The Wire You Lay First
You are six years old, and someone has put a book in front of you. The symbols on the page mean nothing yet. You are about to spend several hundred hours of concentrated effort turning them into sounds, and sounds into meaning, and meaning into something you will one day call reading without thinking about it at all. What nobody tells you, because nobody in the room fully knows, is that the language printed on those pages is not a neutral choice. It is, in neurological terms, the first wire in a circuit that will carry current for the rest of your life.
The policy debate around bilingual education tends to orbit a single question: should children from minority-language homes be taught in their home language or in the dominant national language? Politicians argue about cultural identity. Economists invoke labour-market outcomes. What almost nobody argues about, with any seriousness, is the cognitive architecture question underneath all of it. The language in which a child first cracks the code of reading appears to shape how that child's brain organises written language processing, potentially for decades. That is not a peripheral detail. It is the whole problem.
What "Reading" Actually Builds in the Brain
Reading is not a natural act. Unlike spoken language, which children acquire without formal instruction, reading must be explicitly taught, and the brain has no dedicated region for it. Instead, it borrows. The visual cortex, the phonological processing areas, the angular gyrus: these regions are conscripted into a network that neuroscientists call the "reading circuit." The specific configuration of that circuit varies by writing system, and the differences are not trivial.
A child learning to read in Finnish, a language with near-perfect grapheme-to-phoneme correspondence (one letter, one sound, almost without exception), builds a reading circuit that leans heavily on phonological decoding. A child learning to read in English, with its notoriously irregular orthography, develops stronger reliance on whole-word recognition and lexical memory. A child learning in Mandarin Chinese, a logographic system, recruits visual-spatial processing areas more intensively than either. Think of it less like learning different dialects of the same software and more like running three different operating systems on hardware that has to be reconfigured each time. Research by Ovid Tzeng and colleagues at the National Yang-Ming University, and work from Usha Goswami's lab at Cambridge, has documented measurable differences in the brain regions most active during reading tasks across these populations.
Now introduce a second language. The child who learned to read first in Spanish, a transparent orthography, and then encounters English spelling, is not starting from scratch. They are applying an existing phonological strategy to a system that punishes over-reliance on it. The child who first decoded characters in a logographic system and then meets an alphabetic one faces a different mismatch entirely. Neither transfer is impossible. Both carry friction that pure language-acquisition models do not fully account for.
Two Children, One Classroom Policy
Consider two plausible children in the same school district. Both are seven. Both have Punjabi as their home language. Amara has been read to extensively in Punjabi since infancy and has basic literacy in Gurmukhi script. Rajan's family spoke Punjabi at home but used English-language picture books almost exclusively. When both enter a classroom where reading instruction happens only in English, they are not in the same situation.
Amara arrives with a functioning phonological awareness system already exercised by Gurmukhi, an alphasyllabary with a regular, transparent structure. The research literature, including a widely cited body of work by Virginia Yip and colleagues on cross-linguistic transfer, suggests she will likely transfer phonological awareness skills across the language boundary even though the scripts differ. Her first-language literacy was, in effect, a scaffold. Rajan, despite his oral English being stronger, has no such scaffold. He is building the reading circuit from the ground up in a language whose orthographic irregularities make the task harder than it needed to be.
A policy that treats them identically because they share a home language is not just imprecise. It is missing the actual variable that matters.
What People Get Wrong About "Mother Tongue" Instruction
The phrase "mother tongue education" carries a warm, intuitive logic: children learn best in the language they know. This is true for oral instruction. It is not straightforwardly true for reading acquisition, and conflating the two has cost a lot of children a lot of time. That conflation amounts to a policy failure hiding in plain sight, one that education ministries on several continents have been slow to acknowledge.
The relevant distinction is between oral proficiency and literacy. A child can have deep oral fluency in a language and still face a harder reading-acquisition path in that language if its orthography is opaque. English-speaking children in England and in North American school systems take, on average, roughly twice as long to reach basic reading fluency as their peers learning in Finnish or Italian, controlling for general cognitive ability. That is not a failure of teaching. It is the orthography. Recommending that a child with oral fluency in English learn to read in English first, simply because English is the language they speak at home, may not serve their cognitive development if a more transparent script is available and teachable.
The opposite error is equally common. Immersion advocates sometimes argue that placing children directly into reading instruction in the dominant language is efficient because it avoids the need to transfer literacy skills later. The evidence is unkind to this position. Jim Cummins's interdependence hypothesis, developed through several decades of research, holds that strong first-language literacy actually accelerates second-language literacy acquisition, because the underlying cognitive skills are largely common across languages. You do not learn to comprehend twice. You learn it once, in the language where the learning is most efficient, and it transfers.
The caveat worth sitting with: Cummins's model was developed primarily with Indo-European language pairs. The transfer effects between Arabic and English, or between a tonal Southeast Asian language and French, are less well mapped. Researchers including Ellen Bialystok have pushed the field toward more nuanced accounts of when transfer helps and when structural distance between languages limits it. The policy world has not kept pace with those nuances. It rarely does.
The Consequence Nobody Is Calculating
Literacy is not just a reading skill. It is the medium through which almost every other academic subject is eventually taught. A child who builds a fragile, effortful reading circuit at age six carries that fragility into science, history, and mathematics at age ten. The compounding is real, and it is large. Early literacy gaps in low-income multilingual populations do not narrow with time in most school systems.
So ask yourself this: if the circuit laid down in first reading shapes everything downstream, why is the choice of that first language still being made on the basis of political symbolism or administrative convenience rather than orthographic transparency and the child's existing phonological toolkit?
It is a decision with cognitive consequences that will outlast every policy paper written about it. The debate about bilingual education deserves to be more technical than it currently is. Not colder. The children sitting down with those first books deserve adults who are honest about what is actually being decided when the page is turned.