The Thirty Million Words That Weren't There
By the age of three, poorer children have heard 30 million fewer words. It is the most quoted number in early education, and it came from 42 families and about an hour of tape a month. What has replaced it is less frightening and far more useful at a kitchen counter.
Somewhere in the past twenty years you have probably met this number: by the age of three, a child from a poor family has heard 30 million fewer words than a child from a well-off one. It appears on nursery walls and in government campaigns. Charities have been built on it.
It is one of the most quoted findings in education. It also came from 42 families in Kansas and about an hour of tape a month.
That doesn't make it a lie. But what happened when other researchers went looking for the same gap is worth knowing, mostly because the finding that has replaced it is far more useful to a parent standing at a kitchen counter.
So in the 1980s they began recording families: one hour a month, for two and a half years, 42 families in all. They counted the words each child heard, found large differences between the professional families and those on benefits, and multiplied those hourly rates out across the first four years of life.
That is where 30 million comes from. Not from counting 30 million words, but from scaling up an hour of tape. Published in 1995, the figure gave a moral argument something it had never had before, which was a number.
The difference was in who counted as talking. Hart and Risley had mainly counted words a primary caregiver directed at the child. Sperry's team counted everyone within earshot: grandparents, older siblings, cousins, the adults talking over the child's head.
Once all those voices were in, the tidy class gap largely dissolved. Some of the poorest children heard more words than better-off ones. The variation within each income group was wider than the gap between the groups.
The paper caused a proper row. Roberta Golinkoff and a group of fellow language researchers replied in the same journal, pointing out that Sperry's sample contained no genuinely affluent families, and warning that denying the gap would cost poor children the help they need. That argument deserves respect.
So here is the honest version. Children do turn up at school with very different language, and those differences track family income from early on. What is shaky is the specific number, and the assumption that volume is the thing doing the work.
They tracked the words children heard, the words children said, and the number of times child and adult went back and forth in a real exchange.
Only the back-and-forths mattered. Children who had more of them showed stronger activity in the part of the brain that puts speech together, and that held whatever their parents earned or however long they had stayed in education. The sheer quantity of words heard did not predict it.
A study in Pediatrics found something similar over a much longer run. Researchers recorded 146 babies and toddlers across six months, then tested the same children between the ages of nine and 14. The number of back-and-forth exchanges at 18 to 24 months predicted a sizeable share of the differences in their language and thinking a decade later, with income and parents' education accounted for.
A turn is a small, ordinary thing. Your toddler points at a puddle. You say something about the puddle. They offer something half-formed back, and you answer that. Five of those beat a beautifully narrated trip round the supermarket where nobody waits for a reply.
When YouGov asked more than a thousand teachers about this in March 2025, 44 per cent of primary teachers said up to a fifth of their class could not communicate at the level expected for their age. Asked what was behind it, the most common answer, from 58 per cent, was less conversation at home.
There is something encouraging in a trial reported this year. A team at the University of Otago in New Zealand trained early years staff in what they call serve and return: follow whatever the child is interested in, ask something that can't be answered with yes, then wait.
More than 1,400 children across 136 centres took part. By the age of three they were ahead on oral language and early self-control, and the gains showed up soonest for children from the poorest homes. The approach worked best during book reading and nappy changes, which is to say during the ordinary parts of the day.
Turn-taking asks for the opposite, and it is harder. You cannot take a turn while you are still talking. You have to stop and leave a gap wide enough for a two-year-old to build a sentence, which can take an uncomfortably long time.
Most of us fill it. We finish their sentence. Or we answer our own question while glancing at the phone. What a child needs in that moment is not more words coming at them. It is the floor.
It is one of the most quoted findings in education. It also came from 42 families in Kansas and about an hour of tape a month.
That doesn't make it a lie. But what happened when other researchers went looking for the same gap is worth knowing, mostly because the finding that has replaced it is far more useful to a parent standing at a kitchen counter.
Where the Number Came From
Betty Hart and Todd Risley were psychologists chasing a puzzle. Children who did well in a preschool programme they ran seemed to lose those gains once they left. They wondered whether the answer was at home.So in the 1980s they began recording families: one hour a month, for two and a half years, 42 families in all. They counted the words each child heard, found large differences between the professional families and those on benefits, and multiplied those hourly rates out across the first four years of life.
That is where 30 million comes from. Not from counting 30 million words, but from scaling up an hour of tape. Published in 1995, the figure gave a moral argument something it had never had before, which was a number.
What Happened When Someone Checked
In 2019, a team led by Douglas Sperry published a re-examination in the journal Child Development. They pooled home recordings from five American communities, from very poor rural areas to comfortable middle-class suburbs.The difference was in who counted as talking. Hart and Risley had mainly counted words a primary caregiver directed at the child. Sperry's team counted everyone within earshot: grandparents, older siblings, cousins, the adults talking over the child's head.
Once all those voices were in, the tidy class gap largely dissolved. Some of the poorest children heard more words than better-off ones. The variation within each income group was wider than the gap between the groups.
The paper caused a proper row. Roberta Golinkoff and a group of fellow language researchers replied in the same journal, pointing out that Sperry's sample contained no genuinely affluent families, and warning that denying the gap would cost poor children the help they need. That argument deserves respect.
So here is the honest version. Children do turn up at school with very different language, and those differences track family income from early on. What is shaky is the specific number, and the assumption that volume is the thing doing the work.
The Bit That Actually Travels
Around the same time, Rachel Romeo and colleagues at MIT and Harvard measured something different. They recorded four, five and six-year-olds at home for two days, then scanned their brains while they listened to stories.They tracked the words children heard, the words children said, and the number of times child and adult went back and forth in a real exchange.
Only the back-and-forths mattered. Children who had more of them showed stronger activity in the part of the brain that puts speech together, and that held whatever their parents earned or however long they had stayed in education. The sheer quantity of words heard did not predict it.
A study in Pediatrics found something similar over a much longer run. Researchers recorded 146 babies and toddlers across six months, then tested the same children between the ages of nine and 14. The number of back-and-forth exchanges at 18 to 24 months predicted a sizeable share of the differences in their language and thinking a decade later, with income and parents' education accounted for.
A turn is a small, ordinary thing. Your toddler points at a puddle. You say something about the puddle. They offer something half-formed back, and you answer that. Five of those beat a beautifully narrated trip round the supermarket where nobody waits for a reply.
Why It Matters Here, Now
Speech and Language UK puts the number of children in the UK currently struggling to talk and understand words at 1.9 million, roughly one in five. Ofsted inspectors have described more children arriving in reception who cannot easily make themselves understood, many of them born in the lockdown years.When YouGov asked more than a thousand teachers about this in March 2025, 44 per cent of primary teachers said up to a fifth of their class could not communicate at the level expected for their age. Asked what was behind it, the most common answer, from 58 per cent, was less conversation at home.
There is something encouraging in a trial reported this year. A team at the University of Otago in New Zealand trained early years staff in what they call serve and return: follow whatever the child is interested in, ask something that can't be answered with yes, then wait.
More than 1,400 children across 136 centres took part. By the age of three they were ahead on oral language and early self-control, and the gains showed up soonest for children from the poorest homes. The approach worked best during book reading and nappy changes, which is to say during the ordinary parts of the day.
Harder Than Talking
The 30 million figure caught on because it behaved like calories. Something you could count and fall short of. Plenty of parents came away believing they were supposed to be generating words, as though quiet were a kind of neglect.Turn-taking asks for the opposite, and it is harder. You cannot take a turn while you are still talking. You have to stop and leave a gap wide enough for a two-year-old to build a sentence, which can take an uncomfortably long time.
Most of us fill it. We finish their sentence. Or we answer our own question while glancing at the phone. What a child needs in that moment is not more words coming at them. It is the floor.
Sources & Further Reading