Somewhere in your house, probably on a phone, there is an app that will tell you what your child should be doing this month. It is confident. It has a tick. It was built by people acting in good faith, using numbers that are, in their original setting, entirely sound. And read the way a worried parent reads it at eleven at night, it is close to useless.
The problem is not the numbers. It is what happens to a number when it moves from a population to a person.
What a milestone age actually is
When guidance says that children typically do something by a certain age, that sentence is a compression of a distribution. Somebody measured a large group of children, recorded when each of them did the thing, and produced a summary: a median, or a percentile marker such as the age by which most of the sample had got there.
Both of those are descriptions of a crowd. Neither is a prediction about a member of the crowd. If you are told that most people reach a certain height, you do not conclude that a shorter adult has failed at something. In physical growth, this is intuitive. In development, the same statistic is routinely read as a pass mark.
The spread is the information, and it is the part that gets cut
Every summary has a shape behind it. Some things cluster tightly: nearly everybody does them within a narrow window. Others are enormously spread out, with entirely ordinary children arriving a year apart.
Early language is emphatically the second kind. Vocabulary size at two years, across large samples in several languages, varies by an order of magnitude between children who are all developing typically. First words, first combinations, first questions: all of them wide.
When guidance prints the average without the spread, it is not simplifying. It is publishing the least useful half of the finding. The average tells you where the middle is. The spread tells you whether being away from the middle means anything.
Language is not one dial
The second thing that gets flattened is that language is several partly separable systems, and children are frequently uneven across them.
- Comprehension generally runs ahead of production, sometimes far ahead. A child who says little may understand a great deal.
- Vocabulary and grammar develop in related but not identical ways. Some children accumulate many words before combining any; others combine early with a smaller stock.
- Articulation, the physical business of producing sounds clearly, is a different matter again. A child can have plenty to say and be hard to understand, or be crystal clear and say little.
- Social use, knowing how to open, hold and close an exchange, develops on its own timetable and is the part least captured by any tick list.
A chart that reports one number per age has folded all of that into a single line. The unevenness that a parent notices and worries about is often the most ordinary thing in the picture.
Why the ordinary spread is so large
Several reasons, none of them mysterious.
Children differ in temperament, and temperament interacts with observable language. A child who is cautious about performing in front of others may have a great deal of language and produce very little of it where anyone can hear. That child appears in a study as a low scorer.
Measurement is also harder than it looks. Much early language data comes from parent report, because you cannot follow a two year old around with a microphone for a fortnight, and parent report is systematically noisy. Some parents count a sound as a word; others require the full form. The instrument itself widens the distribution.
And children are allocating effort. A child deep in learning to walk often pauses on language, and the reverse happens too. Development is not parallel processing with unlimited resources.
This does not mean nothing ever matters
It would be a poor magazine that used the width of the ordinary range to argue that concern is always misplaced. It is not. Some children need help, help works better when it starts sooner, and the systems that exist to notice this exist for good reason.
What follows from the width of the range is something narrower and more useful: the judgment does not belong to the chart, and it does not belong to the internet. It belongs to somebody who can observe the child directly, more than once, and set what they see against everything else they know about that child.
In the United Kingdom that is what universal health visiting reviews are for, and it is what the progress check within the early years foundation stage is for. Both are designed around the fact that a single observation is weak evidence and a pattern over time is strong evidence.
How to read published guidance without being frightened by it
Three habits make the difference.
Ask what the number is a summary of. Is it the average age, or the age by which most children have done it? Those two are very different and are frequently printed in identical typeface.
Look for the spread, and notice its absence. If a source gives you an age and no sense of the range around it, it has given you half a fact. Sources produced for professionals nearly always carry the range. Sources produced for parents nearly always drop it, on the reasonable theory that ranges are confusing, which is true and is not a good enough reason.
Distinguish a snapshot from a trajectory. Clinicians are interested in direction of travel. A child who is progressing steadily from a later start is a different picture from a child who has stopped or gone backwards, even if the two look identical on one afternoon.
The underlying point
The parenting information industry runs on precision it does not have. Precision is reassuring to publish and profitable to sell, and the width of an ordinary distribution is neither. So the ranges get trimmed, the caveats get cut for space, and a wide, forgiving, thoroughly researched picture of early development arrives at the reader as a series of dates that feel like exams.
If you would like to see how the same compression happens to the underlying research, the mechanics are set out in our pieces on effect size in plain words and why the parenting press gets research wrong. And if you want the developmental picture behind the numbers, start with how language arrives.