Dr. Patricia Kuhl, one of the world's leading scientists of early language, ran the experiment every parent of a baby should know about: same lessons, same words, same person, delivered live or on a screen. Only one of them worked.
Dr. Patricia Kuhl is co-director of the Institute for Learning & Brain Sciences at the University of Washington, and a world authority on early language acquisition: the question of how a baby who arrives knowing no language at all manages, within a couple of years, to crack one of the most complex codes humans have ever produced. Her laboratory has spent decades mapping what is happening inside infant brains as they listen, and her findings have shaped how scientists everywhere think about the first years of language.
That standing matters, because Kuhl is not someone with an opinion about screens who went looking for a study to support it. She set out to answer a precise scientific question about how babies learn the sounds of a language, and arrived at one of the most striking results in the whole screens-and-children literature almost as a by-product.
The experiment at the centre of Kuhl's work, which she describes in her TED talk The Linguistic Genius of Babies, is beautifully simple. Babies growing up in English-speaking homes were given sessions of exposure to Mandarin Chinese, a language they had never heard, delivered by a live tutor who played with them, read to them and talked to them in Mandarin. Afterwards, Kuhl's team tested whether the babies' brains had learned to tell apart the phonetic distinctions of Mandarin: the fine differences between speech sounds that the language uses and English does not. They had. After a remarkably small amount of live exposure, these babies discriminated Mandarin sounds in the way babies raised hearing Mandarin do.
Then came the part that should be printed on the box of every "baby learning" video ever sold. Kuhl ran the same exposure again with new groups of babies, except this time the tutor appeared on a television screen, or was reduced to an audio recording. Same language, same material, same warm human voice. The result was not weaker learning. It was no learning. The babies who watched the video or heard the audio performed as if they had never encountered Mandarin at all. The single variable that separated learning from nothing was whether the human being was actually in the room.
The source
Patricia Kuhl, The Linguistic Genius of Babies, TED · Read the original
Kuhl's explanation is that the social brain gates language learning. A baby's brain does not treat speech as a stream of data to be absorbed from any source. It takes statistics on the sounds it hears, but it only seems to open the gate and do that work when the sounds come wrapped in a live social exchange: eye contact, shared attention on the same object, a person who responds to the baby and can be responded to. A screen can carry the sounds perfectly. What it cannot carry is the relationship, and for a baby, the relationship appears to be the mechanism. A live human is not one delivery channel among several, faster or slower than the alternatives. On this evidence, the live human is how the learning happens at all.
The first implication is liberating rather than alarming. If you have ever felt a pang of guilt that you are not buying the right language apps or playing the right enrichment videos, Kuhl's work suggests the entire question is misframed. For babies, there is no right video. The most sophisticated language-teaching technology that exists for a child under two is sitting in your chair. You speak it fluently, you respond in real time, and your face does things no screen can do: it looks where your baby looks.
The second implication is about what to protect. The finding does not mean a glimpse of television will harm a baby, and Kuhl's experiment was about learning, not damage. The honest reading is narrower: time a baby spends with a screen is time the language gate is mostly shut, while time spent in face-to-face exchange is time it is open. The currency is shared, responsive moments, and they are cheap. Narrating the washing-up, naming what is in the shopping basket, the absurd repeated drama of peekaboo: these are not filler before the real education starts. On Kuhl's evidence, they are the real education.
It is also fair to say what the study does not show. It was an experiment about phonetic learning in babies, in one language, over a defined period. It does not prove that older children can learn nothing from screens; the picture for a four-year-old watching well-made television is genuinely more mixed, and video calls with a responsive grandparent are a different thing again, precisely because they are social. But for the youngest children, the direction of the evidence is unusually clear, and it points at people.
Picture the scene Kuhl's research keeps describing, transplanted from a laboratory to a sitting-room floor. A father sits opposite his one-year-old with a basket of wooden animals between them. He holds up the cow. She looks at the cow. He looks at the cow, then at her, then says the word. That tiny triangle, two faces and an object, is shared attention, and it is the exact social machinery the Mandarin babies had with their live tutor and lost the moment she went behind glass. The baby reaches, he hands it over, she drops it, he says gone, she shrieks, he says the word again. To an adult it barely counts as an activity. To the infant brain, the gate is wide open.
This is why simple, solid toys earn their keep in the first two years: not because wood teaches vocabulary, but because a toy with no screen, no speaker and no script leaves a silence that an adult naturally fills. A stacking ring becomes a naming game. A block becomes on, off, up, gone. A toy held at eye level keeps two faces pointed at the same thing, which is where Kuhl's evidence says the learning lives. At Jaques London we have been making toys built for exactly this kind of two-person play since 1795, longer than any other games company in the world, and the quiet pieces in our wooden toys collection → are quiet on purpose. A toy that talks does the talking for you. A toy that says nothing hands the conversation, and with it the language, to you and your child.
None of this requires a programme or a timetable. The tutor in Kuhl's experiment was not doing anything a parent cannot do. She was simply there, in person, paying attention to the same thing as the baby. That turns out to be the technology.
Kuhl tells the story of the Mandarin babies herself, with the brain images to go with it, in The Linguistic Genius of Babies, and it remains one of the most watchable introductions to infant science ever recorded. For the wider case that live exchange beats screens, read why real play beats screens, see how her finding fits the broader evidence in the science, or meet more researchers like her in the Expert Library. If you are working out house rules for a baby or toddler, our screen time guide puts this research into ages and stages.
For the phonetic learning Kuhl measured, the answer in her experiment was no: babies exposed to a new language by video or audio recording showed no learning, while babies given the identical material by a live person learned readily. That does not make the videos harmful in themselves, but it does mean that for a baby, screen-based language products are best treated as entertainment, not education. The teaching claim is the part the evidence does not support.
Kuhl's interpretation is that the social brain gates language learning. Babies are superb statisticians of speech sounds, but the analysis seems to switch on only inside a live social exchange: real eye contact, an adult who follows the baby's gaze, attention shared on the same object, responses that depend on what the baby does. A recording carries the sound but cannot respond, and for the infant brain that difference appears to be decisive rather than marginal.
Not necessarily. Kuhl's screens were recordings, which cannot react, and the heart of her explanation is responsiveness rather than the screen itself. A grandparent on a video call who answers, repeats and plays along is socially contingent in a way no recording can be, and that is a reasonable, widely shared basis for treating video chat differently from television. Keep it interactive, keep an adult alongside, and treat it as a conversation rather than viewing.
More than silence, much less than a performance. The exchanges that open the language gate are short, ordinary and repetitive: naming the toy she is holding, echoing her sounds back, peekaboo for the fortieth time. You do not need a large vocabulary or constant commentary, you need shared attention in person, little and often, over a shared toy or a nappy change.
Britain's oldest toy maker. Every toy open-ended, wooden, and built for exactly the kind of play the research recommends.