Every classroom has children who look ready to read, yet when print appears on the page, something slows down. To parents or teachers this struggle can look puzzling: at times it can present as poor intelligence or motivation, and oftentimes, the first guess is that the child has a problem in seeing the text.
That instinct has a long history.
Dyslexia entered scientific language under names such as ‘word blindness’ and ‘congenital word blindness’, labels that made the difficulty sound as if it belonged to the eyes.
Research over many years has since shown that dyslexia is not a defect of eyesight, nor is it a problem of laziness, intelligence, or poor motivation. It is a neurodevelopmental difficulty, classically marked by slow or inaccurate word recognition, and it has been reported across languages and cultures. For decades, the strongest scientific account has focused on phonology — the ability to recognise and manipulate the sounds of spoken language. That insight has been powerful. It has shaped screening, diagnosis, and evidence-based reading instruction.
But it has also become clear that phonology is not the whole story. Dyslexia is not one problem with one cause. It is better understood as the outcome of multiple risk factors that can combine differently across children. A child’s visible struggle with reading may look similar from the outside, but the underlying pathway may not be the same. So, instead of asking whether dyslexia is a language problem or a visual problem, a better question is: for which children do visual-processing abilities matter, and how?
A need for scale
That question cannot be answered by small sample sizes. It needs scale. To study heterogeneity, researchers need large and diverse samples, because hidden profiles do not reveal themselves in small groups of children who share similar backgrounds. And to study vision meaningfully, researchers need measures that are not merely ‘visual’ in a broad sense, but specific to the demands of reading. The relevant question is not whether a child can see clearly; it is whether the child can rapidly encode the identity, order, and spacing of several symbols at once — the kind of visual information that print requires.
Now the challenge is translational. Can a task be built to be brief, reliable, child-friendly, and robust enough for real schools? Can it work with young children who differ in language, exposure, and opportunity? And can it reveal reading risk that conventional language-based screeners miss?
How our study did this
It was these questions that our study sought to answer; these challenges that it sought to overcome. Published in Current Biology, our study takes up that challenge through Multitudes, a digital platform and research system developed by the UCSF Dyslexia Center, San Francisco, to support early reading screening, funded by the State of California to build demographically appropriate screening tools. This platform allowed theory-driven visual-processing measures from the laboratory into California public schools, where they could be tested at scale, in real classrooms, and followed over time.
My colleagues and I used this platform to study a large, socioeconomically and linguistically diverse group of kindergarten and first-grade children. We measured children’s performance on standard language-based reading screeners alongside visual-processing tasks designed to capture how efficiently a child could encode several symbols — letters or letter-like forms — in a brief glance.
We checked whether children clustered into hidden subgroups with distinct patterns of language and visual-processing skills. If dyslexia and reading difficulty arise from multiple risk factors, then the important question is not simply which measure predicts reading for the average child. It is whether different measures reveal different performance profiles.
What the study found
We reported five subgroups, of which one subgroup had strong language scores but weak rapid visual-processing abilities. These children would likely be missed if the screening consisted of only language-based measures, leaving everyone confused as to why these children present with poor reading outcomes a year later.
Another subgroup showed the opposite pattern: weaker language scores but stronger visual-processing abilities, followed by reading outcomes that were better than expected. Their visual strengths may have helped them compensate during the earliest stages of reading. The supporting findings strengthened this interpretation. The visual-processing measures predicted 12-16% variance in later reading outcomes. And, importantly for diverse classrooms, they showed little evidence of bias across home language or socioeconomic status. This contrasted with reading outcome measures themselves, which can carry the imprint of unequal exposure, instruction, and opportunity.
The analysis also identified three more familiar profiles: children who performed broadly high, average, or low across both language-based and visual-processing tasks. These groups matter because they show the expected developmental gradient. Children with strong language and visual skills had the strongest later reading outcomes; children with average profiles showed more typical progress; and children with broad weaknesses showed the greatest risk. In these groups, visual measures did not overturn the language-based picture so much as sharpen it: they showed whether a child’s reading profile was broadly strong, broadly typical, or broadly vulnerable.
The two visually defined groups were powerful because they broke this pattern. Thus, the value of visual measures is not only that they find children missed by conventional measures. It is that they help distinguish broad readiness, broad risk, hidden vulnerability, and hidden strength — distinctions that matter if screening is meant to guide support rather than merely label children.
The takeaway
Thus the study shows that rapid visual encoding can be one pathway into reading risk — and, in some children, one source of strength. Early screening should therefore not only ask whether a child is struggling; it should ask why, so that we move away from one-size-fits-all intervention strategies.
In essence, it shows that some children with strong language skills but weak rapid visual-processing abilities can be missed by conventional reading screeners.
For India’s multilingual classrooms, the finding is both a scientific insight and a public opportunity and the stakes are immediate. In many classrooms, a child’s home language, school language and test language do not fully align. Language-heavy screeners can make two mistakes at once: they can over-identify children whose low scores reflect limited exposure rather than disability, and under-identify children who speak well but struggle with the rapid visual encoding that print demands. Both errors have consequences. One can label a child unfairly; the other can leave a child unsupported until reading failure hardens into shame. This is why language-agnostic measures are not just a technical advance; they are an equity issue.
From lab to classroom
The California study also shows what public investment can make possible. Supported by dedicated State resources, school partnerships and funding, it moved a precise question from the laboratory into real classrooms and to follow children over time.
India has an opportunity to do something equally ambitious: governments, public institutions, and major philanthropic funders can build large-scale, multilingual, school-based research programmes that identify different pathways into reading difficulty and match support to children’s strengths.
Some work has already begun at the Functional Vision Lab at IIT Gandhinagar, to study the developmental rules of visual processing and attention mechanisms that scaffold variability in later cognitive abilities.
To understand heterogeneity, this work will need schools willing to partner, teachers willing to collaborate, parents willing to help researchers understand children beyond a single score, and funders willing to invest in long-term evidence rather than quick fixes.
A child who struggles to read is not showing us the limit of their ability. They are showing us where our tools are still too blunt — and where science must become sharper, fairer, and closer to the classroom.
(Dr. Mahalakshmi Ramamurthy is an assistant professor, department of cognitive and brain sciences, IIT Gandhinagar. mahalakshmi.r@iitgn.ac.in)
