For about three decades, advice to parents about early language has been organised around a number. The number varies by whoever is repeating it, and its origin is a single small study conducted in one American city in the nineteen eighties. It has been printed on posters, built into policy, and used to tell families that they are running a deficit.
This piece is about why the framing was wrong, and what the research actually points at instead. It is a good case study in how one striking finding can outrun its own evidence, which is a subject we treat more generally in what one study can tell you.
What happened to the famous number
The original work followed a small number of families over time, recording speech in the home, and reported a large difference in cumulative words heard between children from different socioeconomic groups. Extrapolated over years, the difference produced a headline figure large enough to be memorable, and memorability did the rest.
Several problems became clear later. The sample was small, and extrapolating from a small sample of recorded hours to a multi year total multiplies whatever error is present. Subsequent attempts to reproduce the gap using larger samples and better recording methods have found substantially smaller differences, and considerable variation within every group, which is the part the headline erases.
And the framing itself carried an assumption: that what matters is the quantity of speech arriving at a child, as though language were rainfall.
What separating turns from volume showed
When researchers were able to record and code long stretches of home audio automatically, it became possible to separate two things that had been bundled together: how much adult speech a child was exposed to, and how many conversational turns they took part in. A turn means an exchange, adult then child or child then adult, within a short window.
Once those are separated, the turn measure tends to carry more of the relationship with language outcomes than the volume measure does. A house full of talk that never lands on the child does less than a quieter house with more genuine exchange.
This is not surprising, given everything else known about how children learn language. Turn taking is what allows an adult's speech to be about what the child is attending to at that moment, which is exactly the condition under which a word can be attached to a meaning.
Contingency is the mechanism people keep rediscovering
The technical word is contingency: the response depends on what the child just did. A contingent response arrives at the right moment and is about the right thing.
Contingency explains several otherwise puzzling findings. It explains why background television does very little for language while a person doing the same commentary does a great deal. It explains why video calls with a responsive relative outperform recorded video of the same relative. It explains why the effect of reading with a child depends heavily on whether the adult talks around the book or reads the words and turns the page.
It also explains why so much specific advice fails to replicate. Techniques that preserve contingency work. Techniques that increase output while reducing responsiveness do not, and can make things slightly worse by turning an exchange into a broadcast.
What quality does and does not mean here
Once you say quality rather than quantity, there is a risk of implying that some families talk to their children properly and others do not, which is both unpleasant and unsupported.
Responsive interaction is not a middle class technique. It is a human default, present across every culture that has been studied, expressed in different forms. Some communities favour direct dyadic conversation with infants; others place children in multi party settings where they participate as observers before participants. Children learn to talk in all of them.
What varies is not whether families do this but how much uninterrupted opportunity they have to do it, which is largely a question of time, work, housing, noise and stress rather than of knowledge or willingness. Framing it as a parental knowledge deficit misdescribes the problem and puts the burden in the wrong place.
Where shared books fit
Reading with young children has a better evidence base than most early language activities, and it is worth being precise about why. A picture book creates a stable shared object of attention, supplies vocabulary that does not come up in the routine of a day, and invites the adult to talk about something other than what is happening right now.
That last property matters. Talk about the here and now is easy and useful; talk about elsewhere and elsewhen is harder and appears to be particularly valuable, because it is the beginning of the decontextualised language that school runs on. The National Literacy Trust and the Education Endowment Foundation both treat shared reading as one of the better supported early practices.
What the evidence does not support is the idea that the number of books read per week is the operative variable. A book read once with a great deal of conversation around it does more than three read in silence.
Why quotas are the wrong instrument
Suppose the research were much stronger than it is, and suppose it did support a daily target. It would still be a poor thing to give a parent, for three reasons.
A target converts a relationship into a performance. Adults who are counting are less responsive, because attention has moved to the count.
A target has no upper bound and no exit. There is no point at which a parent is told that they have done enough, which is a design feature of guilt rather than of guidance.
And a target measures the wrong unit. The thing that appears to matter is the shape of an exchange, and no count of words captures shape. This is the same problem, in miniature, as the one we describe in effect size in plain words: choosing a measurable proxy and then optimising the proxy.
What is left when the number goes
Something less dramatic and considerably more reassuring. Children acquire language through exchange with people who respond to them. Most families do this most days without thinking about it, in the car, at the sink, in the bath, over a book. It does not need to be scheduled, counted or improved.
The families for whom this is genuinely harder are usually facing constraints on time and attention rather than a shortage of information, and the appropriate response to that is not a leaflet. For the developmental picture underneath all this, see how language arrives, and for how wide the ordinary spread is, the range of normal.