"If you don't interact with the children, they won't learn language at all."
Professor Paul Seedhouse, Emeritus Professor of Human Spoken Interaction at Newcastle University
A toddler stands holding an object. He can crawl and he can stand but he cannot quite walk yet. The object itself doesn't matter; it might be a toy, it might be a piece of wood. What matters is what he does next. He looks at it, looks at his grandfather, looks at it again, and hands it over. The grandfather says thanks and hands it back. The child is delighted. He hands it over again. They could keep doing this for half an hour, and they do. For a few weeks, in his small life, it is the most important thing happening anywhere.
A couple of months later, the grandfather tries the same game. The child says no and runs off.
He's done with it. He's moved on.
The grandfather is Professor Paul Seedhouse, Emeritus Professor of Human Spoken Interaction at Newcastle University, and what he was watching, inside his own family, was a developmental window opening and closing in a matter of weeks. The thing his grandson had been gripped by, the simple act of taking turns, is the foundation of every conversation he will ever have. The window for learning it does not stay open indefinitely. Almost no-one remembers being inside it. By the time language arrives, the earlier system has done its work and faded into the background. Paul has spent his career arguing that it is the most important system in human communication, and that we have been looking right past it.
The system that comes before language
Paul calls it the interaction engine, though the term isn't his. The linguist Stephen Levinson coined it in 2006 and brought out a book of the same name last year. The interaction engine is, in plain terms, how a baby communicates before they have any words to do it with. They use everything that isn't language. Gaze. Touch. Facial expression. Gesture. They take turns. They initiate repair when something goes wrong in the exchange. They build short sequences of interaction with whoever will play. They are extraordinarily busy.
The system is not static. As children grow, the engine grows with them, and when language eventually arrives, the two systems learn to work together. Adults are using a version of it constantly without noticing. We just have words sitting on top.
Why the foundation matters
The evidence for why this matters tends to come from cases where it didn't happen. The nineteenth-century German boy Kaspar Hauser, kept in near-total isolation through his childhood and surfacing as a teenager unable to speak, became famous because linguists arrived to teach him language and could not. There have been other cases since. The pattern is consistent. A child denied human interaction in their first years of life cannot, later, simply be given words and brought up to speed. There is no foundation for the words to rest on. The window has closed.
Paul puts it as plainly as it can be put. If you don't interact with children, they won't learn language at all.
Older than language by a long way
One of the things that makes Paul's argument interesting is the scale of it. Levinson estimates the interaction engine goes back around two million years to the early humans. Language, the system we tend to treat as the deeper one, is thought to be no more than a hundred thousand years old. The engine has been doing its work for roughly nineteen times longer than words have existed.
The two systems do entirely different jobs. The interaction engine is social. It is how a child learns their position in a family and in a community. It is how meaning gets made between two people, how one human comes to understand another. Language is something else. Language is for information storage. You can put any quantity of information into language and pass it on, and that is precisely what humans have done with it for the last hundred thousand years. But you cannot pass information to a baby who has no social foundation to put it on. As Paul says, you cannot dump an information system on a baby.
That the two systems are genuinely separate has been shown in unsettling clinical detail. Paul describes a study by the linguist Charles Goodwin of a man who had suffered a stroke that took out almost all of his language. He could say three words. Yes. No. And. His interaction engine, untouched by the stroke, was completely intact. Video evidence showed him carrying out full conversations using those three words and the non-verbal machinery he still had. Two separate systems, working independently, each capable of functioning without the other.
The same engine, everywhere
Languages have fragmented. There are around seven thousand of them now, each shaped by a culture that needed to describe a different landscape, a different climate, a different way of being human in a particular place. The interaction engine has not. There is only one.
Levinson ran a large project comparing ten completely unrelated languages on turn-taking, on the organisation of repair, on sequence organisation. The conclusion was that all of them were running on the same basic system underneath. Cultural variations exist. In some cultures you are expected to decline an invitation twice and accept on the third asking. In some you gesticulate constantly; in others you don't. But the variations sit on top of a shared frame. Paul has taught students from across the world for almost twenty years and the response is always the same: yes, this is how it works for us too.
The cultural changes inside a single lifetime are interesting in their own right. Paul has taught Chinese students for decades. The students who came to Newcastle in the early years used few facial expressions and few gestures. The students who arrive now, in the era of selfies, use far more. The fundamentals stay put. The surface adapts.
It is also, Paul argues, the reason the modern megacity works at all. Pull millions of people from different language groups into the same square mile, and a sociologist in the 1960s would have predicted disaster. Instead, people gesture. They make eye contact. They take turns. The system underneath holds.
What screens actually replace
On screens, Paul is direct. The worry isn't really about the content. The worry is that screen time is interaction time that is no longer happening. Children cannot develop the interaction engine without interacting, and every hour spent in front of a phone is an hour not spent in the back-and-forth that builds it. The link with language delay is well documented. The reason it exists, in his framing, is the displacement, not the device.
He is also sceptical of the tech-industry line that the addictiveness isn't real. Most parents he meets, he says, don't need a study to tell them their child finds it hard to put down.
Covid still casts a long shadow here. Teachers continue to flag the cohort that went through it, children who lost a critical stretch of socialisation during the months when their engines were doing their most active growth. That damage is visible in classrooms now.
What you can actually do
Paul's advice for parents is built around getting yourself onto the floor and seeing the world from a child's height. The specific activity matters far less than the exchange itself.
Peekaboo is a useful illustration because it has been researched since the 1950s, when the psychologist Jerome Bruner first looked at it. The structure is simple. Eye contact. The parent's face disappears behind a cloth. The face reappears with a boo. The child laughs. The bit that matters, Paul says, is the child's contribution. The laugh. The reaching for the cloth. Eventually, as they get older, covering their own face and waiting for you. Without the child's active role, the game is decoration. With it, it is the interaction engine working out loud.
He has his own version. Years ago he developed a game with his own children where he would puff out one cheek. They would stare, transfixed. He would pull a face and the air would move silently to the other cheek. They would stare again. After enough rounds, they would come over to him and press his cheek themselves, waiting for him to transfer the puff to the other side, and then they would press that one. They could keep going for ages. The point of the cheek-puffing isn't the puffing. It's that the child has a role in what happens next, and learns they do.
His advice for the parent of a newborn fits into a sentence. Have fun, and see the world through their eyes.
The Cinderella of communication
Paul has a phrase for the position the interaction engine occupies in modern research. It is, he says, the Cinderella of communication. Language has had the funding, the frameworks, the institutions and centuries of academic attention. What happens before language has largely been treated as the waiting room.
That is starting to change. New digital recording technology, in the form of very small cameras and AI-assisted analysis, is letting researchers look at what babies are actually doing from the first days of life. The findings are arresting. The onset of turn-taking, and of socially directed gaze, can now be observed days after birth. Not months. Days. The old measurements weren't wrong because babies were doing less than we thought. They were wrong because the techniques had been borrowed from research on adults using language, and they didn't fit. A new generation of methods is finally catching up to what babies have been doing all along.
For policymakers, Paul's case is unambiguous. Invest in this. The years before language are the years that decide whether language will arrive at all, and we have spent very little money studying them properly. The tools to change that now exist.
Chocolate or melon
We asked Paul, in a round of quickfire questions, what his grandchildren had taught him that his research hadn't. He told us about the youngest, who is just beginning to use his first words.
After his main meal, Paul's wife asks him a question. Would you like chocolate, or melon? Whatever she puts last is what he chooses. If she names the chocolate last, he wants the chocolate. If she names the melon last, he wants the melon. He almost always ends up with the melon, because she almost always offers it second.
It is, Paul says, the engine doing its work in real time. The child has learned that a question expects an answer. He hasn't yet worked out that a question with two options is asking him to choose. That layer is on its way. The family is watching, and waiting, to see when it lands.
Somewhere between the older grandson with the object he wouldn't stop handing over, and the younger one with the chocolate and the melon, is more or less the whole foundation of human conversation being built. Almost none of us can remember being inside that period. We're all still using what we learned there, every day, in every exchange we have. Paul's argument is that it deserves a great deal more attention than it has had.
Listen on YouTube, Spotify, Amazon Music, Apple Podcasts and more.