Bilingual Brain: One Grammar Engine for Every Language

💡 TL;DR: A new 2026 study from New York University, published in the Journal of Neuroscience, suggests the bilingual brain runs a single grammar engine for all the languages a person speaks, not a separate system for each one. Using magnetoencephalography (MEG) to track brain activity millisecond by millisecond while Spanish-English bilinguals turned singular nouns into plurals, including invented words, the team found the same left fronto-temporal pattern firing about 100 milliseconds after each prompt, in both languages and even for words that do not exist. Grammar, it seems, is computed by rule, not recalled from memory.

One engine, many languages

For decades the intuitive picture of the bilingual brain was two tidy compartments: an English box and a Spanish box, a Vietnamese shelf and a French shelf, each with its own rules. A study released in June 2026 by researchers at New York University pushes hard against that picture. Their conclusion is that the brain does not keep a separate grammar machine per language. Instead it appears to run one shared, abstract grammar engine that every language you speak draws on. As senior author Esti Blanco-Elorrieta, an assistant professor of psychology and neural science at NYU, put it, brains seem to have a single grammatical engine that fuels all of the languages we speak, rather than separate engines for each one.

That single sentence rearranges a lot of assumptions about how we learn, switch and translate between languages. If it holds, the second or third language you pick up is not a brand-new factory built from scratch. It is new material fed into a production line you already own.

How NYU tested it: plurals and made-up words

The paper, titled A Shared Neural Mechanism for Abstract Grammatical Computations Across Languages in Bilinguals, was led by NYU doctoral student Xuanyi Jessica Chen with Blanco-Elorrieta as senior author. The team used magnetoencephalography, or MEG, a technique that reads the tiny magnetic fields produced by neural activity with millisecond precision. That timing matters, because grammar happens fast: the brain has to decide how to combine pieces of a word or sentence in a fraction of a second.

The task itself was deliberately simple. Spanish-English bilingual volunteers saw a singular noun and had to produce its plural form. In English that means boat becomes boats; in Spanish, barco becomes barcos. The researchers mixed in cognates, words that look alike across the two languages, and crucially they also included pseudowords, invented strings such as paple that obey the sound rules of a language but carry no meaning and live in no dictionary. Asking someone to pluralize a word that does not exist is a neat trick, and it is the heart of the study.

Why the invented words are the clever part

If you ask me to turn cat into cats, a skeptic could argue I simply remembered the word cats. Memory, not computation. But nobody has ever stored a plural for paple, because paple is not a word. To pluralize it you have no choice but to apply a rule on the fly. So when the same brain regions light up for paple as for boat and barco, that activity cannot be memory retrieval. It has to be live grammatical computation.

That is exactly what the team found. A left-lateralized fronto-temporal network switched on roughly 100 milliseconds after the cue, and the pattern of activity generalized across languages, across different plural forms, and across real words and invented ones alike. The brain was not looking up an answer. It was running a procedure, and it ran the same procedure no matter which language the word belonged to or whether the word was real at all.

What a shared grammar engine actually means

It is worth being precise about the claim, because it is easy to oversell. The study is not saying that English and Spanish are the same, or that vocabulary lives in one undifferentiated pool. Words are clearly language-specific: you do not accidentally reach for barco when speaking French. What the data point to is something more abstract. The operation of grammar, the act of taking a stem and combining it with structure to make a well-formed plural, seems to be carried out by a shared neural system rather than duplicated language by language.

Think of it like a single calculator that can run sums in any currency. The numbers and symbols differ, but the arithmetic engine underneath is one and the same. The NYU work suggests human grammar may be built from neural computations that sit above any individual language, which fits a long-standing idea in linguistics that beneath the surface variety of the world's tongues lies a common computational core.

What this changes for language learners

For anyone learning a second or third language, the practical reading is encouraging. If grammar runs on shared machinery, then learning a new language is less about constructing an entirely new mental system and more about feeding fresh vocabulary and a few new patterns into a template you already have. This helps explain a familiar experience: the third language often comes faster than the second, because by then the engine is warm and you mostly need to learn what is genuinely different.

It also takes some of the sting out of the myth that some people are simply bad at languages. If the underlying grammar engine is standard equipment, then struggle is usually about exposure, practice and method, not a missing faculty. And because the brain keeps adapting, a point I have written about in this piece on how the brain keeps improving at any age, adult learners are feeding a system that is still very much open for business. The finding was widely reported, including a clear write-up by Neuroscience News.

A translator's view from inside four languages

I make my living moving meaning across English, Vietnamese, Chinese and French, and this result matches what the work feels like from the inside. When I translate, I almost never swap words one for one. I take the grammatical shape of an idea in one language, strip it back to the relationship it encodes, who did what to whom, and then rebuild that relationship using the structures the target language prefers. A single shared grammar engine is a fair description of that experience: the surface forms change, but the structural reasoning underneath carries across.

It also explains why fluent code-switching feels so effortless and why professional Vietnamese translation is more than vocabulary substitution. The hard, valuable part is the grammar: word order, classifiers in Vietnamese and Chinese, gender and agreement in French, tense and aspect in English. A bilingual brain that computes structure abstractly is exactly the brain that can hold four grammars at once and route an idea cleanly from one to another. That is the muscle a human translator trains for years, and it is why certified and specialized work still resists naive machine swapping.

Honest limits and caveats

One study, however elegant, is not the last word. The participants were Spanish-English bilinguals, two languages that share an alphabet and a great deal of structure, so it remains an open question how cleanly the shared-engine picture extends to pairs that are far more distant, such as English and Vietnamese or Chinese, where plurals, classifiers and word formation work very differently. The task also focused on one narrow grammatical operation, pluralization, rather than the full machinery of syntax. And like all neuroimaging, MEG shows where and when activity occurs, not the complete story of how the computation is implemented. The honest summary is that this is strong, well-designed evidence for a shared core, not proof that every grammatical process in every language pair runs on identical hardware.

FAQ

Source: Journal of Neuroscience (JNeurosci)

About the author

I am Dao Huy (Lucas), a professional translator working across English, Vietnamese, Chinese and French (EN to VI to ZH to FR), with more than seven years in medical, legal, financial and academic translation. A study about the bilingual brain is personal for me: switching between four grammars and routing meaning cleanly from one to another is the exact skill this research describes, and it is the work I do every day.

If you need that level of care, I offer professional Vietnamese translation services, certified Vietnamese translation and multilingual localization across four languages. Tell me what you are working on, in medicine, law, finance or anything that has to read perfectly in another tongue, and I will send a tailored quote at daohuy.com.