Play occupies an odd position in the way we talk about childhood. It is either sacred, invoked as self evidently precious and beyond examination, or it is trivial, the thing children do when nothing important is happening. Both positions avoid the interesting question, which is what play is doing.
The most productive framing, and the one that has generated the most usable research, treats play as a form of exploration conducted under unusually safe conditions.
Exploration is expensive, and play makes it cheap
Any system that has to learn about a world faces a version of the same problem. You can do the thing that has worked before, which is reliable and teaches you nothing new, or you can try something else, which might be better and might be a disaster. The second is how information gets acquired and the first is how you survive the afternoon.
Play is a way of buying the second cheaply. In play, the stakes are suspended. A tower that falls costs nothing. A social role tried on and abandoned costs nothing. The consequences of being wrong are removed, which is exactly the condition under which trying the unlikely option becomes rational.
This framing has a useful property: it predicts the observed features of play rather than merely describing them. It predicts that play should be more common when an animal is safe, fed and unstressed, which it is. It predicts that play should involve variation rather than repetition of the optimal move, which it does. And it predicts that play should decline when circumstances become serious, which is one of the more reliable observations in the whole area.
Object play looks like informal experimentation
Watch a young child with an unfamiliar object and the sequence is remarkably consistent. They will do something to it, observe what happens, do a variant, and continue. If the object behaves as expected, interest fades quickly. If it does something surprising, exploration intensifies and becomes targeted at the surprising property.
Researchers have used exactly this to study early causal reasoning. Give children a toy whose behaviour is ambiguous, and they play in ways that would disambiguate it: isolating variables, trying the parts separately, repeating the confusing case. Give them a toy whose behaviour is already clear and they move on.
This is not a metaphor for experimentation. It has the formal structure of experimentation, conducted without instruction by people who cannot yet tie their shoes.
Pretend play and the cost of holding two worlds
Pretend play is the strangest thing children do and the least easy to explain in simple learning terms. A child treats a block as a telephone. They know it is a block. They act, coherently and at length, as though it is not.
Whatever else this is, it is cognitively expensive. It requires maintaining a representation of what is actually the case alongside a representation of what is being treated as the case, keeping them separate, and generating behaviour consistent with the second while remaining governed by the first. Adults do a version of this when reading fiction, planning, or considering a hypothetical.
The tempting move is to conclude that pretend play trains this capacity, and that more pretend play produces better hypothetical reasoning. That step is where the evidence thins considerably. The association is there. The causal claim has proved much harder to demonstrate than its popularity suggests, and several well known demonstrations have not replicated cleanly.
Rough play, social play and the rules nobody writes down
Rough and tumble play makes adults uncomfortable and is one of the more interesting categories. It looks like fighting and is systematically not fighting: participants take turns being on top, self handicap against smaller partners, and use signals that mark the activity as play.
What appears to be happening is calibration. Children are learning how much force is acceptable, how to read whether a partner is still enjoying it, and how to repair when they have got it wrong. Those are not skills that can be acquired by instruction, because they are continuous and contextual.
Group play with rules is a further step again, and one of the few settings where small children voluntarily submit to constraint. The rules of a playground game are negotiated, enforced, argued over and revised, usually with more procedural sophistication than the adults nearby assume.
Why play does not teach in the way a lesson teaches
Here is where a great many well meaning products go wrong. If play generates learning, the reasoning goes, then play designed around a target skill should generate that skill faster.
The trouble is that the mechanism runs on the absence of a target. The features that make play useful, variation, low stakes and the freedom to pursue the surprising, are the features that a designed activity removes. An activity with a correct outcome is not exploration; it is a task with attractive packaging.
This does not mean adults should stay out of it. Adult involvement that follows the child's lead, comments, wonders aloud and supplies materials appears in most quality frameworks, including the early years foundation stage, for good reason. It means the involvement is a contribution to the exploration rather than a redirection of it.
The evidence problem, stated plainly
Play research is genuinely difficult, and it is worth understanding why so that the confident claims can be discounted appropriately.
You cannot randomly assign childhoods. Studies comparing children who play more with children who play less are comparing families, settings and circumstances at the same time. Short interventions can be randomised, but a six week programme is a poor model for a developmental process that runs for years.
The outcome measures are also awkward. What would count as evidence that play worked? Vocabulary is measurable and only partly relevant. Creativity measures are notoriously unstable. Attempts to measure the things play plausibly develops run into the fact that those things are diffuse.
The result is a field where the theoretical case is strong, the observational evidence is rich, and the causal evidence for specific claims is weak. Saying so is more useful than pretending otherwise, and the general shape of this problem is set out in correlation and cause.
What follows
Not a programme. Play is not something a parent needs to administer, and the market in developmental toys rests on a causal claim the research does not support. This magazine recommends no products, and the reason is set out on our about page.
What does follow is a defence of unstructured time against the pressure to fill it, which we take up in unstructured play is not idleness, and a more respectful reading of the state that produces play in the first place, in boredom and what it is for.