I recently interviewed Una Padel, director of the Centre for Crime and Justice Studies, a research foundation that advocates alternatives to prison and restorative justice. A fortnight before we talked, her 13-year-old daughter had been mugged.

If the muggers could be brought to account (they cannot, even though the daughter knows who they are), what would Padel have in mind for them? True to her principles, she does not want the muggers jailed.

"I remain angry with them, but I don't want anything horrible to happen to them," she said. "I want them to stop robbing people, that's the bottom line . . . In an ideal world I would like them to be made aware of the impact they've actually had on my daughter and, ideally, apologise."

Padel is no dewy-eyed naif. She has dealt with criminals for years and is easily as knowledgable and unsentimental as any judge likely to try the case. It's her own daughter who has suffered the harm.

I want to suggest a thought experiment: if she had the power, would Padel be morally entitled to give the muggers a sentence that does not punish them instead of one that does? I am even willing to stipulate that her sentence inspires genuine remorse in the muggers and that they stop mugging (generous stipulations indeed). Would justice be done if Padel had her way?

The principles of the kind of simple justice I propose today say no. Justice does not consist of successful therapy. It consists of just deserts. The just desert for terrorising a 13-year-old and robbing her must entail punishment, whether or not the muggers feel bad about what they've done and whether or not they will do it again.

Padel and her daughter have the moral right and perhaps a spiritual obligation to forgive the muggers. They do not have the moral right to abrogate the community's obligation to punish wrong behaviour.

The reason I raise this is that in an article in these pages last week I argued that the criminal justice elites in England resist get-tough policies on crime, and that this opposition explains why the practice of criminal justice has changed much less than the flurry of new legislation and policies would lead one to expect.

I further argued that this opposition flows from a progressive view of justice that seems to be almost universally shared among the criminal justice elites a view that sees a "just" sentence as a complicated balancing of the nature of the offence, the particulars of the offender's situation, protection of the rights of the offender and concern with the effect that a sentence will have on the offender.

The progressive view of justice has qualities that people with high IQs tend to like: it is nuanced and uses complicated explanations for human behaviour. It also has the quality that tends to make such theories so disastrous as public policy: what can be put into practice bears no resemblance to the promises of the theory.

The reality is that nobody is smart enough to get justice right when attempting to balance the factors that progressive justice tries to balance.

Consider how hard it is for parents to be sure whether a misbehaving child needs an explanation with a hug, a stern warning or a lesson that won't soon be forgotten. Now compare how much parents know about their own children with how little judges know about convicted offenders who appear before them for sentencing.

The judge has seen the offender for a matter of hours at most, more likely a matter of minutes. He may have read a sentencing report on the defendant, often sketchy, often factually wrong. The judge is then supposed to assess whether the offender feels genuine remorse and whether the offender is likely to reoffend.

How? Everyone who stands before the judge is highly motivated to say whatever the judge wants to hear and be whatever the judge wants him to be. The better the criminal, the better the act.

The judge is supposed to be able to see through all that and accurately assess whether the offender is really, truly sorry for what he did and is going to change his ways. It is an absurd premise. Indeed, even when the offender does feel genuine remorse, that doesn't mean the offender will stop offending. Two other characteristics of criminals are impulsiveness and a short time horizon. Offenders can be really, truly sorry for what they did today and be back on the streets doing it again a week from now. As police and probation officers will tell you, it happens all the time.

The judge is supposed to decide whether the offender will respond better to a community sentence or to prison. How? Individual case histories of offenders reveal every kind of response, from the offender who is set straight by a tough prison sentence to the one who sees the error of his ways by apologising to his victim. A judge has to guess. I know judges prefer to call it the use of judicial discretion, not guessing--but guessing is what it really amounts to. Making matters worse, the guess usually reflects not the unique characteristics of the defendant and the offence, but the judge's personal ideology.

Everything I have said about judges applies equally to the people in the Crown Prosecution Service who are making decisions about whether to plea-bargain, drop charges or bring a case to trial.

For practical purposes, the question of whether the justice system should take personalities, background, remorse or predictions about future behaviour into account when deciding what should be done to an offender is moot. Whether or not it should, it can't. Prosecutors and judges cannot be that smart about the parade of offenders who come before them. Progressive justice doesn't do what it claims to do.

The technical label for the simple alternative to progressive justice is "retributive justice". It is the modern version of the systems of justice that came into being at the dawn of human history, and it is based on the same reasoning.

The primal function of a system of justice is to depersonalise revenge. The agreement, perhaps the most ancient of all agreements that make it possible for communities to exist, is that the individual will take his complaint to the community. In return, the community will exact the appropriate retribution--partly on behalf of the wronged individual, but also to express the community's moral values. Justice means retribution through punishment and upholding the supremacy of the good members of the community over the bad.

The word "retribution" is jarring to modern sensibility. Someone who wants retribution is harking back to the bad old days of an eye for an eye, we think. Retribution is something civilised societies ought to rise above. The victim's desire for retribution is atavistic and unworthy.

Is it? As a way of testing your own views, consider a thought experiment that Immanuel Kant, the German philosopher, posed two centuries ago. He imagined an island society that is to disband tomorrow. Its citizens must decide whether a murderer awaiting execution should be executed. (If you're against the death penalty, substitute some other suitable punishment.) Executing him will have no expedient benefit for members of the society. It will certainly have no benefit for the prisoner.

We will assume that the prisoner, if released, will not kill again. The only purpose of the punishment is retribution. Should the murderer be executed? Kant says yes, that "the last murderer remaining in prison must first be executed so that everyone will duly receive what his actions are worth". What do you say?

This way of looking at the function of justice has a distinguished intellectual pedigree, but the principle itself is deeply ingrained in most people's sense of the rightness of things. It feels instinctively wrong when someone does something bad and gets away with it. When we say that someone "gets away with it", we mean that the person suffers no punishment or too little punishment.

The core tenets of retributive justice are simple. The necessary and sufficient justification for punishing criminals is that they did something for which they deserve punishment. Here "something" refers to the behaviours that society has defined as offences; "deserve" means that the offenders are culpable--morally responsible. Society not only has the right to punish culpable offenders; the moral responsibility of the offender imposes on society the duty to punish.

That's it. Nothing about rehabilitation, remorse or socioeconomic disadvantage.

Nothing about the bad effects that the punishment might have on the offender or, for that matter, its good effects. The purpose of a sentence is punishment. When a system fails to punish culpable offenders, it has failed, full stop. It is unjust.

Before rejecting retributive justice out of hand as far too harsh, you should realise that you can be a retributivist and still be against disproportionate punishments. You can oppose the death penalty, for example, or think that the appropriate punishment for painting graffiti on a park bench is not a jail term, but being forced to scrub graffiti from 10 park benches.

Being a retributivist does not mean you must give up on rehabilitation. Add all the educational and therapeutic services you want to the sentence, as long as the sentence itself constitutes a punishment.

You can be a retributivist and still be a civil libertarian of sorts.

A just system requires that culpability be judged correctly, which means that criminal justice procedures should protect the innocent.

But if retributive justice is not as harsh as it may first appear, neither is it warm and fuzzy. Before you decide that you are a retributivist after all, you have to decide what you think of this collateral view of crime and justice that is necessary to make retributivism work. Here it is: for practical purposes, society can be divided into citizens and outlaws. In conflicts between them, the law should favour the citizens.

The progressive view of justice vehemently rejects this notion, starting instead from the premise that we are all sinners. Some of us pad out expense accounts while others mug pensioners -both acts are thefts. All of us could be arrested for something, and probably even be put in prison for something.

People fit on a continuum, not into black and white categories, and justice should reflect that continuum. In defending a dichotomy between citizens and outlaws, I am allying myself with an old English legal maxim: because there is twilight does not mean there is neither night nor day. We are all sinners in God's eyes, but the everyday world contains millions of decent, law-abiding people--the people whom I label citizens--who are different in kind from a much smaller number of people whom I label outlaws.

Perhaps citizens pad out their expense accounts, but they never come close to killing, wounding, robbing, burgling, or raping -the elemental predatory acts.

The person who does kill, wound, rob, burgle or rape has stepped over a line and become an outlaw. While he is in a state of outlawry, he has lost many rights that citizens enjoy.

During the actual commission of the crime, the outlaw's rights have nearly disappeared. The extreme view is that a mugger has no cause for complaint if a young woman victim not only sprays him with Mace but pulls out a handgun. The burglar has no complaint if the homeowner pulls out a shotgun and shoots him as he attempts to flee. In both cases the outlaw has caused his own wounding or death--if he had chosen to be a neighbour instead of a predator, he would have gone unharmed.

Other retributivists take a less wide-open view of the conditions under which lethal force is justified, but the common principle is that criminals in the act of committing a crime are taking their chances. Victims are not expected to respond "proportionally" to being victimised.

The state of being an outlaw also implies reduced rights during the judicial process. The principle to be upheld is that the judicial process is not a game but a solemn search for the truth. The objective of the judicial process is to know as much as possible about all that can assist in determining the truth.

Rules against self-incrimination make sense not because confessions are a bad thing, but because states have had a habit of coercing inaccurate confessions. The only way to prevent coercion has been to make sweeping rules against forced self-incrimination -but that has been a practical necessity, not to be confused with an offender's "right" to hide incriminating evidence about himself. People who are accused of offences have a right not to be tortured.

If an infallible truth serum were invented, with no side effects, and was used only to question people about the specific allegation for which they have been arrested, justice would be served by compelling its use. Forcing defendants to give DNA samples is an analogous resource already with us.

Should evidence of prior arrests be excluded from a trial? If justice is a game, yes. If justice is a search for truth, no.

In every human endeavour other than justice, we take it for granted that arriving at the truth is aided by assembling as much relevant information as possible. The employer hiring someone for an important post does not limit his assessment to the impression he gets from the job interview but tries as best he can to discover how the applicant performed on previous jobs. A woman whose husband comes home with lipstick on his collar and a plausible explanation takes his prior history into account before deciding whether to believe him. Only the criminal justice system has the wacky notion that data about prior behaviour should be excluded in deciding upon a person's guilt or innocence.

The only reason why excluding evidence from prior arrests makes sense is if we don't trust jurors' good sense. Jurors must combine many different types of evidence, none of them absolutely conclusive, and come up with their best estimate of the probability of guilt; the criterion of "beyond a reasonable doubt" gives them no choice but to work with probabilities.

If the defendant's prior record consists of one arrest and no prosecution 10 years ago, that record shouldn't have more than a tiny part in that calculation. If the prior record consists of five arrests during the past two years with much circumstantial evidence but no convictions, it can legitimately play a substantial part in the calculation. Why not rely upon jurors being able to make such distinctions in the same way that they make such distinctions in other aspects of their daily lives?

On this point, I suspect that just about everyone who has served on a jury, and has discovered later that evidence was kept from them, has had the same reaction I had when it happened to me: fury that the system had impeded me and my fellow jurors from doing a difficult job as fairly as possible.

If we admit prior criminal records in testimony, are we not conflating the defendant's character with his guilt for the crime in question? In the case of defendants with many prior arrests and convictions, yes -but this is a virtue, not a defect. One of the most consistent findings about crime around the world is that a small proportion of offenders commit a very large proportion of the crimes.

Divinely accurate retributive justice would not punish a burglar for the one burglary out of dozens when the burglar got caught, but for the aggregate harm that the burglar has done. Judging whether an offender in the dock is a habitual offender is useful in deciding on the probability of guilt in the present case and essential to deciding on the severity of the punishment. A record of many prior arrests is an indicator, more reliable than most, that should feed into both judgments.

All of the issues I have treated so sketchily raise genuinely difficult problems that need to be considered at length. The "citizens versus outlaws" perspective merely says that in making judgments about the benefits and risks of alternative policies, it's not supposed to be a level playing field: citizens get preference.

As the debate stands, citizens are in effect held hostage to threats by the outlaws ("If you put me in prison, I'll just become a worse criminal"). As the debate stands, the legal protections of an innocent defendant are treated as if they were identical to the legal protections of the guilty defendant (they aren't). As the debate stands, nobody is supposed to mention the obvious: everything good in British life is produced by the citizens. That they may go about their lives freely and in peace is the highest goal that government can serve.

Retributive justice is not only simple in concept, it is simple to implement.

Require the legislature to set punishments that are commensurate with the gravity of the various offences. Focus police and court resources on the most serious offences. Give juries all the evidence about the offender in the dock. Limit the ability of judges to depart from the prescribed penalty. Administer the punishment.

Simple, but expensive. In modern Britain the only authentic punishment for serious crimes is imprisonment. To implement retributive justice would mean an increase in the prison population that dwarfs anything any politician, even those most vehemently anti-crime, are prepared to countenance.

To see just how extraordinary the increase in the prison population would be, consider Britain as it was in 1954. If you divide the number of people convicted of robbery who were incarcerated that year by the number of robberies reported to the police during that year, the ratio in 1954 was 1:3. In 2002, the comparable ratio was 1:22. For burglary in a dwelling the ratio was 1:18 in 1954 compared with 1:59 in 2002. For serious wounding the ratio was 1:5 in 1954 compared with 1:12 in 2002.

Because misunderstanding of this issue is so common I should emphasise that the leniency of crown court judges after a conviction is a minor reason for these contrasting numbers. Judges dealing with the most serious offences are almost as tough on convicted offenders in contemporary Britain as they were in 1954.

Leniency resides in the ways that bringing an accused to trial has been made so difficult and time-consuming, and is reflected in the large proportions of cleared-up cases that are not prosecuted as serious crimes and in the decreasing clear-up rates for serious crimes.

The bottom line is staggering: if Britain today imprisoned the same ratio of people relative to the number of the most serious offences (murder, attempted murder, serious woundings, rape, burglary and robbery) that it imprisoned in 1954, the British prison population in 2004 would be in the order of 290,000, more than 3A times the 74,000 that is currently causing such a stir.

David Blunkett is attracting criticism for a plan to add 5,400 prison spaces. To get back to the system of 1954 he would have to add about 200,000. The changes produced by the reforms of the past half-dozen years have been trivial when the frame of reference is British criminal justice as it used to be.

Using that frame of reference is a good idea. The view of justice I have advocated must seem radical and barbarically punitive to many readers. But that's the system Britain used to have -not in the Dark Ages but at the beginning of the reign of Elizabeth II. Britain still practised retributive justice back in 1954, although nobody called it that. It was simply justice, the only kind that crossed anyone's mind.

It worked. In the first half of the 20th century Britain had one of the freest societies ever created, and at the same time one with such extraordinarily little crime that the civility of British life was an international cliche. The professionalism of the British police and the fairness of British courts set standards for the world. It was not a perfect system, but perfection is not what complex governmental systems can aspire to.

By any realistic measure, British criminal justice was superb. Its philosophical core was what I have been calling retributive justice, applied consistently and without apology. Isn't it time for its return?

Charles Murray is the W. H. Brady Scholar at AEI.