In 1993 two ten-year-old boys led a toddler away from a shopping precinct near Liverpool and killed him. The case broke something in England. Within five years the country had scrapped the old presumption that a child between ten and thirteen might be too young to tell a real wrong from mere naughtiness. It left the age at which a child can be tried at ten, where it sits today — the joint-lowest in Europe. A child had done something monstrous, and the law's answer was to make children answerable sooner.1
Thirty years on, that answer is spreading. In February 2026 Argentina cut the age at which a child can be tried from 16 to 14; Sweden, unnerved by gangs that arm children as gunmen, is lowering its own for the gravest crimes; last month Northern Ireland refused to raise its age from ten. The pressure comes each time a child does the unthinkable — a twelve-year-old stabs her mother to death, a fifteen-year-old shoots a man running for president — and each time it pushes the number down.2
This piece is prompted by the strongest case against that wave, made this summer by The Economist: do not lower the age, because punishing children is unjust and, in any case, does not work. It reaches the right conclusion, and argues it well — which is why it is worth taking apart rather than cheering. The strong form of an argument is where you can see what the whole debate, its own side included, has quietly agreed not to ask.
A number is a strange answer to the thing being asked. When a child does grave harm, three questions arrive at once and we treat them as one. Is she to blame — did she, in any sense the law can use, choose a wrong? Is she old enough to be blamed — is her mind formed enough to carry it? And does blaming her help anyone? Three questions, three different kinds of answer, one birthday made to hold all three. When the first grows too hard to argue in public, the argument reaches for whichever of the others can be made to look like science.
What a child knows
The one of the three that science can truly speak to is the second, and it complicates the slogan. “Incapable of evil” — the old phrase for a child too young to try — locates the missing thing in what the child knows. But children tell a real wrong from a broken rule by the age of three or four. Shown a rule against hitting and a rule against wearing pyjamas to school, they judge the first wrong even where no rule exists and no adult is watching; the second, only against the rules.3 They know. What comes later is not the knowing but the power to act on it — foresight, the brake on an impulse, the nerve to refuse an older boy with money. A ten-year-old can know exactly that what he is doing is wrong and be, at that moment, the person least able to stop himself.
That gap is why the science cuts both ways. When America's Supreme Court spared juveniles the death penalty and mandatory life, it leaned on the unfinished adolescent brain, the reward-seeking that matures before the restraint.4 Humane rulings. But an immaturity that spares a child today can be measured differently tomorrow, and a scan describes a thousand brains, not the boy in the dock; it cannot say whether this child could have done otherwise. Reach for the image to settle one child's guilt and you are no longer letting science inform the law — you are asking it into the judge's chair, which it cannot fill.
What works
So the argument moves to the third question, where the footing feels surer: never mind blame — does treating children as criminals work? But “work” hides a swap. To the man whose garage the same boys keep emptying, locking them up works perfectly well: while they are inside, his garage is safe. What the evidence actually says is narrower — that punishment does not cut future offending, that where the age drops gangs recruit younger, that the earlier a child is drawn into the machinery of justice the likelier he is to come back to it.5 Both are true, and they are not the same claim. The practical case — the one the Economist leans on hardest — wins the argument about reoffending rates and quietly drops the one the burgled man was making, which was never about reoffending rates. It is the ground almost everyone can share, because standing on it means never having to say whether the child was to blame, or what is owed to the man whose garage he emptied.
It is also where the best model of prevention sits. Treat violence not as a wickedness to be repaid but as a hazard to be designed out, the way a century of engineering cut deaths on the roads.6 As prevention it works, and its appeal is that you can count it. But the model gives way where you lean on it hardest.
The programmes with the strongest evidence are the ones that keep the wrong inside them. Focused deterrence — a plain message that the shooting is wrong, a credible threat to stop it, and a way out for anyone who takes it — has the best record in the field; Boston's version was followed by a two-thirds fall in youth homicide.7 It bundles condemnation, sanction and help together, and no trial has pulled them apart, so no one can say for certain which does the work. But the model built to manage without the condemnation — the shooter is not at fault, only his circumstances — has the thinner and shakier record. That is not proof that the blame is the ingredient; it is a reason to distrust a model that designs it out. And a model that ranks its measures by how little they depend on anyone choosing puts the person who chose last of all, which is a strange place to build a courtroom.
What no number reaches
That courtroom answers to something the prevention model never sees. Beneath it lies the oldest reason we punish, older than deterrence and untouched by any of this. Before the state, a killing was answered by the dead child's kin; the long labour of law was to take that answer out of their hands — the blood-price, the king's peace — so that a killing ended in a court and not a feud. The state's monopoly is less on violence itself than on who may rightfully avenge.8 Remove that office and the demand it holds does not vanish. It goes back to the family.
Here the science is real, and it discomforts both sides. Revenge does not heal: people handed the chance to punish end up in a worse mood than those refused it, and dwell on the wrong longer, and every one of them expects the opposite.9 Yet the impulse is real, and old, and everywhere. The court's task was never to hand down a satisfaction that never comes; it is to answer the demand in the open, under rules, so it is not answered in a stairwell instead. And whether a given child's act should be answered at all, and with what, is closer to what the family is asking — and no study, no scan, no reoffending rate reaches it. It is not the kind of question a measurement was built to take.
What we owe
That limit is not a licence for feeling over evidence. The innocent are convicted by unchecked certainty — the confident witness who is wrong, the forensic method that was never sound — and it is science, DNA at its plainest, that overturns them.10 The line to hold is narrow, not a wall: science for the facts — did he do it, is his mind formed, what keeps the next child out of the gang — and not for the last question, of what he deserves and what he is owed. That one was never in the instrument.
So why do we keep handing it to the instrument? Because the honest answer is dear. Everyone in the argument already knows what a child in that dock is owed — a childhood safe enough that the gang was not the best offer he ever got — and few of us want to pay for it. “The science is complicated”; “who is to say what evil is”: these are not puzzles waiting on a solution. They are the paperwork a society files to keep from acting on what it already knows, because acting costs money it would rather spend elsewhere. The brain scan wheeled into the courtroom is the grown, credentialed version of a boy who shuts his eyes and thinks he has disappeared.
Lower the age in fear, raise it in pity, and both mistake a choice for a discovery. The number is not waiting in the tissue to be found. It is a line a society draws between what it owes its children and what it will pay for them, and calling the line a scientific finding is how it keeps from admitting the line is a choice. Science can tell us the child knew. What he is owed it cannot — and we go on asking it because we already know, and would rather the answer came from the machine than from us.