Printer Friendly

Is it better that ten guilty persons go free than that one innocent person be convicted?

I Introduction

The relevance of answers depends, of course, on getting the questions right. Sometimes, however, we are not quite sure that we actually know what the question is. The doctrine that underlies the title question to this essay is commonly ascribed to William Blackstone. in Commentaries on the Laws of England (1769), Blackstone wrote that "the law holds that it is better that ten guilty persons escape, than that one innocent suffer." (1) This formulation is not the only version of the doctrine. For example, in his book on Evidence (1824), another British scholar, Thomas Starkie, insisted "that it is better that ninety-nine ... offenders shall escape than that one innocent man be condemned." (2) The various versions of this idea in circulation inspired Jeremy Bentham to make the following skeptical comment in A Treatise on Judicial Evidence (1825):

At first it was said to be better to save several guilty men, than to condemn a single innocent man; others, to make the maxim more striking, fixed on the number ten, a third made this ten a hundred, and a fourth made it a thousand. All these candidates for the prize of humanity have been outstripped by I know not how many writers, who hold, that, in no case, ought an accused to be condemned, unless the evidence amount to mathematical or absolute certainty. According to this maxim, nobody ought to be punished, lest an innocent man be punished. (3)

II A Doctrine and Its Interpretation

I shall take it for granted that no reasonable person could possibly endorse the version implicitly ridiculed by Bentham: "It is better that all guilty persons go free than that one innocent person be convicted." Nevertheless, in the light of Bentham's remarks, it seems equally clear that it would be futile to embark on an exercise in moral mathematics in order to fix and justify the correct ratio of false acquittals of the guilty to false convictions of the innocent. I shall presuppose, moreover, that any ratio one might come up with, say forty-two to one, (4) should not be taken to indicate that it is forty-two times worse to convict an innocent person than to acquit a guilty one. Rather, I think that a more adequate interpretation of the doctrine is to be found in a U.S. Supreme Court case from 1895, Coffin v. U.S. In this case, Justice White, who delivered the opinion of the court, dated the doctrine back to Roman law. He quoted a Roman official who wrote that "it was better to let the crime of a guilty person go unpunished than to condemn the innocent." (5) I take this doctrine, which I shall call the Roman doctrine, to articulate the strong intuition of everyday morality: it is worse to convict innocent people than to acquit guilty people. Significantly, to argue that some phenomena are morally better than some other phenomena is to presuppose that they are comparable with respect to some common moral standard of comparison, for example, a particular conception of criminal justice. There is no further need to presuppose a single scale of units of value allowing for quantification or measurement of how much better it is to reduce the number of false convictions rather than to reduce the number of false acquittals.

Most people take for granted that the Roman doctrine is basically sound, yet it does not follow that it is easy to articulate a rationale for it. Still, in a certain narrow interpretation, it seems as if the doctrine is trivially true. Consider two comparable cases of murder, both of which are wrongly decided. In one case a murderer is acquitted, in another case an innocent person is convicted of murder. According to the doctrine, the outcome in the former case is better, or rather, less worse or less bad, than the outcome of the latter case. How could that fail to be true, we might ask. After all, by punishing the innocent, the guilty person is still out there enjoying his freedom. In the TV series The Fugitive, when Dr. Kimble was wrongly convicted for the murder of his wife, he was the only person who kept looking for the one-armed man. In contrast, by acquitting the guilty, the guilty goes free--and the innocent has yet to be unjustly punished. How could the outcome of one case, one false conviction plus one failure to apprehend the guilty, fail to be worse than the outcome of the other case--the release of the guilty? (6)

This kind of reasoning, however, fails to account for cases in which we already know that there are no alternative suspects. Generally, this is the situation whenever actus reus descriptions are properly ascribed to agents; the crucial question is whether their actions satisfy the mens rea requirements as well Consider another set of comparable cases, both of which are wrongly decided. In one case a murderer is acquitted on grounds of self-defense, in another case a person acting in self-defense is convicted of murder. Clearly, the moral difference between the two cases cannot be adequately explained in consequentialist terms by merely counting the number of direct harms.

III The Reiman--van den Haag Debate

A sophisticated attempt to establish an alternative explanation is provided by Jeffrey Reiman in his interesting exchange with Ernest van den Haag. (7) Reiman identifies three propositions of what he calls strong retributivism. In combination, they yield something like the following argument:


(1) punishing an innocent person for a given crime is as bad as committing that crime against an innocent person, and

(2) failing to punish the guilty is as bad as committing his crime against an innocent person,


(3) failing to punish the guilty is as bad as punishing an innocent person

On the retributivist principle that the punishment shall fit the crime, (1) seems intuitively plausible, but (2) is very hard to accept. How could, say, failing to execute the guilty for a capital murder be as bad as his murder of an innocent person? As Reiman points out, adherents of strong retributivism owe us an account in terms of which the direct harms of the two alternatives (false acquittals and false convictions) are identified and ranked. While it amounts to a tautology to say that the innocent is harmed by the infliction of punishment, who is directly harmed by the non-punishment of the guilty? In a capital murder case, it makes sense, of course, to argue that an additional harm is inflicted on the relatives and friends of an innocent victim whenever the victimizer is acquitted. Nevertheless, it is downright implausible to insist that the additional harm of acquittal equals the harm inflicted by the perpetrator's crime, which, according to the retributivist conception, equals the harm of punishing the innocent. And this point generalizes to the law-abiding citizen, in Reiman's words,

[T]he question reduces to that of whether any given law-abiding person would suffer more from being a victim of undeserved punishment (or of the comparable crime) than from having the criminal who victimized him go unpunished. And that is the same question as the one posed previously to the actual victim of crime. (8)

In contrast, according to van den Haag's conception of retributivism, the duty to punish the guilty and the duty to acquit the innocent are "ends in themselves, independent of the losses or gains of beneficiaries." (9) It is consistent with this version of retributivism to argue, as van den Haag does, that the question of which duty has moral priority--to punish the guilty or to acquit the innocent--cannot be answered.

That this is the case is explainable in terms of Judith Jarvis Thomson's aggravation principle of stringency. According to Thomson, the stringency of a claim, or a corresponding duty, must be taken to vary with the harm that is inflicted on the claim holder whenever his claim is infringed. (10) Needless to say, any theory of duties and rights that fails to address the problem of stringency is radically indeterminate. The indeterminacy of van den Haag's retributivism hinges on his claim that the ranking of crimes and their punishments require cardinal measurement: "The retributivist theory of punishment tells us that punishment must be deserved, but cannot tell us how much punishment crime A deserves, or how much more punishment it deserves than crime B." (11)

This objection fails as an objection to what Reiman elsewhere has called proportionate retributivism. (12) Proportionality is readily recognized as a fundamental principle of criminal justice, regulating the amount of punishment to be meted out for criminal wrongdoing. Impartial justice is frequently visualized as a blindfolded woman carrying a pair of scales. Let me offer an example of this idea: Smith's harmful action, the intentional, unjustified and inexcusable killing of Jones. Prior to the murder there was moral order; the scales were evenly balanced. When Smith's wrongful act of murder is put in one of the scale's pans, the scale tilts, and the moral order is disturbed. Although proportionate retributivism may ultimately fail to explain why we are morally warranted in putting an act of punishment in the other pan in the first place, it explains why an act of just punishment will bring the scales back into horizontal harmony. This is because an act of punishment essentially articulates moral criticism of behavior, and hence the strength of our criticism of Smith must carefully reflect the seriousness of his wrongdoing. Or, as the slogan will have it, justice requires that the punishment fit the crime.

The seriousness of Smith's wrongdoing is a function of the harm he has inflicted on Jones and the extent of his culpability in inflicting it, whether he intentionally, knowingly, recklessly, or negligently brought about the death of Jones. In contrast to proportionate retributivism, the doctrine of lex talionis (an eye for an eye, a tooth for a tooth, a life for a life) contends that the scales tilted by Smith's intentional killing of Jones can be re-balanced only by the state's intentional killing of Smith. Although some proportionate retributivists might argue that the death penalty should fix the upper bounds of the penalty scale, there is, as Andrew von Hirsch has shown, no insoluble problem of ordinal proportionality. (13) It seems possible to reach agreement on, first, the ranking of crimes and punishments in terms of seriousness and severity, respectively; second, the appropriate coupling of the two rankings, so that the most severe penalty is reserved for the most serious crime, and so on; and, third, the appropriate spacing of the penalties, so that if crime Y "is considerably more serious than X but only slightly less so than Z ... there should be a larger space between the penalties for X and Y than for Y and Z." (14)

Consequently, within an ordinal framework it makes sense to say that since (1) punishing an innocent person I for crime x is as bad as committing crime x against I, and (2) failing to punish a guilty person G for crime x is less bad than committing crime x against I, then (3) failing to punish G for crime x is less bad than punishing I for crime x. However, van den Haag has a further and more telling objection against retributivism:

[T]he lex talionis only concerns the victim's right to revenge or compensation for the harm suffered. But the harm retribution is concerned with is to the legal, social, and moral order, not to the victim. Retribution is not a tort action. Punishment may be imposed independently of the suffering of any specific victim, or even when no one individual was harmed by the crime. (15)

Although van den Haag's remarks on retributivism are hermeneutically inadequate, they do contain the valid observation that punishable harms are not reducible to harms inflicted on individual victims. Consider, for example, burglary and theft. When a burglar breaks into somebody's house in order to steal, the burglarized person suffers from the economic (and psychological) loss of goods stolen and windows broken; the emotional turmoil caused by the knowledge that strangers have invaded one's private sphere; and so on. Crucially, however, to the extent that individual crimes induce citizens to believe, truly or falsely, that burglaries are becoming more frequent, law-abiding citizens may suffer from fear of being burglarized as well, incurring further costs of preventive measures to reduce the likelihood of victimization. To these more or less tangible harms must be added the abstract harm of burglars taking unfair advantage of law-abiding citizens' self-restraint.

Because the fears of law-abiding citizens, and the preventive measures they feel compelled to take, are generated by beliefs about the aggregate level of crimes, it would be unfair, on retributivist grounds, to ascribe full-fledged responsibility for those harms to individual wrongdoers. Still, it would not be unreasonable to argue that the fears and costly counter-measures generated by high-level crimes like burglary and theft should have some bearing on the ordinal ranking of those crimes, justifying increases in punishments without compromising principles of proportionate retributivism. Likewise, since the benefits of a well-ordered criminal justice system are the aggregate effects of citizens playing by the rules, the unfair advantage constituted by the refusal of wrongdoers to respect the rights of other people while free-riding on their self-restraint might be properly included in the moral assessment of their acts as well. This was suggested by Reiman in his debate with Louis Pojman on the death penalty. (16)

Clearly, then, it is possible to accommodate van den Haag's objections against retributivism without undermining the Roman doctrine. The simple explanation is that the inclusion of social harms in the ordinal assessment of crimes is canceled out by the proportionate, ordinal increase in punishments. So, because it is still true that (1) committing a crime against an innocent person is as bad as punishing an innocent person for a comparable crime, and because it is still false that (2) failing to punish the guilty is as bad as committing his crime against an innocent person, it is still not the case that (3) failing to punish the guilty for any given crime is as bad as punishing an innocent person for a comparable crime.

IV The Fallibility of the Criminal Justice System

However, what from a retributivist and nonconsequentialist point of view seems true in the ordinal comparison of harms, need not be true at the level of the criminal justice system when the aggregate effects of false acquittals and false convictions are taken into account. This is the essence of van den Haag's worry: "However, the dilemma before us does not concern choosing between punishing guilty persons and innocents, but choosing between punishing more guilty (and innocent) or fewer guilty (and innocent) persons by means of lower or higher standards of proof." (17) An ideal criminal justice system would be one that consistently convicted each and every guilty accused and acquitted each and every innocent person. Unfortunately, given the fallibility of human beliefs and practices, the ideal system is essentially unattainable. It seems reasonable to assume that the current criminal justice systems of constitutional democracies succeed in convicting a majority of guilty persons and acquitting a majority of innocent persons among the accused. Nevertheless, we know that a fallible system like ours is doomed to generate false acquittals as well as false convictions.

In the technical terms of statistics, false convictions are analogous to what are called Type 1 errors, the error of coming to accept a false theory (by way of rejecting a true, so-called null hypothesis). In contrast, false acquittals are analogous to what are referred to as Type 2 errors, the error of coming to reject a true theory (by way of accepting a false null hypothesis). Which error is worse: to accept false beliefs or to reject true ones? There is no statistical answer to that question; it depends upon the content of the belief and the consequences of acting on it. Generally, any attempt to reduce the probability of one type of error is likely to increase the probability of the other type. Accordingly, an attempt to reduce the number of false convictions by raising the threshold of evidentiary support for any conviction is likely to increase the number of false acquittals. Nevertheless, the doctrine that it is better to let the crime of a guilty person go unpunished than to condemn the innocent can plausibly be taken to imply that the reduction of false convictions must always have moral priority over the reduction of false acquittals. This is so, irrespective of the fact that reducing the number of innocent persons convicted will increase the number of guilty persons acquitted.

It would be nice were we able to establish a coherent rationale for the Roman doctrine, which is more or less taken for granted as self-evidently true--a moral datum "to be argued from rather than to be argued for," as Thomson puts it in another context. (18) We might disagree about whether such a rationale is to be found at all or what kind of status it might have. Still, we should be able to agree on the fact that false acquittals, no less than false convictions, are moral harms that officials of the criminal justice system have undertaken an obligation to avoid, at least if everything else is kept constant. Moreover, I shall presuppose an overlapping consensus to the effect that there can never be cases in which state officials would be permitted to convict a person they knew to be innocent or acquit a person they knew to be guilty. False convictions and false acquittals are moral mistakes, and I shall not waste my time arguing against positions that would allow for the conviction of the innocent under that description, that is, knowingly or intentionally.

V The Presumption of Innocence

In the U.S. Supreme Court case referred to earlier, Coffin v. U.S., Justice White provides the sketch of a justifying rationale for the Roman doctrine. The doctrine, he says, connects with another fundamental maxim of criminal justice, namely, the presumption of innocence. "The principle that there is a presumption of innocence in favor of the accused," White insists, "is the undoubted law, axiomatic and elementary, and its enforcement lies at the foundation of the administration of our criminal law." (19) He proceeds to illustrate how Roman law and English common law were pervaded "with the results of this maxim of criminal administration"--the presumption of innocence--by citing precisely the Roman doctrine and the different versions of it that were ridiculed by Bentham.

The presumption of innocence has been enacted as a crucial part of modern human rights legislation. For example, Article 6(2) of the European Convention on Human Rights declares that "everyone charged with a criminal offence shall be presumed innocent until proven guilty according to law." The flip side of the right of the accused to be presumed innocent until proven guilty is the duty of the prosecution to take on the burden of proof. At first glance there might seem to be a kind of performative contradiction here, since no criminal court case would ever get off the ground unless the prosecutor seriously believed that the accused actually was guilty. Indeed, as Randy Barnett has effectively pointed out, there are good reasons to believe that the great majority of the accused in criminal court cases are guilty. (20) But, he maintains, this fact is precisely "a consequence of having first adopted a presumption of innocence" and cannot therefore count as an argument in favor of rejecting it. The presumption of innocence serves to filter out cases based on weak evidence, leaving the prosecutor with cases that are much more likely to generate convictions. However, it will not do to reverse the presumption of innocence as cases reach the courts. Jurgen Habermas and Robert Brandom have forcefully argued that communicative rationality relies on normative preconditions. (21) For example, when we articulate beliefs in dialogue with others, we thereby undertake to justify the truth of our beliefs and to take part in a practice of "giving and asking for reasons." A criminal court case is initiated by the prosecutor's public articulation of a certain belief, namely, that a person has committed a crime and that he is liable to punishment. Thus, it is for the prosecutor, and not for the accused, to establish, in public, the truth of that belief beyond reasonable doubt.

However, when we do seem to attach great moral significance to the presumption of innocence in criminal court cases, this is hardly fully accounted for by a normative theory of communicative rationality. As Barnett points out, "by adopting a presumption of innocence we necessarily accept a greater risk that guilty people will escape a legal sanction than with a presumption of guilt." (22) But when we knowingly accept that greater risk, it is precisely because we intuitively think that a presumption of guilt, with its greater risk that an innocent person will be convicted, is morally worse. In other words, although the Roman doctrine and its various versions might be said to be pragmatically connected with the presumption of innocence, it is not morally justified by it. Rather, it is the other way around. We seem to adopt the presumption of innocence, partly at least, because the Roman doctrine is pragmatically implied by it. Therefore we still have to look for an independent rationale. (23)

We might call Justice White's attempt to justify the Roman doctrine an internal attempt, that is, an attempt to justify it in terms of a principle to which practitioners within the criminal justice system are already committed. Had we succeeded in this line of argument, we would have avoided any reference to external and inevitably more controversial accounts of criminal justice. An internal principle, however, does not seem to be available. So, in order to compare false acquittals to false convictions, we must invoke external moral standards of criminal justice in order to establish why, if at all, the former is better than the latter. The two main camps of external positions, consequentialism and deontology, share one fundamental assumption, namely, the assumption that punishment is properly inflicted on a wrongdoer when certain conditions of criminal liability are satisfied. Although they disagree on why this is so, and on how severely the wrongdoer should be punished, they agree that both the failure to convict the guilty and the failure to acquit the innocent are, strictly speaking, examples of miscarriage of justice, even if this phrase is commonly reserved for convictions of the innocent.

VI Consequentialist Rationales

It is certainly possible to reconstruct a consequentialist rationale for the moral primacy of protecting the innocent against convictions. Bentham claimed that the suffering of the guilty was worse than the suffering of the innocent. Nevertheless, in A Treatise on Judicial Evidence, even he recognized that false convictions based on weak evidence were likely to become publicly known as actual or potential miscarriages of justice. (24) Hence, not merely the knowledge that an innocent man had been convicted, but also the suspicion that this might be the case, could have a much more disastrous effect on people's trust and confidence in the criminal justice system than the knowledge or suspicions of false acquittals could have. Bentham's observations have certainly been confirmed by modern experience. For example, in 1989, Gerard Conlon, Paul Hill, Patrick Armstrong, and Carole Richardson (the so-called Guilford Four) were acquitted after fifteen years in prison, falsely convicted for having planted a bomb in a Guilford pub in 1974. This scandal reinforced the impression of many British citizens that their system had become repressive, brutal, and dishonest. And in the U.S., the O.J. Simpson case illustrated how distrust of the police might lead a jury to discount apparently convincing evidence because time jury feared that the evidence might have been planted by police officers to frame a suspect they thought was guilty.

Despite such examples, there is something peculiar about Bentham's argument. Crucially, the disastrous effects of actual and potential false convictions are mediated by, or dependent upon, people's beliefs. Those beliefs, however, are not reducible to beliefs about utility or consequences. When people are shocked to learn that an innocent person has been convicted, the fact that they are shocked is not first and foremost explained by their subsequent realization that a false conviction is going to reduce the overall utility of the system. Their shock is primarily due to the fact that something unjust has occurred, something that is wrong independently of whatever else might follow from it.

In other words, the success of Bentham's utilitarian argument relies on the premise that most people's beliefs are non-utilitarian. Derek Parfit is certainly right to insist, in his impressive Reasons and Persons, that a position like Bentham's is not directly self-defeating. (25) Bentham's position relies on an asymmetry between his own beliefs, which he necessarily must hold to be true, and the beliefs of others, which are taken into account merely because of the effects that flow from the fact that other people hold them to be true. Clearly, it is impossible to reconcile the beliefs of someone like Bentham and the non-utilitarian beliefs of everyday morality in a coherent and symmetrical way. An attempt to do that is likely to generate paradoxes that are analogous to the paradox illustrated by the story about Raymond Smullyan, the logician. When Smullyan was asked if he believed in astrology, he replied that he certainly did not; he was born under the sign of Gemini, and people who are born under the sign of Gemini are much too skeptical to believe in astrology. (26)

However, there are other, non-utilitarian versions of consequentialism that escape the problems that haunt Bentham's position. But although they escape Bentham's problems, it is hard to see how they can provide a rationale for the Roman doctrine. Consider the most refined and sophisticated version of consequentialism, the so-called contractarian-consequentialist theory developed by John Rawls in A Theory of Justice. (27) What principles of criminal justice would we accept, in terms of instrumental rationality, were we to choose those principles behind a veil of ignorance, not knowing whether we would end up as criminals or law-abiding citizens when the veil was lifted? Although some traditional principles would likely be endorsed, others would more likely be rejected.

Rawls explicitly maintains that strict liability and reversal of the burden of proof for specific crime categories can be derived from the principle of liberty. (28) According to Thomas Pogge, the parties in Rawls's original position would even endorse a death penalty lottery scheme for drunken driving. (29) Mens tea requirements are irrelevant for criminal liability in drunken driving; the relevant conditions of liability are identified by means of natural-scientific measurements. It is estimated that 10,000 people are killed annually by drunken drivers in the U.S. Although the parties in the original position are denied knowledge of their own positions in society, they are not denied knowledge of social science. Suppose criminological research indicates that a lottery scheme involving the execution of, say, 200 drunken drivers annually would save the lives of 400 persons. Because the parties' deliberations behind the veil of ignorance are exclusively prudential, there is no reason to believe, Pogge plausibly argues, that they would reject a lottery scheme of draconian punishments and/or strict liability conditions for high elasticity crimes. Yet, as Rawls himself has acknowledged, strict liability cannot prudentially be accepted in the original position as a general non-ideal principle of criminal justice. (30) The device of the veil of ignorance is supposed to articulate certain moral assumptions deeply embedded in modern liberal societies, namely, the ideas of persons as free and equal citizens. As Herbert Hart points out in Punishment and Responsibility, the elimination of mens rea requirements would radically undermine "the individual's considered and informed choices" because the capacity to predict and coordinate the outcome of such choices would be "immeasurably diminished" to the extent that the agent's intentions, beliefs, motives, and desires were discounted as proper grounds for penal sanctions. In short, the individual's status as a choosing being would be jeopardized. (31)

What, then, about the standard of proof? Suppose that we are going to live in a criminal justice system that accepts the death penalty. Thus, we could be killed in either of two ways, by acts of murder committed by citizens and by lethal acts of punishment committed by state officials. Since the contractarian-consequentialist perspective does not draw a moral distinction between how death falls on us, both kinds of harms are equally serious. As rational agents, we want to reduce the overall risk of death. By raising the threshold of evidence, say, beyond reasonable doubt, we can reduce the risk of being falsely convicted of capital murder and hence of being executed by state officials. Unfortunately, by raising the standard of proof, more murderers are likely to be acquitted, weakening the incapacitative and deterrent impact of the criminal justice system and thus increasing the overall risk of being killed by acts of murder. To the extent that there is a net reduction in the overall risk of death by lowering the standard of proof, that is the option we should endorse behind a veil of ignorance. In a consequentialist-contractarian account, then, it seems as if we are incapable of providing a rationale for why it is better to acquit the guilty than to convict the innocent. On the contrary, it seems more urgent to punish the guilty at the cost of a higher risk of convicting the innocent.

This is van den Haag's conclusion as well. Conceding that the conclusion would be valid were we merely aiming for a reduction in the overall risk of harm, Reiman nevertheless maintains that we structure our world in terms of various risk zones, distinguishing between zones in which we deem our selves to be reasonably secure (our homes) and zones in which we deem our selves to be much more vulnerable to criminal attacks. Reiman argues, "it is rational for people to prefer a world with some very secure zones even at the expense of the lowest possible overall risk. But then it seems to me that it is rational to want one's government to be one of these very secure zones." (32) However, although it is rational to insist that one's government should refrain from inflicting harm on its citizens, it is equally rational to insist that the government should protect citizens' homes against illegitimate intrusions by fellow citizens. The crucial dilemma--which risk should be minimized at the cost of increasing the other?--is not resolved by the introduction of risk zones.

VII A Puzzle about Evidence

A rationale for our current standard of proof, favoring the non-punishment of the innocent at the cost of acquitting the guilty, must therefore be found elsewhere. There is a puzzle about the standard of proof, however, that we should note before we move on. We are used to attaching a probability of 51 percent to the standard of proof in civil cases, "a preponderance of the evidence," implying that propositions are more likely than not to be true. In criminal cases, in which the standard of proof is "beyond reasonable doubt," we tend to think that the probability should be somewhere between 90 percent and 100 percent, or absolute certainty. When judges are asked to translate "beyond reasonable doubt" into numerical probabilities, they frequently choose 95 percent. Of course, criminal acts are individual, non-reproducible events, so it usually does not make sense to attach probabilities to them. Rather, different probabilities articulate different degrees of belief that a given act actually occurred. Still, even in cases in which it does make sense to say that a certain act occurred with a 90 percent or 95 percent probability, we seem to be disinclined to convict.

Alan Dershowitz, the well-known defense attorney and Harvard law professor, presents his students with the following case. (33) A person is hit and killed by a blue bus. There are no witnesses, but 90 percent of the blue buses in town are run by bus company A and 10 percent by bus company B. Should the court decide that company A is criminally liable because there is a 90 percent probability that the accident was caused by a bus from that company? No, the students say, a 10 percent likelihood of innocence is too great. But suppose, Dershowitz continues, that an eyewitness comes along who is sure he saw that the blue bus had company A's logo on it. That changes the matter, the students reply; an eyewitness testimony is something different. But suppose, the professor says, an expert on witness psychology can establish the fact that this kind of witness is accurate in only 85 percent of cases; would you still decide on the basis of the eyewitness account? Yes, the students answer. So, in one version of the case they decide to convict on the basis of an 85 percent likelihood of truth; in another version of the case they refuse to convict on the basis of a 90 percent likelihood of truth.

Presumably this is so because the students classify eyewitness accounts as direct and individualized evidence, and statistical evidence as indirect and non-individualized. In a similar (although of course not equivalent) manner, there might be something about our way of classifying the harm of (unintentionally) punishing the innocent, and our way of classifying the aggregate harm that flows from acquitting the guilty, that explains why the latter is less serious than the former. In ethics, one way to find out whether this is the case is to carry out thought-experiments. The logic of ethical thought-experiments is similar to the logic of experimentation in physics and the other natural sciences. We attempt to control the impact of factors by comparing pairs of cases in which possibly relevant differences are isolated and "experimentally manipulated," while everything else is kept constant. If a morally relevant difference between pairs of cases persists or disappears, we can more plausibly identify the moral impact of the isolated factor. Some of these cases might strike non-philosophers as bizarre, crazy, or even grotesque, but they nevertheless serve to illustrate matters of principle that we would otherwise have failed to recognize.

VIII Doing and Allowing

In order to isolate the impact of one possibly relevant factor, namely, the number of innocent lives, let us make one concession, by way of stipulation, to those (the consequentialists) who argue that the practice of punishment is justified in terms of its beneficial consequences. The concession we should make is to accept, for the sake of argument, a consequentialist assumption to the effect that for every innocent person who is mistakenly executed by state officials, an indeterminate number of innocent citizens are saved from being killed by acts of murder. Moreover, let us accept another consequentialist assumption to the effect that for every guilty murderer who is mistakenly acquitted by state officials, an indeterminate number of innocent citizen are killed by acts of murder. By making these stipulations by way of a thought-experiment--comparing a case in which an indeterminate number of innocent lives is saved by mistakenly executing one innocent with another case in which an indeterminate number of innocent lives is lost by failing to execute one guilty--we are able to compare the impact of innocent lives.

I suspect most of us are strongly inclined to think that punishing the innocent is morally worse than acquitting the guilty, despite the beneficial consequences that flow from the first alternative and the fatal consequences that flow from the latter alternative. Why is this so? I assume the answer has at least something to do with a fundamental distinction of everyday morality, namely, the distinction we intuitively make between the harm we bring about by doing something and the harm we merely allow to happen by failing to act. In ethics there is the well-known and grotesque example of a small child who is about to drown in a lake unless I, who happen to pass by, pull the child out of the water. Certainly a very small effort is required on my part to save the child from drowning, yet I ignore the child's cry for help and continue on my daily walk. Needless to say, that is an awful, terrible thing for me to do, or rather, not to do. But according to the doctrine of doing and allowing, it is not as terrible and awful as my throwing the child into the lake in the first place and keeping his head under the water to make sure he dies. That is plain and deliberate murder.

One way to explain this difference is to say that in the first version, the child would have died anyway had I decided not to go for a walk to the lake on that day. By contrast, in the second version, the child's death is a direct causal consequence of the terrible things that I did at the lake. Unlike the child's death in the first version, the child's death in the second version would not have happened anyway. (34) Surely I deserve to be punished in both versions, but not as severely in the first version as in the second version.

However, on reflection, it does not seem that the doctrine of doing and allowing is capable of providing a strictly symmetrical explanation of why convicting the innocent is morally worse than acquitting the guilty. Evidently, by pressing the button which activates the electric chair, or less drastically, by inflicting punishment in other more humane ways, the harm that falls on the innocent is a direct causal consequence of something we, as state officials, do, although we act on the mistaken belief that he is guilty. By contrast, by failing to execute the guilty, or, alternatively, by failing to punish him in less drastic ways, the harm that falls on innocent citizens is not quite like the harm that falls on the child when I refuse to interfere in the course of nature to save him from drowning. Rather, the harm that falls on innocent citizens falls on them as the result of our failure to prevent other people's criminal acts.

True, as state officials, we have undertaken an obligation to protect innocent citizens from being victimized by acts of crime, but again, according to the doctrine of doing and allowing, our failure to exercise that obligation is not as bad as our failure to exercise an obligation not to harm innocent persons in more direct ways. By contrast, in their joint introduction to the debate on the Roman maxim, Reiman and van den Haag reject the doctrine of doing and allowing by arguing that the relevance of its distinctions is undercut by the existence of prior obligations. State officials have, they rightly insist, pre-existing obligations both to acquit the innocent and to punish the guilty. But the failure of state officials

to punish the guilty cannot get an automatic moral discount because it is an omission. If their obligation to punish the guilty is as strong as their obligation not to punish the innocent, then failure at the first obligation is as bad as failure at the second. This leaves us with the task of determining the relative strength of these obligations, which is where we were before taking up the acts-omissions distinction. (35)

But is it the case that the doctrine of doing and allowing is rendered irrelevant when there are positive duties to prevent harm? This seems counter-intuitive: a lifeguard who (deliberately) fails to act on his positive duty to save a swimmer from drowning is liable to less punishment than a lifeguard who (intentionally) kills the swimmer by keeping his head under the water. It is precisely the stringency of positive versus negative duties that is addressed by the acts-omissions distinction. According to Thomson's aggravation principle of stringency, the stringency of obligations correlates positively with the harm we inflict on a claim holder whenever we fail to fulfil our obligations. However, the magnitude of that harm cannot be identified independently of how the harm falls on the claim holder: whether it is intended as a means or merely foreseen as a side-effect; whether it is positively brought about or merely allowed; and so on. So although more people might be harmed by failing to convict murderers than by punishing innocents for murder, there are good reasons to reject the contention that state officials' obligation to punish the guilty is as strong as their obligation not to punish the innocent. Reiman and van den Haag are right in pointing out that we are left with the task of determining the relative strength of our obligations, but they are wrong to suggest that this predicament undercuts the relevance of the distinction between doing and allowing. (36)

IX Conclusion

In conclusion, is it better that ten guilty persons go free than that one innocent person be convicted? Yes, I think it is, although, as I have said, I do not think there is much moral significance attached to the number ten rather than, say, the number forty-two. To explain why it is better, or, rather, less bad, however, is not as straightforward as it might seem. Still, I think there is something to be said in favor of the sketch I have provided; at least, I think it points in the right direction.


(1) William Blackstone, Commentaries on the Laws of England (1769; Boston: Beacon Press, 1962), Bk IV, ch. 27, 420 [*358].

(2) Thomas Starkie, A Practical Treatise on the Law of Evidence (London: John Fairburn, 1824), quoted by the U.S. Supreme Court in Schlup v. Delo, 513 U.S. 298 (1995).

(3) Jeremy Bentham, A Treatise on Judicial Evidence (London: J.W. Paget, 1825), 197.

(4) Some readers may recall that Deep Thought, the super computer in Douglas Adams's The Hitchhiker's Guide to the Galaxy, insisted that the correct answer to the ultimate question, "What is the meaning of Life, the Universe and Everything?" was 42. To a somewhat disappointed audience he went on to suggest that "the problem, to be quite honest with you, is that you've never actually known what the question is ... once you do know what the question actually is, you'll know what the answer means" (London: Pan Books, 1979, 136).

(5) Coffin v. U.S, 156 U.S. 432 (1895). See also In Re Winship, 397 U.S. 358 (1970, Justice Harlan concurring).

(6) Within the narrow interpretation, then, not only are the direct effects entailed by descriptions of actions--"acquitting the guilty" and "convicting the innocent"--relevant, but the pragmatic implications of those descriptions are included as well. While the release of the guilty is merely the flip-side of his acquittal, the conviction of the innocent implies that it would be irrational to keep searching for the guilty, since we are acting on the erroneous assumption that he has already been identified. In a sense, the fact that the guilty is still unjustly enjoying his freedom is cancelled out in the moral comparison of the two cases, and we are left with the harm of punishing the innocent. However, we might object that although it is irrational to keep searching for the guilty when we have convicted the innocent, this is precisely what we are morally obliged to do in (false) acquittal cases in which doubt remains due to lack of evidence, procedural error, or jury distrust of the criminal justice system. So, although it is hardly likely that police officers would start looking for a different suspect in cases such as the O. J. Simpson case, some acquittals of the guilty may generate the additional future risk of punishing the innocent. Nevertheless, when comparing the actual infliction of harm now with the possible infliction of harm at some future occasion, we are surely right to prefer the latter rather than the former.

(7) Jeffrey Reiman & Ernest van den Haag, "On the Common Saying that it is Better that Ten Guilty Persons Escape than that One Innocent Suffer: Pro and Con," in Crime, Culpability, and Remedy, ed. Ellen Frankel Paul, Fred D. Miller Jr., and Jeffrey Paul (Cambridge, MA: Basil Blackwell, 1990), 226-48.

(8) Reiman & van den Haag, "On the Common Saying," 234.

(9) Reiman & van den Haag, "On the Common Saying," 242.

(10) Judith Jarvis Thomson, The Realm of Rights (Cambridge MA: Harvard University Press, 1990), 154. Thomson suggests a further refinement of the aggravation principle, the Large-Grain Conception of Stringency (156), but that need not concern us here.

(11) Reiman & van den Haag, "On the Common Saying," 242.

(12) See "Why the Death Penalty Should Be Abolished in America," in The Death Penalty: For and Against, by Louis P. Pojman and Jeffrey Reiman (Lanham: Rowman & Littlefield, 1998), 67.

(13) Andrew von Hirsch, Censure and Sanctions (Oxford: Oxford University Press, 1993).

(14) von Hirsch, Censure and Sanctions, 18.

(15) Reiman & van den Haag, "On the Common Saying," 244.

(16) See Reiman, "Why the Death Penalty." Reiman's contribution is particularly impressive because, as Pojman observes (133), his principled attempt to provide a coherent rationale for punishment combines insights from retributivism, Herbert Morris's "unfair advantage" position, and deterrence.

(17) Reiman & van den Haag, "On the Common Saying," 245.

(18) Thomson, The Realm of Rights, 135.

(19) See Coffin v. U.S, 156 U.S. 432 (1895).

(20) Randy Barnett, The Structure of Liberty (Oxford: Oxford University Press, 1998), 207-8.

(21) Jurgen Habermas, Theorie des kommunikativen Handelns (Frankfurt am Main: Suhrkamp, 1981); Robert Brandom, Making It Explicit (Cambridge MA: Harvard University Press, 1994).

(22) Barnett, The Structure of Liberty, p. 206.

(23) Might it not, however, be possible to endorse the presumption of innocence for reasons that have nothing to do with the Roman doctrine? On the principle of equality of arms, which is implicit in the jurisprudence of the European Court of Human Rights, interactions between citizen and state should be fair in the sense that neither party of a criminal case should enjoy any advantages over the other. Therefore, to compensate the relative powerlessness of the individual citizen, the state accepts its obligation to justify its charges against the suspect, presuming the suspect to be innocent until the state's charges against him or her have been successfully established in terms of the required standard of proof. Still, I think it is difficult to explain why the value of fairness should override the aggregate costs of acquitting guilty suspects without to some extent invoking the Roman doctrine. The same challenge faces a libertarian who accepts the presumption of innocence because it is more compatible with a minimal state. According to a libertarian account, a minimal state is legitimate precisely because it provides protection against force, theft, and fraud and ensures the enforcement of contracts. So, why not accept increased protection at the cost of (unintentionally) punishing the innocent? Again, the Roman doctrine provides a plausible answer.

(24) See Bentham, A Treatise on Judicial Evidence, 197.

(25) Derek Parfit, Reasons and Persons (Oxford: Clarendon Press, 1984).

(26) The Smullyan story is taken from John Allen Paulos's amusing book, Innumeracy (London: Penguin, 1988), 49.

(27) John Rawls, A Theory of Justice (Cambridge, MA: Belknap/ Harvard University Press, 1971).

(28) Rawls, A Theory of Justice, 242.

(29) Thomas Pogge, "Three Problems with Contractarian-Consequentialist Ways of Assessing Social Institutions," Social Philosophy and Policy 12 (1995): 241-66.

(30) Rawls, A Theory of Justice, p. 241.

(31) H.L.A. Hart, Punishment and Responsibility (Oxford: Oxford University Press, 1968), quoting from pp. 45, 47.

(32) Reiman & van den Haag, "On the Common Saying," 240.

(33) Alan Dershowitz, Reasonable Doubts: The O.J. Simpson Case and the Criminal Justice System (New York: Touchstone, 1997), 40.

(34) The counterfactual account is open to objections, as effectively pointed out by Frances Howard-Snyder in her entry in Stanford Encyclopedia of Philosophy (Summer 2002 edition), "Doing vs. Allowing Harm" <http://plato.stanford. edu/entries/doing-allowing/>. One of her examples is as follows:

Suppose a man is lying asleep on the ground. He is awoken by a crash and notices a large rock rolling down the hill towards him. He can easily move out of its way, but realizes that if he does so the rock will gain momentum and kill a group of small deaf children further down the hill. He tenses his muscles, fights his desire to run away and stands his ground. The rock hits and seriously injures him. But he stops it. And he saves the children. Donagan's [counterfactual] account [in The Theory of Morality, 1977], however, seems to imply that he merely allows the rock to stop, since, had he remained asleep, the rock would have struck and been stopped by his body.

In the counterfactual account, had the man moved out of the rock's way, it would have killed the children, and hence his action would have to be classified as an instance of doing and not allowing. But clearly there is more to instances of doings than mere counterfactuality: merely deciding to move out of a certain position is not a plausible candidate for positive agency. The counterfactual account needs to be supplemented by what Howard-Snyder calls "the Transfer of Energy Account": by pressing the child's head under the water in the lake, I bring about its death by physically acting on it; by moving out of the rock's way the man in the above scenario does not act on the children further down the hill at all.

(35) Reiman & van den Haag, "On the Common Saying," 229.

(36) Of course, the doctrine of doing and allowing needs a philosophical rationale that adequately explains our moral intuitions in individual cases. For a very interesting attempt to ground the doctrine in "our practice of treating one another as responsible agents," see Samuel Scheffler, "Doing and Allowing," Ethics 114 (January, 2004): 215-39.

Vidar Halvorsen is Associate Professor at the Institute of Criminology and Sociology of Law, University of Oslo, Norway.
COPYRIGHT 2004 Institute for Criminal Justice Ethics
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2004 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Halvorsen, Vidar
Publication:Criminal Justice Ethics
Date:Jun 22, 2004
Previous Article:An ethics consult with Kierkegaard.
Next Article:Idealizing the other? Western images of the Japanese criminal justice system.

Terms of use | Privacy policy | Copyright © 2022 Farlex, Inc. | Feedback | For webmasters |