Serve and Return

Welcome back. Lesson 2 ended with a claim: ordinary family life is what the growing brain expects. This lesson is about the single most important pattern inside that ordinary life. Researchers at Harvard's Center on the Developing Child gave it a name that stuck: serve and return.

The idea is borrowed from tennis. Your baby serves: a look, a coo, a reach, a squawk, a cry. You return: you look back, you say something, you pick them up, you make the sound they just made. Then they serve again. That's it. That's the whole mechanism. It is happening dozens of times an hour in every home with a baby in it, and almost nobody notices they're doing it.

Over the next few minutes we'll look at what a serve actually looks like at two months, at what happens inside a baby when the return doesn't come (there is a famous experiment about exactly this), at why these exchanges predict language and brain development better than how many words a baby hears, at the voice you already use that draws a return out of a baby, and at what phones do to the pattern.

And we'll spend real time on the finding I most want you to leave with. Researchers who film parents and babies together, second by second, find that even happy, well-attached pairs are out of sync most of the time. Not some of the time. Most. What builds the brain is not staying in sync. It's coming back.

Let's start with what counts, because parents often picture serve and return as something that begins when a baby can point or say a word. It begins much earlier.

A serve is any bid for you. A newborn locking onto your face is a serve. A two-month-old's coo. A four-month-old's squeal when you walk into the room. A reach. A cry. Crying is the loudest serve a baby has, and answering it counts. Later come the pointing, the held-up toy, the babble aimed straight at you. A return is any answer at all: eye contact, a word, an imitation of the sound they just made, naming the thing they're looking at, picking them up.

Here's what the research adds. Ann Bigelow's lab followed babies from the first week through three months, and found that somewhere around two months, babies start making bids on purpose: they smile or vocalize while looking right at you, even when you're not doing anything. And the babies who did this most were the ones whose parents had been answering their sounds most consistently in the weeks before. In other words, a baby learns to serve by being returned.

So the first thing serve and return teaches is not language. It's something underneath language: I can make things happen. When I reach out, the world answers. Every ordinary answer you give in the first months is building that expectation, and the expectation is what everything else grows on.

Now, what happens when the return doesn't come? In the 1970s, Ed Tronick and his colleagues designed an experiment to find out, and it has been repeated in well over 100 studies since.

It works like this. A parent plays with their baby, face to face, for a couple of minutes. Then, on a signal, the parent goes still: a blank face, no words, no expression, still looking at the baby but not responding. Two minutes of that. Then the parent comes back and plays again.

Babies as young as two or three months notice within seconds. They smile at the blank face. Then they try harder: they coo, they reach, they wave their arms, sometimes they point at something as if to say, look at that. When none of it works, they turn away. They fuss. They chew their hands. Their heart rate goes up. Some cry. A 2009 review pooled more than 80 of these studies and found the effect is remarkably consistent: across ages, across countries, across families with different parenting styles. Babies in rural Ecuador react the way babies in urban Germany do.

Two more findings from that review matter here. Babies whose parents were more responsive in everyday life held onto more positive feeling during the blank-face minutes; they seemed to trust that the return was coming. And babies who coped better with the still face were more likely to be securely attached at their first birthday. The baby's reaction is not random distress. It's the reaction of someone with an expectation of being answered, watching that expectation fail.

That's the sobering half of this experiment. The reassuring half is what happens next, when the parent comes back.

When the parent comes back after the still face, most babies recover. They fuss for a moment, and then they re-engage. And it turns out that recovery, not the smooth part, is where the real learning is.

Tronick's lab went on to film ordinary mothers and babies playing together, coding their expressions second by second. In pairs with no problems at all, mother and baby were in matched, positive states only about 30% of the time. About 70% of the time they were out of step: one looked away, one was bored, one was too loud, one missed the cue. And then, over and over, they found each other again. Tronick called it mismatch and repair, and he argued that the repair is the point. A baby who experiences a thousand small mismatches that get fixed learns two things: I can recover, and the people I love come back.

Beatrice Beebe's lab, filming four-month-olds and following them to their first birthday, found something that surprised even attachment researchers: more coordination was not always better. Pairs whose moment-to-moment coordination was unusually high, where every flicker of one face was tracked by the other, were more likely to be insecurely attached a year later, just as pairs with very low coordination were. The healthiest pattern was in the middle. Room to look away. Room for the baby to do something on their own. Then a return.

I want you to hear that clearly, because so much parenting advice pushes toward more: more responsiveness, more attunement, more eye contact. The evidence points to enough, with repair. Missing serves is not a failure of the system. It is the system. What matters is that, over the pattern of a day, you keep coming back.

For a long time, the popular version of early language was about volume: hear more words, learn more words. In the last decade, researchers have been able to record entire days in real homes with a small device in a baby's vest, and the picture has changed.

What predicts language best is not how many words a baby hears. It's how many turns they take: a baby makes a sound, an adult answers within a few seconds, the baby answers back. In a study that followed children for 10 years, the number of these back-and-forth turns between 18 and 24 months explained a meaningful slice of IQ and vocabulary scores at age 10, after accounting for family income and education. The raw word count did not hold up once income was accounted for. The turns did.

The brain research says the same thing. Rachel Romeo's team found that children who had experienced more conversational turns showed more activity in the front left language region of the brain, regardless of income and regardless of how many words they'd heard. And a 2023 study from Patricia Kuhl's lab found that turn-taking measured as early as six months predicted how well insulated the brain's main language pathways were at age two. The effect held after controlling for the total amount of adult talk. It was the exchange that mattered, not the exposure.

Notice what this means for a six-month-old who can't say a word. They are already in the conversation. When your baby babbles and you answer, and they babble again, that's a turn. It counts in the recordings, and it counts in the brain.

So how do you get more turns out of someone who can't talk? It turns out you already know how. You just may have been told it's silly.

Almost everywhere in the world, adults talking to babies shift into a particular voice: higher pitched, slower, more musical, with exaggerated ups and downs. Researchers call it parentese. It is not baby talk in the sense of made-up words. It's real words, real sentences, delivered in a register that babies find irresistible. Studies of full-day home recordings find that babies are measurably more likely to vocalize right after an adult speaks to them in parentese than after the same adult speaks in a flat, grown-up voice. Parentese draws the return.

And it can be taught. In a randomized trial from the University of Washington, parents of six-month-olds were coached three times over the first year: they listened to their own home recordings, got feedback on their use of parentese and turn-taking, and talked about everyday activities that invite conversation. By 18 months, the coached families were taking more turns, and their children knew more words. Family income didn't change the result. A follow-up found the advantage was still there at 30 months, a full year after the coaching ended. A separate long-term study found that parents who used parentese consistently in the first year had children with richer language at kindergarten entry.

So when you catch yourself saying, "Whooo's got the toes? Yooou've got the toes," in that ridiculous voice, please keep going. That voice is a tool. It exists because it works.

You may have already noticed that the still-face experiment looks a lot like something that happens in every home now: a parent looking at a phone.

Researchers noticed too. Several labs have rerun the still-face experiment with a twist: instead of a blank face, the parent picks up a phone and scrolls for two minutes. In a 2020 study of more than 200 babies, the result matched the original. Less smiling, more fussing, more turning away, more attempts to get the parent back. In an earlier study, babies of parents who reported heavier everyday phone use had a harder time settling back in during the reunion. A study using paired brain recordings found that the synchrony between a mother's brain activity and her baby's dropped while she scrolled, and then came back when she re-engaged.

Now the honest version of what this adds up to. A 2025 meta-analysis of 21 studies found that parents' technology use around young children was associated with slightly lower cognitive scores and slightly more behavior problems. The associations were real, and they were small. Reviews of this research also note that being absorbed in a device for long stretches seems to matter more than brief interruptions. And in the brain recordings, the connection came back when the parent did.

So this isn't a lesson about never looking at your phone. You'll look at your phone. The phone is also how you reach your partner, your mother, your friends, at three in the morning. The lesson is the one from the still face: what matters is the repair. When you look up, look up all the way. Say their name. Answer whatever serve was waiting. The pattern of a day, not the interruption, is what the baby learns.

So here is what to do, and what not to worry about. Look at the four moves on this slide. They come from the same research, and you already do most of them.

Notice the serve. Follow your baby's eyes. What are they looking at? What are they reaching for? A serve is easier to return when you've seen where it went.

Return it. Any answer counts: a look, a word, the same sound back, the name of the thing. "You see the dog. That's the dog." Parentese welcome.

Wait. This one is the least obvious and maybe the most useful. After you return, pause. Give your baby a few seconds to serve again. Babies need more time than adults do to take their turn, and a lot of would-be turns get talked over. The pause is where the back-and-forth becomes a conversation.

Let them end it. When your baby looks away, they're usually done for now, not rejecting you. Let the exchange close. They'll open the next one. That's the room-to-look-away finding from Beebe's lab, in practice.

Now what not to worry about. You don't need to respond to everything, or instantly, and the evidence says trying to can backfire. You don't need to narrate all day; a running monologue is exposure, and turns are what count. You don't need baby sign classes: a randomized trial found infants taught signs learned language no faster than infants who weren't, though signing does no harm and some parents enjoy it. And if you've heard that children in some families hear 30 million fewer words by age three, know that the study behind that number was tiny and that a larger replication did not find the gap. The turn-taking research holds up. The word-gap slogan does not.

Let's bring it together.

Your baby has been serving since the first week. By two months they're doing it on purpose, because you've been answering. When the answer stops, even for two minutes, they feel it, and when it returns, they recover. Filmed second by second, even the happiest pairs are out of sync most of the time; what builds the baby is the coming back. The turns, not the word count, are what predict language and shape the brain. The voice you're embarrassed by is the tool that draws those turns out. And the phone matters less as an interruption than as a question of whether you look back up.

You do not have to be in sync. You only have to come back.

In Lesson 4, we go to the hardest hours: crying, sleep, and the states a newborn cycles through. We'll look at why a baby can't calm themselves yet, and what it means that they're built to borrow your nervous system while their own is under construction. I'll see you there.