Having juries issue multiple verdicts according to multiple standards of guilt

Copyright Ó 2003, by Alec Rawls


I   Introduction

Because juries in criminal cases are only asked to apply a single standard of guilt, our system of criminal law admits no distinction between the certainly guilty and the almost certainly guilty, some of whom will be innocent. This lumping together of the certainly guilty with a proportion of innocents greatly limits how we can deal with the certainly guilty, and it greatly increases the harm done to those innocents who are wrongly convicted. We could achieve more justice for both the innocent and the guilty simply by asking juries to discriminate further. Instead of rendering one verdict, have them render two: is the defendant guilty according to a true certainty standard of guilt, and is he guilty according to an explicit somewhat less than certainty standard. The idea would be, not to lower the standard for conviction, but to complement what currently is in practice far less than a certainty standard with an actual certainty standard.

Perfect certainty is an impossibility. At the very least, radical skeptical doubts about human perception and the possibly conspiratorial nature of human testimony will always remain. Still it is perfectly possible to define a standard of guilt where no significant or real doubts remain. Our current "beyond a reasonable doubt" criterion of guilt is just such a standard, requiring "human" or "moral" certainty, meaning certainty as far as such a thing is possible in human affairs (see section II below). The problem is to get juries to adhere to such a standard.

Jurors will disobey a stand alone certainty standard because to let a defendant go unless there is complete (human) certainty is to let the tiniest risk of convicting an innocent person outweigh almost a near certainty of letting a guilty person go free, effectively giving no weight to the latter. Such a choice stands in direct violation of the fundamental principle of moral reason, which is to account all value.(1) Ask moral people not to account the costs of letting the guilty go free and they won't do it. Thus our nominal certainty standard devolves into far less than a certainty standard in practice.

But suppose we were to make our current less than certainty standard an explicit less than certainty standard, conviction under which would yield reduced punishments, then complement it with a certainty standard, conviction under which would incur enhanced punishments. The less than certainty standard would give jurors a way to account the expected costs of freeing the probably guilty, allowing them to deliberate without moral ambiguity on the question of certainty. Thus the key to achieving a true certainty standard of guilt is to complement it with an explicit less than certainty standard. The way to achieve a true certainty standard for use in a two verdict system is simply to have a two verdict system. With some additional safeguards, we can achieve a true (human) certainty standard of guilt, under which there is virtually no chance of an innocent person being found certainly guilty.

This achievement would open up avenues of prosecution that yield tremendous leverage over the guilty while greatly reducing risks to the innocent. Consider the death penalty, which presents a unique problem in that, once carried out, it cannot be commuted if evidence of innocence arises. The proper response to that limitation is to apply the death penalty only when a true certainty standard is met.(2) Then we would be certain that evidence of innocence would not arise. If a true certainty standard can be achieved then executions can be carried out quickly and cheaply with virtually no risk to innocents. A large majority of Americans favor the death penalty but also want to take whatever steps are necessary to make sure that innocents are not at risk.(3) A true certainty standard would allow both desiderata to be satisfied.

An even more powerful avenue of prosecution that would be enabled by a true certainty standard of guilt would be severe additional punishments for those who plead innocent and are found certainly guilty. At present we must tolerate the ultimate act of recidivist defiance—the perjurous declaration innocence, where a guilty party continues to try to get away with crime, even while standing in the dock of justice—because to punish it would heap a double injustice on those innocent accused who plead innocent and are found guilty. But if we can separate the almost certainly guilty (who will include some innocents) from the certainly guilty (who will not) then we can punish this act of perjury and recidivism with virtually no risk to the innocent. We just have to limit extra punishment pleading innocent to those who are found certainly guilty.

Imagine the leverage this would afford over the guilty, who would always have to fear that actual proof of their guilt might emerge. If punishments are doubled for defendants who plead innocent and are found certainly guilty there would be extreme incentive for the guilty to plead guilty, even absent any bargaining down of charges. Case loads and costs would tumble. Yet not only would the innocent would stand clear of this heavy artillery, but the injustice done to them when they are wrongly convicted would be greatly lessened because, just as separating the certainly guilty from the almost certainly guilty would enable increased punishments for the certainly guilty, so too it would call for reduced punishments for the less than certainly guilty, reducing harm to the percentage of innocents who are caught in that wider net.

Altogether, changing to a system of dual verdicts should yield large efficiencies all round: quite a bit more justice for both the innocent and the guilty with greatly reduced expenditure of judicial resources (both from fewer guilty defendants going to trial and from cheap and permanent disposal of the worst offenders).

The first step is to verify that a true certainty standard is really possible. This requires detailed dissection of the conflict between our present stand alone certainty standard and moral sense. Some such conflicts will remain in a two verdict system, but with regard to the certainty standard they are orders of magnitude smaller standard and should not be significant. This analysis is the subject of sections II through IV. Sections V through VII consider what backup mechanisms and appeals processes would be necessary to support a true certainty standard. Sections VIII through XIII take a closer look at the avenues of prosecution that a true certainty standard enables.

Section XIV notes some natural law requirements of the concept of certainty, while section XV describes how refinement of natural law concepts could lead to further innovations that would allow restrictions on evidence gathering to be relaxed. The large resulting increase in evidence gathered would then feed into the extreme leverage created by the system of multiple verdicts of guilt. Thus the two schemes, each independently powerful for fighting crime, would combine multiplicatively. The second scheme is not described in depth, just enough to indicate the complementarity with the present scheme.

Section XVI looks into some questions about the less than certainty side of the two verdict formula. Section XVII considers the merits of adding yet a third standard of guilt for cases of terrorism and the worst serial crimes, and section XVIII examines the constitutionality of the proposed changes, concluding that they can be instituted, certainly at the state level, without federal constitutional amendment.


II   The conflict between our present one verdict system and moral sense

Our present "beyond a reasonable doubt" standard guilt is already, according to the derivation of its terms, a certainty standard. "Beyond reasonable doubt" refers to a complete absence of reasonable, or rational, doubts.(4) The absence of rational doubts is a way of expressing the concept of "moral certainty," which historically was a reference the highest degree of certainty that is possible in human affairs and on the basis of testimony.(5) This derivation of the "reasonable doubt" standard is widely accepted and has been embraced by the Supreme Court, in Sandra Day O'Connor's majority opinion from Victor v. Nebraska .(6) Yet despite being nominally a certainty standard, the "reasonable doubt" standard of guilt has for the most part not been treated as a certainty standard either by interpretation or in practice.

As for interpretation, Judge John W. May of Boston, who began the modern historical research on the "beyond reasonable doubt" standard, saw it as an expression of Sir Matthew Hale's famous maxim that: "It is better that five guilty persons should escape unpunished than one innocent person should die."(7) To consist with this maxim, jurors on the cusp should estimate that the defendant is five times as likely to be guilty as innocent, which implies about an 83% certainty standard of guilt.(8) Blackstone urged a ten to one standard (91%), Fortescue a twenty to one standard (95%) and Starkie a ninety nine or more to one standard (99% plus).(9) Taken literally, the "beyond a reasonable doubt" standard calls for 100% certainty before a verdict of guilty can be rendered, implying that it is better to free an infinite number of guilty persons rather than convict one innocent.

In practice, the "reasonable doubt" standard probably works out to something less than the one in five standard that Hale called for. Surveys show that judges tend to regard the "reasonable doubt" standard as equivalent to something like a 90% certainty of guilt, but empirical evidence suggests that it may work out to something substantially weaker, sometimes even falling below the 50% certainty level usually associated with the "preponderance of evidence" standard.(10)

If our nominal certainty standard fails to secure an actual certainty standard, how can a true certainty standard ever be achieved? The answer lies in the unreasonableness of a stand alone certainty standard. As Judge May noted, Hale's maxim is already a "doctrine of leniency," expressing the idea that "it is better to err on the side of mercy than on the side of justice."(11) But pushing this predilection to the extreme of certainty makes justice impossible. "Better any amount of crime than one mistake in well-meant endeavors to suppress or prevent it! Better for whom?" Judge May asked, "for society, or for the malefactor?"(12) May held that such a doctrine, sacrificing the protection of society for the last increment of protection for the almost certainly guilty "logically gives justice to nobody, and mercy to those only who show none and deserve none."(13)

Modern decision theory, introduced to lawyers by John Kaplan in his 1968 Paper "Decision Theory and the Factfinding Process," offers an analytic grounds for coming to the same conclusion. An incremental raising of the standard of guilt increases the chances of freeing the guilty (type I error) and decreases the chances of convicting the innocent (type II error). So long as the benefit from raising the standard of guilt (the decrease in probability that the innocent will be convicted, times the harm of punishing the innocent) exceeds the harm (the increased probability of freeing the guilty, times the harm of freeing the guilty) the standard of guilt should be raised. Similarly, when the benefits of lowering the standard exceed the costs, it should be lowered.(14)

To carry this calculation through, the different components of harm and benefit have to be tallied up. The benefits of locking up the guilty are incapacitation (keeping innocents from being victimized by the defendant in the future), deterrence, and justice (retribution). With punishment of the innocent there is the injury to the particular innocent to account, plus the anxiety for all of living in a society where innocents are prone to be punished. Then there are institutional incentives to consider. How will those who operate the criminal justice system react to a lowered standard of guilt? A lowered standard of guilt would, for instance, make it easier for the police to frame people. To the extent that there are bad actors amongst the police, that has to be accounted. In sum, there are lots of ways that harms to innocents are at stake on both sides of the balance and they all need to be accounted.

The resulting calculation powerfully rejects a stand alone certainty standard. Such a standard only weighs the harm of punishing the innocent, giving the slightest risk of such harm priority over the near certainty of freeing the guilty, no matter how much greater in expectation the harms of the latter are.

This decision-theory analysis accords, in turn, with the principles of moral reason, which are just the principles of reason applied to matters of value. Thinking straight about value requires, firstly, that all of what one can see to value must be fully accounted (value must be valued) whenever it is enough at stake to be worth accounting. Thus our nominal certainty standard of guilt, giving zero weight to the risk of freeing the guilty, stands in direct violation of the first principle of moral reason.(15)

Jurors are on average very moral beings. Possessing moral reason, they are powerfully driven to account all value, including the risk of freeing the guilty. When they are given jury instructions that ask them not to account all value, their moral sense will urge them to violate these instructions. As Judge May put it: "nothing saves our criminal courts from complete paralysis but the stalwart sense of the juror who refuses to adopt methods to the accomplishment of any purpose which seem to him exactly adapted to defeat it."(16) Thus the inevitable consequence of the deep immorality of a stand alone certainty standard of guilt is that our present nominal certainty standard devolves into far less than a certainty standard in practice.


III   A two verdict scheme enables a true certainty standard

What causes our nominal certainty standard of guilt to devolve into less than a certainty standard is the need to account the cost of letting the possibly guilty go free. If the nominal certainty standard were complemented with an explicit less than certainty standard, jurors would not have to find a defendant certainly guilty in order to protect society against the possibility that the defendants is guilty. They could find the defendant less than certainly guilty (with lesser punishments applied). That should allow jurors to deliberate on the certainty standard with little or no conflict between jury instructions and moral sense.

There might still be some conflict with moral sense, but it would be vastly diminished. When a defendant is almost certainly guilty, a full accounting of value could still urge jurors to impose the severe punishments that the defendant almost certainly deserves. What has changed is the alternative. If the defendant is found guilty at the less than certainty level, he will receive perhaps half the punishment he would if found certainly guilty. If he is actually guilty, how much better is it that he get half what he deserves than none? How much better is it that a serial rapist gets twenty years in jail than that he still be on the street (even though he escapes staying in jail until he dies)? How much better is it that a serial murder go to jail until he dies than that he not go to jail at all (even though he escapes the death penalty)? Is it ten times better? A hundred times? A thousand? The conflict is reduced by orders of magnitude.

At the same time, the heightened punishments enabled by a certainty conviction should cause increased horror at the prospect of convicting an innocent person. Altogether, then, jurors should feel vastly less impulse to violate jury instructions in order to account the possibility that the defendant is guilty, and more impulse to obey instructions so as to avoid the risk of harm to the innocent. Add that moral jurors will also tend to appreciate the need for a consistent system of justice with rules and standards that jurors abide by and we can conclude that, under the two verdict system, moral considerations on the whole will rarely if ever provide any net moral impetus for jurors to violate their instructions.

Where there could still be substantial conflict between moral sense and jury instructions in a two verdict system is in a jury's application of the less than certainty standard of guilt. Suppose a juror estimates a 70% chance that the defendant is guilty. Suppose also that the charged crime is especially heinous or suggestive or recidivism, so that the likelihood of further harm to innocents if the defendant is guilty and is released is very high. In such a case, the harm to innocents from freeing a guilty defendant could be greater than the harm of convicting an innocent, so that moral sense would urge conviction all the way down to the 50% certainty level. With 70% certainty of guilt, it will powerfully urge that jury instructions be violated. Thus having the option of finding the defendant guilty at the 80% level does not eliminate conflict between jury instructions and moral sense but in effect shifts the conflict from the certainty standard to the less than certainty standard.

Luckily, that is all we need. Shifting the moral conflict to the less than certainty standard allows the certainty standard to become a true certainty standard. That, in turn, allows us to deal full justice to those who are found certainly guilty. The reason to temper justice with mercy is out of concern for the percentage of convicts who are actually innocent. When we can separate the almost certainly guilty, with its percentage of innocents, from the certainly guilty, who contain no innocents, then the proper way to deal with the latter is to give them unalloyed justice. Because the guilty are always prone to be found certainly guilty, the result will be a terrible leverage over the guilty that can be used to make them come clean.


IV   Error by source

The above is a highly abstract account of how our nominal certainty standard degrades into much less than a certainty standard and how it could be turned into a true certainty standard. Investigations into actual cases where innocent people have been found certainly guilty (according to our present "reasonable doubt" standard of guilt) back up this theoretical account.

In their studies of errors in capital cases, Michael Radelet and Hugo Bedau broke down the main sources of discovered error into four categories: witness error, police error (careless or slanted investigation, coerced or otherwise errant confessions), prosecutor error (withholding of evidence, overselling weak cases) and miscellaneous errors by judges, defense lawyers and crime labs.(17) It is the standard of guilt that determines how wide the door is opened for these various types of error to get through. This can be seen in the cross-section of error examined by Barry Scheck, Peter Neufeld and Jim Dwyer in their book Actual Innocence, which reviews individual cases where convicts were later exonerated by DNA testing.(18) No matter the source of the error, jurors in almost every case of wrongful conviction were faced with truckloads of reasonable doubt and simply violated their jury instructions.

This violation of instructions can be morally sound. It is perfectly appropriate to be just as worried about type I error (acquitting the guilty) as type II error (convicting the innocent). Indeed, in the case of serial crimes, or crimes subject to high rates of recidivism, type I error may do much more harm than type II error, so that a full accounting of value might even prompt jurors to make concern for type I error their primary responsibility, in exact opposition to jury instructions. Post mortems performed on DNA based exonerations show just how far jurors will go in this direction.(19)

This anecdotal cross section of discovered type II error backs up the experimental evidence with mock juries that has found our nominal certainty standard to devolve into something more like a preponderance of evidence standard in practice.(20) Thanks to their moral sense, jurors are loath to acquit a person who they think is probably guilty and they are correct to assume that no one who isn't at least more likely to be guilty than innocent according to the evidence should ever be brought to trial. The upshot is that juries often end up trusting police and prosecutors to point them to their moral duty, leaving the door to conviction wide open to errors and abuses by the state, to witness error and other defects.

So long as we rely on a single standard of guilt this is unavoidable. Any kind of error is prone to be decisive just by being suggestive. But if jurors can be directed to apply themselves to the unbiased application of an actual proof standard, as should occur under a two verdict system, then only the most severe or malevolent errors could be so misleading as to threaten an errant certainty conviction, and such errors would always be countered by at least one clearly rational doubt: that what actually happened could have happened. Unless there is actual proof, the certainty verdict is to be withheld. The concept is simple, clear and easy to apply. The trick is to get juries to accept it. Once conflict with moral sense is eliminated, so that jurors can accept the task of applying the concept of actual proof, the door is closed on most sources of error. Errors may still infect the system, but at the level of the certainty standard they should almost never lead to errant convictions.

Of course this is no reason to be complacent about the various sources of error. Regarding the less than certainty standard of guilt, any substantial error will still tend to be decisive, and if prosecutors and police are allowed to get out of control then the most misleading and compound errors, the kind that can be decisive even at the certainty level, are bound to occur. If we don't want a significant number of errors to get by the catcher, we had better try to minimize the number of fastballs that are thrown into the dirt. Scheck, Neufeld and Dwyer offer a compendium of possible safeguards against various sources of error, many of which would reduce type II error without significantly increasing type I error.(21) The dollar costs are also mostly pretty low, certainly when compared to the extreme price we pay now, both in dollars and in type I error, to avoid conviction of the innocent (even though we don't avoid it very well). Most obviously, we should be fully funding our forensic labs, insuring that they are independent, and checking to make sure they aren't cheating.


V   A fail safe for jury incompetence and standards for appeal

There is also the problem of jury competence to consider. Even absent a conflict between jury instructions and moral sense, getting jurors to properly follow jury instructions and adhere to a true certainty standard is a substantial proposition. Real proof—the question of what constitutes a rational doubt—is largely a matter of logic and reason, which many people are not adept at. Having twelve jurors should add substantially to jury reliability, but it may introduce odd group dynamics as well.

The handiest way to reinforce the certainty standard against jury incompetence would be to add a requirement that the judge in the case must second any certainty verdict before a certainty conviction is considered valid. This might seem like little backstop when going before a judge known for severity, but as with juries, the availability of a less than certainty standard of guilt also removes the conflict that judges at present feel between our nominal certainty standard of guilt and their moral obligation to account the costs of freeing the guilty. Judges, like juries, will be freed from this conflict to apply the concept of proof objectively, in most cases bringing a high level of competence to this determination.

When this competence is lacking the problem of judicial errors arises, either due process errors or failures by judges to recognize absence of proof. The latter could be handled by allowing review whenever any definite persistence of rational doubt can be shown, which would also address problems of incompetent lawyering. If there is a clear possibility of innocence that defense lawyers failed to raise or that the judge and jury failed to credit, this would be grounds for appeal. The burden of evidence on appeal should be proof of non-proof. Proof of guilt is already the ultimate standard. If a hole in proof cannot be demonstrated there is no reason (on grounds of uncertainty) to further withhold the punishments that certainty enables, such as the death penalty.

The executive would also have to have its say. Convicts could use arguments about holes in proof to appeal to Governors and Presidents for commutations of certainty convictions into less than certainty convictions. If proof of guilt is lacking in a case then a sympathetic political ear might be expected, assuming that a substantial majority of the electorate really does want to withhold ultimate severity when guilt is not certain, as polls suggest.(22) As for appellate recourse, overturning of certainty convictions would only be called for in extreme cases of malfeasance, dereliction or incompetence at the trial level. A judge and 12 citizens, applying the concept of proof in an unbiased manner (as the two verdict system promotes) may pass some questionable judgements at the margin about what constitutes a real doubt, but for them all to dismiss a doubt that is not marginal but is clearly real would take a wide confluence of intentional violation of both moral reason and jury instructions. It is important to be able to overrule such instances, but there is no reason to expect them to occur.


VI   Due process errors

Due process errors are a separate matter, but reviewability for certainty would greatly simplify review for due process by separating concerns about possible innocence from the need to enforce due process as a check on arbitrary power. As it is, these concerns are conflated, especially in death penalty cases where extensive due process oversight has been added, not primarily in order to enforce due process, but out of a concern for possible innocence, creating what has been described as a kind of "super due process."(23) But the magnitude of the worst crimes makes technicalities relatively less important, not more.(24) Thus a focus on technicalities is a highly perverse way to compensate for uncertainty, especially in death penalty cases.

In the two verdict scheme, uncertainty is dealt with by issuing the severest punishments only when the certainty standard of guilt is met. Once questions of due process have been untangled in this way from questions of certainty of guilt, they can be handled as a separate matter, with no call for particularly heightened scrutiny in certainty cases or death penalty cases. Such a position is backed up by the Supreme Court, which ruled in U.S. v Calandra that the exclusionary rule "is a judicially created remedy designed to safeguard Fourth Amendment rights generally through its deterrent effect, rather than a personal constitutional right of the party aggrieved."(25) In effect, the court refused to find an interest in protecting, for their own sake, those instances where due process rights have the unfortunate side effect of protecting criminal activity.

This distinction between an interest in protecting due process and an interest in protecting criminal activities that are protected by due process is maximized in a case where the evidence (perhaps improperly collected) is found by a jury to constitute proof of certain guilt under a two verdict system. The case is revealed to be in fact one of those unfortunate cases where due process does not shelter legitimate activity from arbitrary government power, but shelters the arbitrary power of criminals from detection. It is proved to be one of the costs of due process, not one of the benefits. Thus while due process problems could still provide grounds for appeal under a two verdict system, just as they can under the present one verdict system, findings of certainty would lessen rather than heighten the weight that should be given to due process concerns.


VII   Due process review of less than certainty convictions

While due process errors become less important when a true certainty standard of guilt is met, they are of undiminished relevance in cases where only the less than certainty standard is met, and of course they are always a grave concern when the issue is protecting legitimate activity from arbitrary government power. When the latter may be at stake, the exclusionary rule is obviously called for (more on this in section XIII). When the case is clearly criminal, the only concern is to deter violations of due process by the police, and for this purpose the exclusionary rule is clearly irrational. For incentives to work, they must impinge on who is to be affected. It makes little sense to punish the innocent, by withholding justice, when the goal is to affect the police. A more plausible alternative that has been suggested is to install a system of review boards with the power to discipline police who violate due process rules.(26)

Why go into this here? Just to note that the two verdict system offers yet another option for enforcing due process. The presence of two different standards of guilt, with different punishment standards, would provide a way of splitting the difference between applying and not applying the exclusionary rule. When important violations of due process are found in cases of certain guilt, it would be possible to knock the conviction down to a less than certainty conviction, so that instead of the defendant going scott free, he would receive the lesser punishment attendant on a less than certainty conviction. As noted above, certainty cases are actually the least likely to call for the exclusionary rule to be applied. Still, there might be cases where half-applying the rule would be a valuable option. In any case, the two verdict system comes with this possibility built in.

Overall, a certainty standard could be achieved without the need for an elaborate review process. The main failsafe—the trial judge's concurrence in the certainty verdict—would be concurrent with the jury verdict and would cover all sources of error except judicial error. Judicial error in certainty convictions could be open for automatic review for the presence of clearly unrebutted claims of reasonable doubt (proofs of non-proof) and for clear and substantial violations of due process. No more review would be called for before certainty punishments are be applied, including the death penalty. So long as guilt is certain, there is no reason to hesitate further.


VIII   Punishing perjurous declarations of innocence

With certainty in hand, severity is enabled, and can be put to good effect. Consider the act of pleading innocent when guilty. Not only is this perjury, it is the essence of recidivism: still trying to get away with crime rather than admitting oneself to punishment and reform. Criminals whose mentality is recidivist should not be let back out to commit more crimes and anyone who is so brazen as to continue to defy justice even when standing in the dock demonstrates a strongly recidivist mentality, calling for greatly extended incapacitation.

Under a less than certainty standard of guilt it is inappropriate, when a defendant pleads innocent and is found guilty, to add punishment for perjury and recidivism. Some who plead innocent and are found guilty will actually be innocent and additional charges would compound this error. It is a cause for ambivalence that some such additional punishments are administered at present. The punishments for being found guilty are generally substantially higher than a defendant could have obtained through plea bargaining and the person who is wrongly convicted is hammered yet a third time if he continues to maintain his innocence in the sentencing phase, which judges typically answer with a maximum sentence to punish lack of remorse.

The guilty, in contrast, actually deserve much harsher punishment: not just a maximum sentence, but severe additional punishments for perjury and recidivism. Dual standards of guilt would separate these cases. An innocent person who pleads innocent might be found "almost certainly guilty," but under a true certainty standard he will virtually never be found "certainly guilty." He will virtually never suffer a double or triple injustice, as happens with substantial regularity now. Thus the innocent would be far better served by dual verdicts, and the guilty? With innocents out of harm's way, we could focus on giving those who are found certainly guilty the full punishment that they deserve, both for their original crime and for continuing to try to get away with it. A fifty percent increase in the severity of the sentence, or perhaps a doubling of severity, seems reasonable. To the extent that judges retain discretion over sentencing, this result would have be a product both of the range of sentences allowed and the behavior of judges.

Such hefty additional punishment for being found certainly guilty would give guilty defendants tremendous incentive to plead guilty, without any offer of reduced charges. The trash would be taken out of society far more reliably and at much less cost than at present, and for those perpetrators who are brazen enough to plead innocent in spite of the risks, we will be able to take comfort in the percentage that do get squashed by the anvil. This additional severity accords with all three principles of punishment. In addition to deterring the crime of pleading innocent when guilty, it also incapacitates those most in need of incapacitation (those with recidivist tendencies), and it accords with justice. The recalcitrant criminal deserves the harshest punishment.

At the same time as punishments can and should be considerably stiffened for those who plead innocent and are found certainly guilty, punishments should be lessened for those who are found less than certainly guilty. Since our criminal justice system currently lumps the certainly guilty and the almost certainly guilty together, current sentencing guidelines are for this combined set. As a result they are harsher than we would want to apply to the not quite certainly guilty alone. In this way too, a system of dual verdicts would decrease harms to the innocent.

The upshot would be a two or possibly three tiered punishment scheme. There are to be much harsher punishments for those who plead innocent and are found certainly guilty. The remaining question is whether there should be a punishment differential between those who are found to be less than certainly guilty and those who plead guilty. Again, to the extent that judges have discretion, the actual results here will depend on judicial behavior. The question is whether there is reason to punish the one category more harshly than the other.


IX   A two tiered punishment system

Careful analysis shows a two tiered punishment system (with those who plead guilty facing the same punishment standards as those who are found less than certainly guilty) to be best. On the one hand, those who are found almost certainly guilty will include some innocents, whose injuries we would like to minimize, suggesting that those who are found less than certainly guilty should be treated less harshly than those who plead guilty (and are hence certainly guilty). On the other hand, the great majority of those who are found almost certainly guilty will actually be guilty and should ideally receive extra punishment for perjury and recidivism. Neither of these considerations takes any clear priority over the other, leaving no compelling argument for differential punishments. At the same time, if either is punished more harshly than the other, incentive problems arise.

If those who are found by a jury to be less than certainly guilty are punished less harshly than those who plead guilty, that gives incentive for those who are guilty, but don't think they can be proven certainly guilty, to plead innocent. It may be a bad risk, but many criminals are obviously slow learners about risk. Better to keep their incentives as unambiguous as possible rather than offer temptation.(27)

The other incentive that needs to be unambiguous is the incentive for the innocent to plead innocent. One of the horrors of our present system, where defendants typically have to choose between a plea to lesser charges and trial on harsher charges, is that the innocent are regularly presented with powerful incentive to plead guilty.(28) Under a two verdict system this can be eliminated. The worst plausible result for an innocent pleading innocent is to be found less than certainly guilty.(29) So long as the punishment for pleading guilty is not less than for being found less than certainly guilty, innocents face no real downside risk from pleading innocent rather than pleading guilty to the same crime. Thus the key to eliminating downside risk to innocents of pleading innocent is to limit the tendency to plea bargain charges down, and the two verdict system should have a powerful such effect. Prosecutors would obviously prefer not to lower charges, so with criminals having a powerful incentive to plead, prosecutors can be expected to make fewer concessions.

Another consideration amplifies this conclusion. At present there are no strategic consequences to the fact that a defendant enters his plea before the discovery process begins. He can plead innocent and, when he learns the strength of the case against him, he can always decide to change his plea to guilty with no penalty for having first pled innocent. Once severe additional punishments for pleading innocent and being found certainly guilty enter the picture, the timing of plea and discovery suddenly become very important. If proof emerges in the discovery process, the prosecutor will not let the defendant enter only a plea of guilty to the full severity of the crime. The defendant is guilty of the serious further crime of perjury and recidivism for having pled innocent and he should be compelled to plead to this also. That is, in contrast to present practices, we might expect some bargaining up of charges.

Since this scenario depends on the emergence of proof of guilt, or the likelihood of the emergence of proof, innocents would again stand clear. But just the existence of such a dynamic could introduce a new set of expectations, which could go a long way towards eliminating the practice of bargaining down. To the extent that this occurred, downside risk to the innocent of pleading innocent would be completely eliminated, so long as pleading guilty is not punished less severely than being found less than certainly guilty.

Overall, this qualitative analysis suggests a two tiered punishment system, with equal punishments for pleading guilty and for being found less than certainly guilty, and much harsher punishments for being found certainly guilty.


X   Dangers of bargaining up

A defendant is not entirely in the dark when he enters his plea. The indictment against him reveals some of the evidence that incriminates him. Nevertheless, under a two verdict system, having to enter a plea partially blindfolded enhances the risk to the guilty of pleading innocent. This in itself is all to the good, but it does raise questions about the incentives that a two verdict system would create for prosecutors. Will they be tempted to go on fishing expeditions, indicting half-suspects on flimsy evidence in hopes that the guilty will plead guilty out of fear that proof has been or will be found? On the one hand, this prosecutorial leverage could be a wonderful tool, especially when there is a clear suspect in some heinous crime but no proof. On the other hand, if this leverage was used as a substitute for the hard work of evidence gathering then some criminals who could have been found guilty or certainly guilty will get off easy or go completely unpunished by pleading innocent.

Also, bringing charges on flimsy evidence would at some point violate due process rights. The right not to incriminate oneself was in part a rejection of the oath Ex Officio which required defendants before the British ecclesiastical courts to answer all questions on any subject.(30) The Sixth Amendment only requires that charges be specified, but at some point, having to answer enough specific charges on weak enough evidence becomes equivalent to having to answer all charges, violating the intent of the Fifth Amendment.

Several mechanisms could serve to check abuses. First, the demand of the public for competent prosecution should keep a check, through the electoral system, on lazy legwork. Second, the Supreme Court could step in and draw the line on violations of due process. Third, the grand jury, as arbiter of probable cause for indictment, should put a check fishing expeditions. It is impossible to predict how all this would work out. Very likely the increased prosecutorial leverage will lead to some untoward excesses, and optimization will likely call to some steps to moderate the negative possibilities. What we can say for sure is that adding vast prosecutorial leverage over the guilty, while leaving the innocent unscathed, is obviously on the whole a very good thing, highly productive of the main desiderata of a justice system. If we want to, it should be possible to spend the huge gains making sure that we don't go backwards on any single front, but there is no reason to think this would be optimal.


XI   The death penalty

When guilt is certain (whether by plea or by verdict) full severity is called for (both for the original crime alone, or for any perjurous plea that may have been entered). This will call for the death penalty to be applied far more than at present. The problem with the death penalty has nothing to do with its being too harsh for the crimes to which it is applied. The problem with the death penalty is that it cannot be taken back. The correct response to this limitation is to apply the death penalty only when guilt is certain. Since a true certainty standard requires a two verdict system, and we have not had a two verdict system, we have instead been trying to compensate for the irreversibility of the death penalty by curtailing the range of crimes for which it is applied. This will need to be undone.

Child molestation and rape are clear examples of crimes where sentences of death would often be called for if guilt were certain. Sex crimes are both among the most heinous of all crimes and are prone to the very highest rates of recidivism. Yet evidence of these crimes is very often fraught with uncertainty. Child molestation cases often depend on the testimony of children, which has proven to be extremely unreliable and subject to manipulation. Rape segues into many gray areas of pressure that might or might not threaten force and it segues into submission that may or may not communicate non-consent, with accusations further clouded by possible motivations of regret or recrimination. The law responds to these commonplace uncertainties with merciful standards of punishment, thereby often giving mercy rather than justice to the guilty. In contrast, two verdict system would distinguish between cases where uncertainty is and is not present, allowing us to focus our mercy on those cases where uncertainty is present and unleash justice where it is not.

A second way that our current system compensates for the irreversibility of the death penalty is by allowing for an interminable appeals process which costs more than life imprisonment, which delays justice for victims and their families sometimes for decades, and which defeats any deterrent power that the death penalty has by pushing it further into the future than a typical criminal imagines living in the first place.(31) By applying the death penalty only when a true certainty standard of guilt is met we could not only broaden the death penalty to the full range of cases where it is warranted but we could also eliminate most appeals, making an exception only for clear cases of uncertainty or violation of due process. The risk of putting innocent people to death would be far less than it is now (it would be essentially zero), and society's evil predators would be eliminated, quickly, permanently, with huge savings in court and prison costs.

The largest portion of the high cost of death penalty cases at present is the vastly higher cost of the initial case, where the courts compensate for the weak standard of guilt afforded by our present single standard of guilt by imposing "super due process."(32) Under a system of dual verdicts, every case would distinguish the certainly guilty from the almost certainly guilty, without adverting to any heightened due process. Death penalty cases would not be any different from any other case in this respect. What concerns there are for due process per se are no greater in death penalty cases than in other cases (if anything they are relatively less important). What is different in death penalty cases is the need for certainty, which is far better served by dual standards of guilt than by any amount of heightened sensitivity to due process (which typically does not heighten truth seeking, but obstructs it).

Another current death penalty expense that would disappear is the separate penalty trial. Under the system of dual verdicts there would just be one trial with two different sets of penalties, one each for the different degrees of certainty of guilt that may be arrived at by the jury. Neither would there significantly greater reason to entertain appeals in death penalty cases than in any others. So long as guilt is certain, technicalities are relatively insignificant. Automatic review to entertain proofs of non-proof of guilt should be heard, but there is no reason to hear any further appeals than that. Under dual standards, the death penalty really would become fast, accurate and cheap.


XII   Death and the prisons

Interestingly, Judge May's account of Hale's rationale for slanting five to one in favor of freeing the guilty was precisely because the penalty in question was death. He greatly decried the same slant in favor of the defense being applied when lesser punishments are at stake.(33) Today we perhaps are rich enough to extend the "doctrine of leniency" even further, extending it to lesser crimes while punishing by death only in cases of full certainty (not just 83% certainty). But we will never get beyond the need for the death penalty itself, and it isn't just a matter of money.

True, there is no less worthwhile way to spend a million plus dollars of scarce public monies than on housing an un-releasable criminal for life.(34) But more importantly, our policy of long term imprisonment for the vast majority of the most depraved criminals has turned our prison system into one of the most extensive systems of cruel and unusual punishment in history, full of beatings, murders, homosexual rape and gang culture.

We all understand the need to take the trash out of society but it is just as important to take the worst trash out of the prisons if we want prisons to be humane places where lesser criminals can be rehabilitated. That means applying the death penalty to everyone who has been determine to be certainly guilty of unmitigated violence of great extremity. In particular, it should be applied those who are certainly guilty of being predatory in prison. For instance, everyone who is found certainly guilty of rape in prison should be given the death penalty. In America today, far more men are raped each year than women, primarily in prison, and while female rape victims are generally raped only once, many prison rape victims are raped repeatedly by numerous people for long periods of time.(35) Neither the predators nor the victims are being rehabilitated. Prisons need to be safe. That means prison predators need to be dead. There is no other place for them. Criminals in general are the least propitious of all students. Only if we remove the most vicious elements is there any possibility for the rest.

A harder question is whether we should go further and apply the death penalty to recidivist perpetrators of lesser acts of violence or even to non-violent recidivists. If someone is going to keep burgling, he can't be let out of jail. Right now the alternative to keeping a prisoner in jail for life is an even more expensive death penalty process, but once the death penalty becomes fast and inexpensive we will have a monetary choice to make: whether we owe a million plus dollars of prison expenses to someone who already owes far more to society than he could even begin to repay. Estimates are that a million dollars of tax money can be spent saving numerous innocent lives (to say nothing of how much it costs in innocent life to raise a million dollars in taxes).(36) Unless we value the life of a criminal more than we value multiple innocent lives it seems we should impose the death penalty on all un-releasable criminals, no matter their crimes.

Of course, we don't have to impose all the severity that separating the certainly guilty from the almost certainly allows. Its just that, no longer having to moderate punishments out of concern for the innocent, quite a bit more severity is called for.


XIII   The death penalty and punishment for perjurous recidivism

The two enhancements to criminal justice that are enabled by certainty (punishment for perjurous pleading and swift and broadened application of the death penalty) reinforce each other at some points and conflict at others. They work together where the aggravating circumstance of certain recidivism might turn a less than death penalty crime into a death penalty crime. Are there such cases? Simply on the punishment front it seems plausible. Contrast the person who is still trying to get away with child molestation with the person who does not want to get away with it. The latter we might want to let live out his life in jail. Mindset makes a big difference in what a person deserves.

More generally, however, the death penalty scheme and the punishment for purjurous pleading scheme get in each other's way, creating somewhat of an either-or situation.

Recall the ideal incentive structure worked out for the punishment of perjurous pleading. It was seen to be optimal to have guilty pleas punished the same as verdicts of almost certain guilt. But this scheme won't work for death penalty crimes. We can only deal death when guilt is certain. Thus the severest punishment that can be applied for a less than certainty conviction is life without the possibility of parole. If the penalty for pleading guilty is to be the same as the punishment for being found less than certainly guilty, then the severest punishment for pleading guilty would also be life without the possibility of parole. This would allow the most obviously guilty capital criminals to evade the death penalty simply by pleading guilty. Deterrence and punishment would both be greatly weakened and instead of taking the trash out of the prisons, we would be leaving our prisons full of the worst predators.

None of this would be for any good reason. The reason not to apply the death penalty for convictions of almost certain guilt is not because death is too harsh but because it cannot be taken back. This constitutes an artificial limit on harshness that should not apply when guilt is certain. But if we go ahead and deal death to those who plead guilty to capital crimes, that undoes the incentive to plead guilty.

There is a resolution to this dilemma, if we are ready to be severe enough. It is possible, in capital cases, to punish guilty pleas with death and still maintain separation between the punishment for pleading guilty and the punishment for a verdict of certain guilt. We just have to create for the latter a punishment worse than death. Adding time in jail before issuing death would not suffice, as that would act to counteract the punishment of death. Neither would gruesome forms of death be appropriate, as they would tend to debase society. What would work is a to apply, prior to execution, a severe course of corporal punishment, specifically as punishment for perjurous pleading. Thus, for instance, a defendant who is found certainly guilty in a capital case might be caned forty times a day for two months, then hung.

Whether such a scheme would require amendment of the Eighth Amendment prohibition on cruel and unusual punishment is uncertain. On the one hand, corporal punishment is a traditional punishment, especially in military settings, and as a matter of logic, intense but brief punishment might well be less cruel than extended low intensity punishment, which necessarily unplugs a person's entire life. With youths in particular, for whom juvenile detention becomes an education in crime, truly frightening corporal punishment could well be best way to keep errant youth from getting sucked into a life of crime, making it the least cruel punishment. On the other hand, inflicting pain because death is not harsh enough would, in the context of a one verdict system, be just the kind of thing that the Eighth Amendment was meant to ban. But would it still be cruel under a system of multiple verdicts?

The Eighth Amendment was promulgated on the assumption that guilty verdicts will not always be correct. When we can separate out the certainly guilty, the implications of moral logic change. All punishments are cruel in some ways, or they would not be punishments. We just want them to be proportional and not gratuitous. When innocence remains a possibility, mercy is called for and proportionality sets a much lower limit. In contrast, when guilt is certain, there is no call for mercy, and proportional punishments for truly evil crimes goes up considerably. Neither is the infliction of corporal punishment gratuitous. It has a very important role to play, punishing the additional crime of perjurous pleading, so as to deter it. In any case, our options are clear. If we cannot stomach dealing a dose of certainly deserved corporal punishment to those who plead innocent to the worst crimes and are found certainly guilty then we will just have to forgo some of the efficiencies of the two verdict system, essentially excusing continued efforts by the guilty to get away with the very worst crimes.


XIV   The requirement of malim in se

Before any enhancements to severity based on certainty of guilt can be applied, there is a second criterion of certainty that must also be met. There must also be a certainty that the act that is charged is properly criminalizable. Remember, the Fifth Amendment right not to incriminate oneself came into being as a protection for those who are guilty as charged of violating unjust laws (the English blasphemy laws, prosecuted by Star Chamber and the Court of High Commission).(37) Insuring that certainty standard punishments do not infringe on this class of protection requires an expansive concept of what might not be properly criminalizable. In general terms, this would include sex, drugs and rock and roll; guns, abortion and any gray areas of self-defense; and anything that could conceivably be protected by constitutional, common or natural law. Natural law should also be considered a positive requirement. A violation should have to be wrong in itself (malim in se, or mala in se) rather than wrong because laws and social arrangements make it wrong (mala prohibitem) before certainty punishments can apply. It is not rule-breaking per se, or antagonism to over-reaching government, or any other spirit of liberty that is to be answered with the greatest severity, only immanent wrongdoing. This cannot be overemphasized.

We live in a nation that 1) is founded on the belief that, while government is sometimes a necessary evil, it is always an evil, ever seeking improper power, never to be trusted, and 2) has a government that proves the wisdom of this distrust by perpetually committing both mala in se and mala prohibitem on the grandest scales. Mala prohibitem is the overwhelming tendency of government, as majorities exercise contempt for the constitutional limitations that would bind them, and they have little trouble finding Justices who are glad to operate on the theory that the Constitution means whatever the majority wants it to mean and whatever the Court says it means.

We see this in our own United States where over half the federal government is clearly unconstitutional. There is no enumerated power of the federal government to socialize retirement, or health care, or to lay down detailed safety rules for every occupation, etcetera ad nauseum. The majority does not care that it is prohibited by law from doing these things. As bad as this is, Mala in se is government's even more basic prerogative, as majorities and well placed interests are free to declare immanent wrong to be legally right to their hearts' content. More theft is regularly perpetrated through the offices of government than by all private crime combined. The only way to free certainty convictions from this taint of legal injustice is to declare mala prohibitem to be insufficient for a certainty conviction. Jurors must have no rational doubt that an offense is mala in se before a certainty verdict can be reached.

Thus if a parent kidnaps his child in violation of a court custody order, the fact that the kidnapping is in violation of a court order is irrelevant to the question of whether the act was certainly wrong. If a descendant of black slaves engages in a tax protest, withholding taxes in compensation for the forty acres and a mule that an ancestor should have received under the Thirteenth Amendment's guarantee of equal protection of the laws, the fact that withholding taxes is illegal is irrelevant to the question of whether this action is certainly wrong in itself. If this tax protester can convince the Court that he is indeed owed such compensation under the Thirteenth Amendment, he may even avoid punishment at the less than certainty level, but at the certainty level, any doubt about where justice lies demands acquittal.

Requiring certainty about immanent wrong would add a dimension to the trial process, but for the vast majority of violent crimes, and certainly for any heinous crime, it creates no obstacle to applying certainty punishments. Any unmitigated act of any serious violence is certain mala in se. There are cases, however, where the mala in se requirement would occlude certainty punishments for some of the most serious acts of violence. Consider Ellie Nesler's courtroom shooting of Daniel Driver, the man who, after being convicted of molestation and spending token time in jail, went on to molest her son.(38)

Nesler broke the law but her action was not wrong in itself, only wrong given the state of social arrangements, where we have in some ways improved on natural justice by setting up a system for insuring that a suspect is in fact guilty before meting out appropriate punishment. When she took natural justice into her hands she bypassed these safeguards, running an increased risk of either punishing the innocent or of punishing the guilty excessively. Responsibility for that risk is on her. If she had shot an innocent man in this circumstance then she would be guilty of mala in se. She did not shoot an innocent man so she is only guilty of mala prohibitem and hence could not properly be subject to certainty punishments.

The whole idea of a certainty standard is to draw the necessary distinctions so that the full severity called for by justice falls only on to those whose wrongdoing we can be certain about. Thus the very concept of certainty forces the natural law foundations of American liberty to the fore. Unless we can be certain of immanent wrong we lack the necessary certainty to impose full severity.


XV   The direct protection of liberty

The criterion of mala in se, and what can be said about natural right, lies at the heart of a second possible scheme for greatly advancing the efficiency of our system of criminal justice. The full scheme takes a substantial amount of work to describe (see my paper "Direct Protection of Liberty"), but the two schemes are complementary in a synergistic way, so I will take a moment here to briefly describe this second scheme and how it multiplies the leverage over the guilty that the dual verdict scheme creates.

There are two ways to protect liberty. It can be protected directly, by enumerating areas of protected activity and striking down laws that attempt to criminalize any protected activities, or it can be protected indirectly, by placing restrictions on law enforcement. Our own system of liberty is mostly indirect. The only protected activities are those listed in the First and Second Amendments. The rest of our protections are procedural. But suppose we were able to articulate the full ideal of protected liberty, and protect it directly in the Constitution. But suppose we could figure out the full contours of what society must, as a matter of natural right, allow ("natural right" referring to what we have grounds to assert about what is right morally). If this full ideal of liberty could be articulated and protected directly in the Constitution, liberty would be far better protected than under our present system of indirect protection. The protection of liberty would be systematic instead of hit and miss, and indirect protections, having been superseded, would become superfluous, allowing them to be greatly relaxed. We could untie the hands of law enforcement, allowing evidence gathering and crime control to be greatly enhanced, while still providing greatly increased protection for liberty.

As with the present scheme, such an advance would be particularly valuable at a time when we are faced with a war on terrorism that creates powerful inducement to relax restrictions on law enforcement. Instead of giving up liberty as the price of these eased restriction, we can get rid of far more restrictions than anyone has begun to contemplate while greatly strengthening our protection of liberty. All we need to do is figure out how to articulate the full ideal of liberty so that we can protect it directly, and as it turns out, we can. (See my "Direct Protection" paper to find out how.) All I want to point out here is the complementarity between the direct protection scheme and the multiple verdicts scheme.

Under the direct protection scheme, evidence gathering is unleashed. Any kind of non-invasive universal, random, individual or group based search would be enabled, and many kinds of invasive searches would be enabled as well. Suspects would not be able to hide behind lawyers, but would be required to answer questions related to any properly criminalizable activities. In these and other ways, evidence gathering would be greatly enhanced. That enhanced evidence gathering then become the weight that gets multiplied up by the leverage that the multiple verdicts scheme exerts over the guilty. The direct protection scheme and the multiple verdicts scheme each has its own tremendous crime fighting power, and the two join multiplicatively.

Once both are implemented, crime is toast. Nobody is going to get away with it any more, which is certainly what we are going to need if we are to overcome the modern threat of terrorist individuals wielding weapons of mass murder. Not letting them get away with mass murder is not enough to beat them, but if we can get to the point of not letting them get away with preparation for mass murder, then we have a chance. The price for such a thoroughly empowered immune system? Much better protection for liberty and much less risk to the innocent.


XVI   The less than certainty standard of guilt

Our present "reasonable doubt" nominal certainty standard of guilt is a de facto less than certainty standard. In the two verdict system, this would be transformed into an actual certainty standard by complementing it with an explicit less than certainty standard (possibly explained in terms of Hale's maxim or Blackstone's maxim). An explicit less than certainty standard would be new territory and some studies to gauge how jurors would treat different such standards would certainly be warranted, but moral logic suggests that an explicit less than certainty standard would have to be a great improvement upon the moral confusion wrought by the reasonable doubt standard. This can be seen in the steps being taken today to try to make the reasonable doubt standard work.

Jurors are not the only ones who have moral sense. Judges and legislators (and the voters who elect them) are also driven by their moral sense to try to moderate the unreasonableness of our nominal certainty standard of guilt. One strategy has been to eschew explication of the historical meaning of "beyond a reasonable doubt" or "moral certainty."(39) No intent to moderate the standard is acknowledged, but the impact can be predicted by moral logic. Absent any explication of terms, the only logical way to interpret the adjective "reasonable" is as a reference to some reasonable amount of doubt that one should tolerate consistent with conviction, or perhaps to amount and kind, but not to kind alone, as it actually refers.(40) The untutored mind would never adopt an interpretation of "reasonable" that is the epitome of unreasonable.

Similarly for "moral certainty." What could the modifier "moral" mean in reference to the kind of certainty one should feel when deciding to convict rather than acquit? The last thing the untutored mind would consider is that "moral" certainty is a command to violate the first principles of morality and discount the risks of letting a guilty person go free. Thus one way to escape the unreasonableness of the nominal certainty standard is not to explain it. This obfuscation is in contrast to what would work under a two verdict system. Lawrence Solan's survey of the evidence suggests that, for achieving a proof standard, it is critical to emphasize the positive requirements of proof.(41) Under a two verdict system this would be the obvious strategy, because under that system a true certainty standard is desirable. Unlike with the present single certainty standard, the object would be to be clear.

Another tactic that has been used to evade the literal meaning of the beyond reasonable doubt standard is various attempts to specify what reasonable doubts are not. It is hard to go very far in this direction without violating the actual meaning of reasonable doubt, so the higher courts have had to strike some of these attempts down.(42) Missouri has gone to a slightly different tactic, giving a positive explication of reasonable doubt that does not suggest a certainty standard, instructing that: "A reasonable doubt is the kind of doubt that would make a reasonable person hesitate to act."(43) Hesitate when? Before or after deliberation? A reasonable person, abiding by the moral principle of accounting all value, would account the costs of letting a guilty person go free and therefore might not hesitate to convict so long as the defendant seemed substantially more likely to be guilty than innocent. One might hesitate at first but it is the nature of the deliberation that jurors are engaged in to reach a decision whether, in the end, one must do or forbear. Time will tell whether the Supremes will let the states get away with these attempts to move away from a nominal certainty standard.

The reasonable doubt standard is caught between a rock and a hard place. Its literal call for certainty is profoundly immoral yet every attempt to make it reasonable by shaving ten percent off is a direct violation of its essential meaning. The best that can be hoped for is a firm sounding heap of confusion that will appeal to duty and morality in a way that gives some latitude to jurors' moral sense while asserting every grave caution. The problem is that one can never be certain which elements the jurors will pick up on, the literal demand for certainty or the un-explicated calls for reasonableness and morality. This leaves tremendous room jurors to read into "reasonable doubt" what they want to read into it, depending on the whatever biases or snap judgments a case elicits from them. If they want to acquit then the no-reasonable-doubt standard is extraordinarily accommodating. If they want to convict, "reasonable" can bend that way too. At best the standard creates a serious conflict between jury instructions and moral sense. At worst it is a blank check for capriciousness.

In contrast, an explicit less than certainty standard of guilt seems like a far less arbitrary way to achieve a less than certainty standard. The idea of being 80% certain or 90% certain may be hard to apply precisely but the concept itself is perfectly clear and not subject to widely varying interpretations. If we were to do nothing else, swapping the reasonable doubt standard for an explicit less than certainty standard would seem to be a clear improvement in due process.

Another possibility to consider is that jurors might treat an explicit less than certainty standard differently depending on whether it stood alone or was part of a two verdict system. When paired with a certainty standard, jurors might look at it as a dumping ground for cases where the defendant is probably guilty but not at a certainty level or even an almost certainty level. If there were an effect in this direction it could likely be counteracted by tinkering with the level of certainty demanded for a finding of almost certain guilt. On the other hand, it might turn out that the experience of unbiased attention to a true certainty standard would heighten jurors' level of caution and analysis when applying the less than certainty standard. Ideally it would heighten their objectivity, promoting an accurate assessment of likelihoods. Here is where some laboratory studies would be helpful. Compared to ruling on a single 90% standard of guilt, does ruling on both a 100% standard and a 90% standard raise or lower the likelihood of conviction on the 90% standard?

At the least, jurors in a two verdict system should be less prone to capriciousness, which would add to the increased justice for both the innocent and the guilty. If Hale was right that we should worry more about punishing the innocent than about freeing the guilty, the decreased risks to innocent defendants under the two verdict system should be a decisive for preferring it. In terms of harms to the innocent, there is simply no comparison between a process that is well designed to withhold the severest punishments until a true certainty standard is met and one that applies the severest penalties even when substantial uncertainty remains. True, greater harm would be inflicted on the vanishingly small number of innocents who are found certainly guilty in a two verdict system, but this would weigh as nothing against the very large numbers of innocents (one out of five marginal cases, if the Hale standard is being followed) who would have their injuries greatly decreased, some of whom would otherwise receive the very severest penalties. Thus by every measure, harms to the innocent should be greatly decreased.


XVII   A third standard for terrorist threats

When the risk posed by mistakenly freeing the guilty becomes astronomical, as is the case with terrorism, serial murder and serial sex crimes, it might make sense to ask juries to hand down a third verdict on a still lower standard of guilt, with still lower penalties, aimed primarily at being able to monitor the defendant's activities so that terrorist threats can be interdicted and serial crimes prevented or at least not gotten away with. Suppose Theodore Kaczynski had gone to trial and prosecutors were only able to prove with 50% certainty that he was the Unabomber. Must a 50% Unabomber be allowed to walk, with no constraints and no way to even keep track of him? Must he be allowed to secret himself again to where, if he is the Unabomber, he can continue his career of serial murder?

A full accounting of value says "no." It says that the risk of punishing an innocent person should be weighed against the risks posed by freeing the guilty. In terms of Kaplan's analysis, when the dis-utility of freeing the guilty rises, the appropriate burden of proof goes down.(44) This is a standard practice of all legal systems in time of war, when military law takes over and risk assessment becomes the rule, even allowing for pre-emptive incarceration (the holding of suspected enemy combatants). The difference is that, where military law reacts to threats with a rule of caution, putting the interests of the many ahead of the one, civilian law must continue to react to uncertainty with a rule of leniency. In a war on terrorism we will need to use both, and one possibility is to employ a third lower standard of guilt within the civilian process.

If and when a lower standard is employed, less severe punishments would have to go along with it. There is a high likelihood that anyone convicted only of "probable involvement" in terrorist activity will be innocent, which implies that hardships should be mitigated. Indeed, punishment per se should be eschewed. Sentences should be limited to preventative measures, such as barring "convicted suspects" from the instruments of crime and forcing them to submit to whatever is necessary to allow law enforcement to monitor their movements while interfering with their rights and liberties as little as possible. The failure to follow this principle was the one really untenable aspect of the internment of Japanese Americans in W.W.II. Harm to innocents was not minimized, as with the failure to protect the property of the internees.(45)


XVIII   Would multiple standards of guilt be constitutional?

A couple of constitutional questions arise, both about multiple standards of guilt and about the avenues of prosecution that multiple standards enable. Consider multiple standards first.

The Fifth and Fourteenth Amendments require that the federal and state governments respectively not deprive any person of life, liberty or property without due process of law. Under various doctrines of "incorporation" the Supreme Court has progressively included more due process elements from the Bill of Rights into the concept of due process that it requires the states to adhere to. Justices Cardozo and Frankfurter were the leading proponents of a flexible doctrine of "selective incorporation," allowing the states leeway to establish their own due processes of law, so long as those processes were consistent with "the concept of ordered liberty."(46) Justice Black was the great proponent of "total" and "specific" incorporation, looking to the due process requirements in the first eight amendments for specific guidance as to what due process requirements were to be imposed on the states. Under either of these interpretations, multiple standards of guilt should pass muster.(47)

Nowhere does the Constitution specify what standard of guilt is to be applied in criminal cases. So long as the proposed innovation does not violate the concept of "ordered liberty" there should be no obstacle to adopting it at the federal level or by the states. Multiple standards of guilt clearly do satisfy ordered liberty because they yield improvements across the board. The change is efficient in the technical sense, enhancing both crime control and the protection of liberty at the same time. Liberty is better ordered, on every front.

The hitch is that the Court, in In re Winship , has already seen fit to impose the "beyond reasonable doubt" standard on the states.(48) Still, under either doctrine of incorporation, this imposition would have to make way to an improvement in ordered liberty. This is obvious for the "selective incorporation" doctrine, which takes ordered liberty as its criterion. Indeed, the primary rationale for allowing states leeway is so that they can pursue improvements in their systems of liberty. As Justice Frankfurter put it: "A construction which gives to due process no independent function but turns it into a summary of the specific provisions of the Bill of Rights would... deprive the States of opportunity for reforms in legal process designed for extending the area of freedom."(49)

The doctrine of "specific" or "total" incorporation has a very different motivation, but it would also call for flexibility on multiple verdicts. The rationale for "specific" incorporation was to place Supreme Court Justices in a straight-jacket, limiting their discretion to pursue activist agendas.(50) Justice Black feared that the doctrine of selective incorporation placed Justices in a position where they would be forced to exercise unbridled discretion, divining the "natural law" requirements of ordered liberty.(51) But the objective of reigning in Court activism could never justify imposing on the states any non-enumerated elements of due process. The Court's imposition of the beyond reasonable doubt standard would be activism even beyond the doctrine of selective incorporation and a fortiori must be rejected by the doctrine of specific incorporation. Thus there is no coherent position on incorporation that would conflict with the adoption of multiple standards of guilt by the states. Similarly at the federal level, if Congress wills the change, the Court should only override if the Constitution demands it, but the Constitution is silent on the mechanisms of due process in this area. Efficient changes (yielding improvements all fronts at once) should clearly be allowed.

Another possible friction point is the Fifth Amendment right not to incriminate oneself. As yet the requirement to enter a plea has not been thought to conflict with this right. On the other hand, the dual verdict system would explicitly make it a crime to enter a false plea. Does this make a difference? It does not seem that it should. The right not to incriminate oneself is not a right to lie with impunity. The defendant is free to plead however he wants. If he chooses to lie he runs the risk that the evidence will reveal his lie, which is properly punishable. Still, this is new territory and Fifth Amendment challenges would certainly be mounted.

If the dual verdict system were held to violate the Fifth Amendment, that would be decisive at the federal level, meaning that a multiple verdict system could only be implemented for federal trials by passing a constitutional amendment. At the state level, the issue would again become the doctrine of incorporation, where "selective incorporation" should again smile on the dual verdict system. The right not to incriminate oneself has a fairly recent and well known origin in a specific purpose: to protect those who are guilty of violating unjust laws. So long as the punishments enabled by certainty are not applied when there is any concern that the law that has been violated is unjust then the intent of the Fifth Amendment right is satisfied.

Under Black's "total incorporation," a ruling that the dual verdict system violates the Fifth Amendment would automatically disqualify it for use by the states, which shows why, in the end, Black's "strict" method must give way (as it has) to Frankfurter's flexible one. There are many ways to achieve due process and the desire to put the Supreme Court Justices in a straight jacket is no justification for putting the states in a straight jacket.

Obviously the weight of history is behind the single "reasonable doubt" standard, presenting a formidable obstacle. The danger is that the seeming idealism of our nominal certainty standard would prompt the Court to reject a "lower" standard. Nothing sounds more noble than a certainty standard, so long as all attention is focused only on the protection it affords to those wrongly accused, with no mention of its gross failure to account all value. Interpreting this blinkered idealism into the Constitution is a path that must not be followed and there is no better amicus brief to that effect than Judge May's vigorous denunciation of the march toward a certainty standard in his day.(52)

On the positive side, the two verdict standard offers a way to reconcile a rift in current Supreme Court jurisprudence on burdens of proof. In Victor the Court notes that the philosophical and judicial origins of the "beyond a reasonable doubt" standard imply a literal certainty standard. But modern rulings by the Court on burden of proof more generally invoke a balancing of harms, along the lines laid out by Kaplan.(53) Clearly the Court would like to have its cake and eat it too. It wants the certainty standard, and it wants rationality. A system of dual verdicts would allow it to have both. The harshest penalties would be reserved for cases where certainty was present, while the irrationality of relying on a certainty standard more generally would be avoided.

Ultimately, constitutionality is not the issue. The real hurdle is political, and with political will, any amendments that might prove necessary could be passed. The heart of the matter is this: that by the simple innovation of employing two standards of guilt we can have what certainty enables (severity, without risk to the innocent) when certainty is available, and we can limit ourselves to what less than complete certainty requires (mercy) when almost certain is as certain as we can be. It is time to stop lumping the possibly innocent with the certainly guilty. By pairing the reasonable doubt nominal certainty standard of guilt with an explicit less than certainty standard we can achieve a true certainty standard, allowing us extreme leverage over the guilty while doing much less harm to the innocent. I've got the ballot measure ready to go.

The End

Contact the author at alec@rawls.org

A non-academic version of this paper, How to Safely Decimate Crime, appeared in The Thinker, 12/8/1995. The Thinker's archives are available at www.stanford.edu/group/Thinker. Thanks to Evert N. Fowle for helpful discussions.

Worth a buck of your time?

You pick the amount. PayPal's fee schedule is 30 cents + 2.2%, so make any donations lump sum rather than item by item. To hear more, visit:  The decentralized coordination of intelligence.

Site Links

Home       Latest opinion columns etc.      Lawsuit       Direct Protection       Multiple Verdicts       Book on Republicanism       Illiberal "liberalism"       Decentralized coordination of intelligence       Rebel-Yell       Site search      Contact      Email sign-up       Donate


1. Moral reason just means following the requirements of right reason on matters of value, and the most basic principle of thinking straight is to not forget what you already know. With regards to matters of value, that means, account all discoveries of value, wherever they are enough at stake to be worth accounting. The book on republicanism, from which this article is drawn, will derive all in terms of the principles and implications of moral reason, but instances of moral reason do not in general need to be placed in the larger structure of moral reason in order to stand and compel.  Return

2. This has been suggested a couple of times: by myself (Death Penalty Unacceptable when Guilt not Certain, The Stanford Daily, 5/6/88), by James Grippando (For the Death Penalty, a Super-Proof Standard, The National Law Journal, 2/20/95) and perhaps by others.  Return

3. A June 2000 Field poll of California death penalty views found that at the same time as 2/3 of adults in the state are pro-death-penalty, 3/4 of adults favor a moratorium on executions to study the fairness of the system. See the San Francisco Chronicle (sfgate.com) 2/22/00.  Return

4. Precursors to the modern formulation were "beyond any reasonable doubt" and "beyond all reasonable doubt" (see Lawrence M. Solan Some Doubt About Reasonable Doubt, 78 Texas Law Review, 105, footnote 23 (1999)), also, beyond "any rational doubt" (see footnote 7 below).  Return

5.  Barbara Shapiro gives an extensive account of these origins in chapter 1 of her book, Beyond Reasonable Doubt and Probable Cause: Historical Perspectives on the Anglo-American Law of Evidence, University of California Press, 1991. See in particular p. 7 et seq. Justice O'Connor traces American precedent for the identification of "reasonable doubt" with "moral certainty" to Commonwealth v. Costly, 118 Mass. 1, 24 (1875).  Return

6. 511 U.S. 1, at 11 et. seq. (1994).  Return

7. John W. May, Some rules of Evidence, 10 American Law Review, 642 (1876), at 653-657. For Hale's maxim, see his treatise History of the Pleas of the Crown, 290 (1800). May traced the "beyond reasonable doubt" rule to the Irish Treason Trials where Justice Chamberlain instructed jurors that they must acquit when they have "any rational doubt" about guilt (Bond's case, 27 How. St. Tr. 523 Ire. 1798). See May 657. May's thesis was that the need for lenience in these politically charged trials led to the adoption of an extraordinarily high standard of guilt. May's thesis has been challenged by Morano, A Reexamination of the Development of the Reasonable Doubt Rule, 55 Boston Univ. L. Rev. 507 (1975). For a brief modern review of the history of "reasonable doubt," see Solan, supra, or the Court in Victor. For a thorough review of the historical evidence, see Shapiro, supraReturn

8. 83.33% to be exact (five times the remaining 16.66%).  Return

9. 4 William Blackstone, Commentaries, 358 (9th edition, 1873); Sir John Fortescue, De Laudibus Legum Angliae, 93 (translated by A. Amos, Cambridge Eng., 1825); Starkie Evidence , 756 (8th Am. ed. by Sharswood. 1st ed. 1824).  Return

10. For the survey of judges, see Rita James Simon, Judges' Translations of Burdens of Proof into Statements of Probability, The Trial Lawyers Guide, 103, 1969. For a summary of empirical research on what level of certainty the "beyond a reasonable doubt" standard works out to in practice, see section III of Solan, supra at 119-132.  Return

11. May, supra, at 659 and 662.  Return

12. Id. at 655.  Return

13. Id, at 659.  Return

14. John Kaplan, Decision Theory and the Factfinding Process, 20 Stanford L. Rev. 1065 (1968). Kaplan derives a simple formula for the optimal degree of certainty of guilt which should yield a guilty verdict as: P = 1/(1+ (Dg/Di)) where Dg is the harm ("disutility") of letting the accused go free if he is guilty, and Di is the harm of convicting him if he is innocent (at 1071-72).  Return

15. See supra note 1.  Return

16. May, supra, at 659.  Return

17. Radelet, M.L. and Bedau, H.A., Fallibility and Finality: Type II Errors and Capital Punishment, in K.C. Haas and J.A. Inciardi (eds), Challenging Capital Punishment, 91-112. Newbury Park, Cal., Sage Press (1988). In their book Actual Innocence (Doubleday, New York, 2000) Barry Scheck, Peter Neufeld and Jim Dwyer offer a percentage breakdown of discovered errors into similar categories. See their page 246.  Return

18. IdReturn

19. Sheck, Neufeld and Dwyer suggest that one reason for the appalling low standard of guilt that seems to have been applied in many of the cases they discuss is because the crimes that their clients were wrongly convicted of were the worst kinds of crime, where the possibility of letting the perpetrator off becomes almost unthinkable. Of the Bloodsworth case they write:

If Kirk Noble Bloodsworth had been accused of robbing a bank, there is a reasonable chance the prosecutors would not have taken the case to trial. The evidence against him was that weak. And yet, it is easy to see how the jury convicted him. With such gruesome facts, the barrier to conviction dropped considerably. (Id. at 213.)

It is not obvious that such a phenomenon should occur, since the harms of convicting the innocent also rise proportionately when the crime is of the worst sort, but to the extent that the phenomenon is real, it strengthens the call to move to a two verdict system, so that the impulse to convict can be satisfied without imposing the highest punishments.

In addition to Actual Innocence (id.), see also Michael Radelet, Hugo Bedau and Constance Putnam, In Spite of Innocence, Northeastern University Press, Boston, (1992).  Return

20. See footnote 10 above, citing Solan, supra, at 119-132.  Return

21. Scheck et. al., supra, at 170 and Appendix I.  Return

22. Field poll, supra note 3.  Return

23. Mark Costanzo, Just Revenge, St. Martin's Press, New York, at 67 (2000). Costanzo does not say where he gets the term "super due process" from, but he does offer an illustrative quote from Justice O'Connor: "The Court has accordingly imposed a series of unique substantive and procedural restrictions designed to ensure that capital punishment is not imposed without the serious and calm reflection that ought to precede any decision of such gravity and finality." Thompson v. Oklahoma, 487 U.S. 815, at 856.  Return

24. The idea that technical violations that reveal guilt should be tolerated when punishment for the worst crimes is at stake was put forward by the late Professor John Kaplan of Stanford University. See his article, "The Limits of the Exclusionary Rule," Stanford Law Review, May, 1974, V.26, p. 1027.  Return

25. U. S. v. Calandra, 414 U.S. 338, 347-348 (1974).  Return

26. For a discussion if review boards as a substitute for the exclusionary rule, see Caleb Nelson's chapter "The Paradox of the Exclusionary Rule," in Criminal Justice, edited by Robert Bidinotto, Foundation for Economic Education, Irvington-on-Hudson, New York, 1996.  Return

27. Ambiguity still arises from the possibility that the criminal might be found "not guilty," but at least it can be removed from two of the outcomes our hypothetical criminal is weighing: the consequences of pleading guilty vs. the consequences of being found less than certainly guilty. Between these two, pleading innocent presents only downside risk (providing a deterrent to the criminal) so long as being found less than certainly guilty is punished at least as harshly as pleading guilty.  Return

28. See "Plea Bargaining: An Unnecessary Evil," by Ralph Adam Fine in Criminal Justice, edited by Robert Bidinotto, published by The Foundation for Economic Education, Irvington-on-Hudson, New York, 1996, p. 88 et seq.  Return

29. Admittedly, an outcome does not have to be plausible to weigh heavily on the mind. Objectively, there is no reason that being found certainly guilty when innocent needs to be any more likely than winning the lottery. Yes, some people do win the lottery, and a vanishingly small number of innocents will lose the certainty lottery, but for any individual the chance is simply not worth accounting. Still, just as people dream about winning the lottery, the possibility of being found certainly guilty would be a harrowing lottery ticket for any innocent accused to hold. The best we can do is make sure that, between the realistic outcomes, incentives are unambiguous, and hope that no one is so frightened of the unrealistic outcomes as to steer by them. Here the fact that people buy lottery tickets does not imply that innocent defendants are likely to plead guilty to avoid the possibility of being found certainly guilty. The magnitude and the probability of the payoff may be similar in both cases, but the price of the ticket is a dollar for the lottery while the price of avoiding the trial lottery is criminal conviction. It is reasonable to think that few people would buy lottery tickets at that price.  Return

30. See Leonard W. Levy, The Origins of the Fifth Amendment, (Ivan Dee, Chicago, 1968) Chapter 2.  Return

31. On costs, see Costanzo, supra, chapter 4.  Return

32. On cost of trial, see Costanzo, id. at 62 et seq. On "super due process" see supra note 23.  Return

33. May, supra, at 659.  Return

34. R. Paternoster, Capital Punishment in America, Lexington Books, New York, (1991). Paternoster estimates the cost of life in prison without parole at .75 to 1.1 million dollars.  Return

35. Eli Lehrer, Hell Behind Bars, National Review, 2/5/2001.  Return

36. See for instance, T. Tengs, et al., Five Hundred Life-Saving Interventions and their Cost Effectiveness, Risk Analysis, 15, 369-91. Tengs et al. Analyzed the cost effectiveness of safety expenditures by various government agencies. Some costs per life saved are: Federal Aviation Agency, $23,000 per life; Consumer Product Safety Commission, $68,000 per life; Highway Transportation Safety Administration, $78,000 per life; OSHA, $88,000 per life; EPA, $2,780,000 per life. Only compared to EPA expenditures does sparing the life of a criminal cost less than one innocent life.  Return

37. Levy, supra, Chapter 10.  Return

38. Woman Kills Molestation Suspect, San Jose Mercury News, 4/3/93 (online at bayarea.com).  Return

39. This approach has been taken by the Fourth and Seventh Circuits. See Solan, supra, at 115-116 for a review. Shapiro, supra, offers a brief account of ongoing conflicts over how and whether to instruct jurors in the meaning of reasonable doubt and moral certainty (chapter 1, note 138).  Return

40. One of the themes of May's criticism of the "beyond reasonable doubt" standard is that it is only concerned with the kind of doubt, with no concern about the amount of doubt (another way of saying that it is a certainty standard). May supra at 645, 647, 651, and implicitly throughout.

Note that to say that the "beyond reasonable doubt" standard refers to the kind of doubt, not the amount, is not to deny that at some point amount can become kind. An iota of rational doubt can be small enough that it is irrational to treat it as a real possibility or give any weight to it. Still, it is the kind of doubt that matters. As soon as there is any real possibility that the defendant is innocent, acquittal is required. Amount only matters for determining what constitutes a real possibility.

It is important to clarify that at some point amount becomes kind—that too small a possible doubt is no longer a real doubt—because under the two verdict system the no rational doubt standard becomes the operative certainty standard. This standard is meant to be high, but it is not meant to be concerned with radical skeptical doubts and other "mere possible doubts" as the "Webster charge" phrases it (Commonwealth v. Webster, 59 Mass. (1850) 295). They have to be real possibilities of innocence, consistent with the evidence. On this point Shapiro notes that even the "any doubt" origins of the "beyond reasonable doubt" standard never called for acquitting on the basis of irrational or radical skeptical doubts. Supra at 21 et seqReturn

41. Solan, supraReturn

42. Again, see Solan, id. at 112-115, for a review. For instance, in Cage v. Louisiana the Supreme Court rejected "substantial doubt" and "doubt as would give rise to a grave uncertainty" as explications of reasonable doubt. 498 U.S. 39, 40-41 (1990).  Return

43. Missouri Jury Instructions, Section 3.11, following United States v. Conley, 523 F.2d 650, 655, 8th Cir. (1975).  Return

44. See supra note 14.  Return

45. The issue here is quite subtle. Most loss of property came about through the inability of interned Japanese to maintain the solvency of their pre-war operations. Many other Americans were also forced by the war to part for the duration with their civilian enterprises, but at least they were paid, allowing them to remain solvent. It would not have been appropriate to pay the internees. They were not carrying the fight to the enemy, and they were being housed and fed by the rest of America. But they should have been offered liquidity, that is, loans, so that they could have remained solvent through the war, then gotten back to their civilian lives with their earlier lives mostly intact, only with some loans to repay.  Return

46. See Cardozo's majority opinion in Palko v. State of Connecticut 302 U.S. 319, 325; and Frankfurter's concurrence in Adamson v. People of State of California 332 U.S. 46 (1947).  Return

47. See Black's dissent in Adamson, idReturn

48. 397 U.S. 358.  Return

49. From Frankfurter's concurrence in Adamson, id. at 67.  Return

50. This was the reason for Black's dissent in Griswold v. Connecticut. He was against incorporation of the un-enumerated rights of the Ninth Amendment because of the unbridled discretion this would afford to Justices. Griswold v. Connecticut, 381 U.S. 479, 507 et seq.(1965).  Return

51. Black dissenting. Id. at 69-71.  Return

52. May, supraReturn

53. This point was made by Bell, supra, at 557-8. He cites in particular, Santosky v. Kramer, 455 U.S. 745 at 755-56 (1982); Addington v. Texas, 441 U.S. 418 at 423-24, 427 (1979); Patterson v. New York, 432 U.S. 197, 208 (1977); Mullaney v. Wilbur, 421 U.S. 628 at 698, 700-01 (1975); In Re Winship, 397 U.S. 358 at 363-64 (1970).  Return

Site Links

Home       Latest opinion columns etc.      Lawsuit       Direct Protection       Multiple Verdicts       Book on Republicanism       Illiberal "liberalism"       Decentralized coordination of intelligence       Rebel-Yell       Site search      Contact      Email sign-up       Donate

Hit Counter