There is a version of this story that almost every parent is told, and it goes like this. For about a year, nothing. Then, one afternoon, a word. Then more words, faster and faster, until you cannot remember what the house sounded like before. It is a lovely story and it is wrong in the only part that matters, which is the word nothing.
What happens in that first year is not nothing. It is most of the work. The reason it looks like nothing is that almost all of it is invisible from the outside, and the one part that is visible, saying words out loud, happens last.
Sorting the sounds that matter from the sounds that do not
Every spoken language uses a subset of the sounds a human mouth can make, and it draws the boundaries between them in its own way. English treats the first sounds of rot and lot as two different things. Some languages do not. Japanese speakers learning English as adults famously find that particular distinction effortful, and English speakers find distinctions in other languages equally slippery.
Very young infants are considerably more open than adults are. Over the first year, that openness narrows. The child stops treating every possible acoustic difference as potentially meaningful and starts treating the ones their language uses as meaningful and the rest as noise. This is a gain and a loss at once, and it is one of the most robust findings in the field. It happens without instruction, without correction, and without anybody noticing it is happening.
By the time a child says a word, they have already built a filter. The word is the filter becoming audible.
Turn taking arrives before content
Watch an adult with a four month old and you will see a conversation with the words removed. The adult says something. The baby vocalises. The adult waits, then replies to the vocalisation as though it meant something. The baby vocalises again. The rhythm is conversational: one at a time, with pauses in roughly the places pauses go.
This is not the adult being sentimental, or not only that. The structure of conversation is itself a thing to be learned, and it is being learned here, in advance of anything being said. Children who have had a great deal of this kind of exchange arrive at words with the architecture already in place. The content slots into a shape that is already built.
It also explains something that surprises people about the evidence on talking to babies. What appears to matter is not the sheer quantity of speech washing over a child, in the way that a radio left on produces quantity. It is the back and forth: the turns, the responses that follow what the child just did, the pause left for a reply. Evidence summarised by bodies such as the Education Endowment Foundation in its early years work consistently points at interaction rather than exposure.
The moment objects acquire names
Somewhere in the middle of all this, a child works out an idea that is genuinely difficult and that we hand to them without ceremony: that the noise a person makes while looking at a cup is about the cup. Philosophers have written at length about how underdetermined this inference is. A noise made in the presence of a cup could refer to the cup, its handle, its colour, its being on a table, the act of pointing, or the general situation.
Children solve it, and they solve it partly by watching where people are looking. Joint attention, the business of both of you attending to the same thing and each knowing the other is attending to it, is one of the great quiet achievements of the first eighteen months. Naming rides on top of it. This is also why pointing tends to appear around the same time as early words rather than long after them: they are two expressions of the same underlying grasp.
Why the vocabulary spurt looks like an explosion
Many children go through a period where their vocabulary appears to accelerate sharply. Parents describe it as a switch being flipped. Researchers have argued for decades about whether it is a genuine change of gear or an artefact of how we count.
The deflating possibility is arithmetic. If a child adds words at a roughly steady proportional rate, the absolute number added per week rises even though nothing has changed about the process. Ten per cent of forty words is four. Ten per cent of four hundred is forty. From the outside, that looks like an explosion. From the inside, it may be the same steady machinery it always was.
Not everyone accepts the deflating version, and there is a reasonable case that something does change, particularly around the child's grasp that everything has a name. What matters for a parent is the practical implication: the sudden feeling of acceleration is not evidence that you did something right in the preceding fortnight.
Grammar is not taught and is not copied
The single most persuasive piece of evidence that children are doing something active rather than imitating is that they make mistakes nobody said to them. A child who says goed has never heard goed. They have extracted a rule from the language around them and applied it to a verb the rule does not fit. The mistake is the proof of the rule.
What is more, children who previously said went correctly sometimes start saying goed, which looks like going backwards and is nothing of the kind. It is a system being generalised. Correcting it, incidentally, has very little effect. Children reliably ignore direct correction of grammar while absorbing the same information from ordinary exposure. Anyone who has tried to correct a three year old already knows this.
Why the range is so much wider than the leaflet
Charts and apps quote ages. The ages are usually averages or percentile markers drawn from large samples, and they are perfectly reasonable as descriptions of large samples. They become misleading the moment they are read backwards, as though a child who is not at the marker is behind.
Two things get lost in the translation. The first is spread: around any average there is a wide distribution of entirely ordinary children. The second is that language is not one thing. A child may be well ahead on understanding and unremarkable on production, or have an unusually large vocabulary and be slow to combine words. Development is not a single dial.
None of which means that concern is never warranted. It means that the judgment about a particular child belongs to somebody who can see and hear that child over time, which is the whole reason health visitor reviews and the early years foundation stage progress check exist.
What follows from all this, practically
Very little that is prescriptive, which is the point. The evidence on early language is unusually consistent about one thing and unusually quiet about most others. The consistent thing is that responsive, turn shaped interaction is associated with language development, across settings, across income levels and across languages.
The quiet part is everything else. Whether you narrate the shopping, sing, use books at eight months or eighteen, and how many words you produce per hour: the evidence here is thinner, more contested and much less deserving of the confidence with which it is usually delivered. If you want to read more about why that gap between confidence and evidence is so wide, we have written about what a single study can and cannot tell you and about how the parenting press handles research.
Language arrives. It has been arriving for months before you hear it. The first word is a milestone in your life more than in the child's.