with even severe learning disabilities are not identified until second grade or beyond. Each year that a child with a learning disability goes undetected can make problems become more intractable and lead to concomitant problems in attitudes and motivation for learning. Reading and writing involve language in print, and language provides the vehicle for identifying children at risk for learning disabilities well before the beginning of primary school. The potential impact to the field of this project capitalizes on relations between language and literacy to facilitate much earlier identification of children at risk for language and literacy difficulties. This project is also
synergistic with Project 1 in that it will begin the process that could lead to the development of
risk-prediction models that could be used at earlier ages.
Specifically, the proposed research project investigates a broad array of early child, familial,
and environmental correlates and predictors of language and early literacy difficulties and
disabilities to improve our capacity to identify children at heightened risk for these
developmental problems. Although there is significant awareness of the risk to typical language and
early literacy development posed by familial history, early language delay, and socio-demographic
and environmental influences, there remains a sizable gap in the ability to identify which specific
children with one or more known risk factors will achieve normative developmental milestones by
school entry versus which children will remain at or accrue high-risk status and be well below
typical skill levels at school entry. Earlier identification of high-risk children is crucial to
prevent the development of learning disabilities or to minimize their duration and severity. We
focus on a group of children at known risk, namely 2- year olds with a familial history of language
impairment, reading impairment, or both, to explore how to maximize the sensitivity and specificity
of early predictors of learning difficulties at age 5. Once selected from within a large screened
sample, we will trace the development and experiences of these 250 children across four years. Results of the proposed investigations have the potential to improve early identification of learning disabilities by identifying combinations of key predictors and add to knowledge regarding the interplay of language, cognitive and early literacy skills in a critical early developmental period.
Project IV: Reading and Math Co-Development in a Diverse Sample of Twins
Elliot Tucker-Drob, Consultant
Using an ingenious system for identifying twins from an existing state-wide database that contained longitudinal data of reading and reading-related variables, a diverse sample of thousands of twins was collected in a previous funding period. The large-scale twin sample enabled the detection of effects that would not have been possible with typical twin studies. An example is the finding that teacher quality moderates genetic effects on early reading, a finding that also was published in Science (Taylor et al., 2010). The potential impact to the field of this project is replicating the method on a much larger sample that includes both reading and math data, thereby enabling the study of genetic and environmental etiology of co-occurring problems in both domains at a more fine grain level than has been possible for most studies.
Specifically, reading and math problems represent an important public health issue for children in
that they are associated with various negative outcomes including school failure, limited occupational success, and juvenile delinquency (Geary et al., 2001; Reynolds et al., 2001). Of US fourth-grade students, one-fourth fail to reach even partial mastery of grade-level knowledge in reading, and one-fifth fail to reach partial mastery of grade- level knowledge in math (NCES, 2015), highlighting the prevalence of reading and math difficulties in childhood. Given we know that children who struggle in reading often also struggle in math, it is important to identify influences on the development of both reading and math. The overall goal of the proposed research is to uncover salient factors, including genetic and environmental influences, which contribute to the co- development of reading and math performance, at a critical developmental point (elementary school). We will identify the first nationally-representative US twin sample through the proposed National Project on Achievement in Twins (NatPAT). The NatPAT sample will comprise 7,668 pairs of twins located across the US, and will be uniquely situated to address the overall goals.
Project V: Co-Development of Skills Associated with Learning Difficulties/Disabilities in Monolingual and Bilingual Children
Learning disabilities do not develop in isolation but, rather, in the context of developing (a) achievement in other domains, (b) self-regulation, and (c) for some children, developing knowledge of English as a second language. This potential impact to the field of this project is that it will be the first to apply the state-of-the-science method of latent-change-score and parallel-process-growth modeling to study influences among reading, math, self-regulation, and language status, in the development of learning disabilities and difficulties, and to examine how text and reader characteristics affect Spanish-speaking English-learners’ construction of meaning from text.
Specifically, the successful acquisition of reading and math skills during elementary school
represents one of the most significant educational achievements of early education. These skills
provide the foundation for acquiring knowledge both in school and throughout life. Many children
acquire these skills early and maintain them at a high level throughout school. A significant
number of children, however, struggle with reading skills, math skills, or both throughout their
school experiences, resulting in learning difficulties and disabilities. Two factors that may affect development and co-development of reading and math are self-regulation and non- mainstream
language environments (i.e., children from homes in which a language other than the societal
language is spoken). Whereas prior research has demonstrated associations between these two factors
and problems in reading and math, including learning disabilities in these domains, few studies
have employed longitudinal designs that allow a determination of the nature and direction of these
associations. Consequently, the overall goals of this project are to examine issues related to the
development and co-development of learning disabilities and difficulties in reading and math both
in the context of self-regulation and in the context of nonmainstream language environments across the early-elementary-school years using longitudinal-study designs that allow better determination of the nature of the associations.
Project VI: Imaging genetics in SRD: Mega- and Meta-analyses
Laura Almasy, Consultant
Susan Bouregy, Consultant
Michael Milham, Consultant
The neurobiology of reading disability has been a popular topic of study using both neuroimaging
and genetic methodologies. However, replicable results have been difficult to obtain, partly
because of the relatively small sample sizes of most studies. It has not been feasible to image
large numbers of individuals with reading disability, and doing so multiple times in a longitudinal
study has not been practical. Genetic influences are turning out to be multiple and small as
opposed to few and large, requiring large samples for the neuroimaging research and genetic research has been carried out independently without regard to findings from the other domain. The potential impact to the field of this project is that it addresses these limitations by combining imaging and genetics (i.e., imaging genetics), and using meta-analysis to combine results across primary studies.
The goal of the Administrative Core is to help a well-integrated and productive Center become even better by ensuring the efficient and productive functioning of the research activities and the maximal utilization of cores, consultants, and advisors.
Serves as the central dissemination and translation, resource sharing, and training component of the proposed FLDRC.
Serves to provide statistical analysis assistance to the Projects in order to support their efforts to evaluate the project aims and hypotheses
David Kaplan, Consultant