Editor’s note: This article was written by Judge Jon O. Newman during his tenure as the Bolch Judicial Institute’s inaugural Distinguished Judge in Residence. The Institute’s Distinguished Judge in Residence is awarded to an outstanding jurist who has an interest in scholarship, law teaching, and the study of the judiciary. The recipient visits Duke Law, where he or she may give guest lectures, participate in faculty scholarship workshops, and advise students and faculty. The residency also offers the judge an opportunity to work on the judge’s own writing projects, with the assistance of a student researcher. Judge Newman thanks Bolch Judicial Institute Fellow Amelia Thorn and Duke Law student Colleen O’Leary for helpful editorial suggestions.
Trial courts frequently do not make the standard of proof beyond a reasonable doubt sufficiently clear to juries, and appellate courts sometimes do not sufficiently assure that the standard is being observed. Why is this so? What can trial and appellate judges do about it? I start by recalling how the standard developed and what its use is expected to accomplish before considering how the standard functions — and should function — in trial and appellate courts.
The reasonable doubt standard, originally and now, serves to minimize the chances that an innocent person will be convicted, but the reason for striving to avoid that outcome has been the subject of dispute. Although the standard now serves to make conviction difficult or at least to increase the accuracy of a determination of guilt, one scholar, Yale Law Professor James Q. Whitman, has written that its original purpose was to make conviction easier.1
The reason: jurors were fearful that if they found an innocent person guilty, they would be severely punished in the afterlife. According to Whitman, “The reasonable doubt formula was originally concerned with protecting the souls of the jurors against damnation.”2 Whitman further explains, “Convicting an innocent defendant was regarded, in the older Christian tradition, as a potential mortal sin. The reasonable doubt rule was one of many rules and procedures that developed in response to this disquieting possibility. It was originally a theological doctrine, intended to reassure jurors that they could convict the defendant without risking their own salvation, so long as their doubts about guilt were not ‘reasonable.’”3
Former Judge Richard A. Posner has challenged Whitman’s historical contention. Posner points out that the theological concern about convicting an innocent person — and thereby subjecting jurors and judges to damnation for error — though prevalent in the Middle Ages, was not a significant factor centuries later when the reasonable doubt standard came into use.4
Whitman acknowledges that it was not until the end of the 18th century that English judges began to instruct jurors not to convict a defendant if there was “reasonable doubt” about guilt.5 Whitman reports a 1782 trial in London’s Old Bailey where the judge asked the jury, “Have you any reasonable doubt[?],”6 and a 1783 trial where the judge told the jury, “If on viewing the evidence any reasonable doubt remains on your minds, . . . [the defendant] will be entitled to your acquittal.”7
By the time of a 1796 trial, Whitman reports, a jury charge included a now familiar phrasing, “The first point . . . for you to consider is, whether [a signature was genuine], or whether it is a forgery; which, if we should establish beyond any reasonable ground of doubt [would suffice], for you are not to expect mathematical demonstration in the proceedings of the administration of justice.”8
Defense counsel in the 1798 Irish Treason Cases9 urged adoption of the reasonable doubt standard because they believed it would effectively raise the prosecution’s burden of persuasion.10
On the other hand, some prosecutors in that era urged adoption of the same standard because they believed it would effectively lower their burden from having to persuade the jury of guilt beyond all doubt.11 Ultimately, it came to be recognized that proof “beyond a reasonable doubt” was necessary, not to spare the jurors the risk of damnation for convicting an innocent person, but to lessen the risk that an innocent person would be convicted.
That risk has sometimes been quantified in expressions of the idea that it is better to free some number of guilty persons than to convict one innocent person. The number has varied. Perhaps the lowest was expressed by defense counsel (later President) John Adams in the 1770 trial of soldiers involved in the Boston Massacre: “[I]t is better, five guilty persons should escape unpunished, than one innocent person should die.”12 A usual version of the ratio puts the number of guilty persons freed for one innocent person convicted at ten,13 although ratios of 20 to 114 and even 99 to 1 have been mentioned in earlier literature.15
No one suggests that such ratios should be included in a jury charge.
Although the reasonable doubt standard was well known to the Framers of the Constitution, neither the original document nor any provision of the Bill of Rights guaranteed that standard in criminal trials. The standard was regularly included in jury charges in both federal court and state court trials, but it was not until 1970 that the Supreme Court ruled that the Due Process Clauses of the Fifth and Fourteenth Amendments guaranteed a defendant the right have the standard included in a jury charge.16
In 1975, the Court made clear that not only the ultimate issue of guilt, but also each element of the offense charged, must be proved beyond a reasonable doubt.17
A. What Trial Judges Tell Juries
Numerous appellate courts have cautioned trial judges not to explain the reasonable doubt standard to juries.18 I find it astonishing that trial judges are instructed not to give juries any guidance on what the words “reasonable doubt” mean. Although it would be difficult to reach a consensus on the best, or even an acceptable, explanation, the risk of giving no explanation to a phrase of imprecise meaning is that different jurors will apply their own ideas, some of which might well give the accused far less protection than the standard is supposed to provide. This risk becomes particularly evident from consideration of some of the jury charge language various trial judges have used and the defects of such language.
Some judges tell a jury that a reasonable doubt is “a doubt based on reason.”19 This elaboration has three defects. First, it runs counter to the idea that a juror should be entitled to vote “not guilty” based only on a gut feeling, without an articulable rationale.20
Second, it can create ambiguity as to whether the juror has a doubt for which a reason can be thought of in the juror’s mind or a doubt that the juror can articulate to other jurors.21
Third, it might mislead a jury to look to the defendant for an explanation.22 The “based on reason” formulation has encountered some criticism, mostly in an earlier time.23 In Jackson v. Virginia,24 the Supreme Court said that “[a] reasonable doubt, at a minimum, is one based on reason.”25
Another common explanation is that the evidence must persuade the jurors of guilt “to a moral certainty.” Some federal courts have explicitly rejected the “moral certainty” standard, fearing that the word “certainty” would conflict with the concept of “reasonable doubt,”26 although some state courts endorse it.27
The Supreme Court has said that it does not “condone the use of the antiquated ‘moral certainty’ phrase” but has permitted it when amplified with supposedly clarifying language.28 Interestingly, the “moral certainty” standard was originally introduced at the urging of prosecutors to lessen their burden, because “moral certainty” was thought of as “reasonable certainty” — as opposed to the “absolute certainty” they feared jurors were thinking was required.
Still another explanation of proof beyond a reasonable doubt is the “hesitate to act” formulation. In Victor v. Nebraska,29 the Supreme Court approved the following version of this formulation: “‘[r]easonable doubt’ is such a doubt as would cause a reasonable and prudent person, in one of the graver and more important transactions of life, to pause and hesitate before taking the represented facts as true and relying and acting thereon.”30
In Holland v. United States,31 the Supreme Court noted that the trial judge had defined reasonable doubt “as ‘the kind of doubt . . . which you folks in the more serious and important affairs of your own lives might be willing to act upon’” and stated, “We think this section of the charge should have been in terms of the kind of doubt that would make a person hesitate to act, rather than the kind on which he would be willing to act.”32
The defect of “hesitate to act” language is that it is subject to different interpretations. From my own inquiries, I have learned that some people think it means that if they, as potential jurors, were to think that the evidence leaves them with a doubt comparable to the doubt that would cause them to hesitate before deciding some important matter, then they should vote “not guilty.” That understanding seems to be what the instruction literally requires them to do. Other people, however, have told me that they would reject this literal understanding because they almost always hesitate before making important decisions, and they do not think a judge would be telling them to find nearly every defendant “not guilty.” For these people, the instruction suggests caution: if they conclude that the evidence has created a doubt comparable to the doubt that would cause them to hesitate before making an important personal decision, they should take a careful look at all the evidence and vote to find the defendant guilty only if they are then quite sure that he is guilty. In other words, for one group, reaching the point of hesitation ends the process; for the other group, reaching that point permits the process to continue, but with caution. However juries understand a “hesitate to act” instruction, the fact that this formulation is ambiguous ought to cast doubt on its utility.33
Some of those thinking about the reasonable doubt standard have written about it as establishing a high “probability” that the defendant committed the charged crime. For example, Judge Jack Weinstein has written, “Were I the trier of fact, I would require a probability of guilt of no less than 95%.”34 Professor James Franklin has written, “[A]ny probability less than 0.8 should be declared less than proof beyond a reasonable doubt in all circumstances.”35
Professors Peter Tillers and Jonathan Gottfried, using a variant of probability language, favor an instruction that permits a juror to convict “only if the juror believes that there is more than a 95% chance that the defendant is guilty.”36
Professor Rita James Simon sent a questionnaire to 1,200 federal and state judges inquiring what numerical value they placed on the reasonable doubt standard.37 Her questions all asked for a number based on “probability”; a typical inquiry was worded: “Translate the phrase ‘beyond a reasonable doubt’ into a statement of probability.”38 A large number of judges responded and selected a number. For the quoted inquiry, the median number was 8.8, and the mean number was 8.9. However, the questionnaire never explained to the judges (or to the readers of the report) what was meant by the term “probability.”
Expressing the reasonable doubt standard in terms of probability is ill-advised. “Probability” is usually defined as “the chance that a given event will occur.”39 If thus defined, probability would have nothing to do with the likelihood that an accused defendant committed the crime for which he is being tried. Obviously, if he committed the crime, he did so in the past. “Probability” as a statistical concept is appropriately used when we say that if a coin is tossed in the air once, there is a 50 percent probability that it will come up heads.
Perhaps some who speak of a probability that a defendant is guilty are not using the term in a statistical sense. They might mean simply that, as a matter of ordinary speech, the defendant is probably guilty. But a standard of “probably guilty” would be an entirely unacceptable way of conveying to a jury the idea of guilt beyond a reasonable doubt because the vagueness of “probably” would permit a range of interpretations such as “more likely than not,” “quite likely,” and “almost definitely.”
There is a sense in which “probability” as a statistical concept could be applied to the “likelihood” that a defendant committed the crime charged. For example, when some say that “beyond a reasonable doubt” should be understood to mean that the jurors should not convict a defendant unless they conclude that there is at least a very high probability (for example, 95 percent) that he committed the crime, they might mean that if the same evidence was presented to 100 juries considering the same charge against 100 defendants and 95 of those defendants in fact committed the crime, then there is a 95 percent probability that the defendant on trial committed the crime charged.
But I doubt if the explanation just stated is what those advocating a high probability of guilt mean. Indeed, my guess is that there is some variety among such advocates as to what they mean by a “probability” that the defendant committed the crime charged. And that variety itself is reason enough that “probability” should not be a concept communicated to a jury for the obvious reason of the high risk that different jurors will think the term has a different meaning, and at least some of those meanings will be incorrect.
B. What Trial Judges Should Tell Juries
Despite my despair over the reluctance of appellate courts to permit trial judges to explain to juries what “reasonable doubt” means, I agree that any elaborate explanation of the phrase poses a risk of both perplexing juries and possibly lessening the rigor that the standard is supposed to require them to apply to their task. So, instead of attempting to define the words “reasonable doubt,” I believe a better approach to developing a helpful instruction would be to consider the purpose of the reasonable doubt standard.
That purpose is to make sure that the jurors do not convict a defendant unless they have a high degree of certainty that he is guilty. The Supreme Court pointed toward that purpose in Winship, the decision establishing the “reasonable doubt” standard as a requirement of due process of law. The Court stated, “[T]he reasonable-doubt standard is indispensable, for it impresses on the trier of fact the necessity of reaching a subjective state of certitude of the facts in issue.”40 The Court repeated the “certitude” language of Winship in Jackson v. Virginia,41 modifying the language to “near certitude.”42 As the Court explained in Jackson, “[B]y impressing upon the factfinder the need to reach a subjective state of near certitude of the guilt of the accused, the standard [of proof beyond a reasonable doubt] symbolizes the significance that our society attaches to the criminal sanction and thus to liberty itself.”3
Precisely. The standard is met when the jurors have reached “a subjective state of near certitude” concerning the defendant’s guilt.
In 1987, a subcommittee of the Committee on the Operation of the Jury System of the United States Judicial Conference proposed a model jury charge that included these words: “Proof beyond a reasonable doubt is proof that leaves you firmly convinced of the defendant’s guilt.”3 Justice Ruth Bader Ginsburg has endorsed this charge language, stating, “This Model instruction surpasses others I have seen in stating the reasonable doubt standard succinctly and comprehensibly.”43 I agree.
If certitude (or certainty) is thought of as a continuum, “beyond a reasonable doubt” means that the probative force of the evidence of guilt has reached a point very far along a continuum of certainty. If the continuum were to be expressed in numerical terms with the scale of certainty running from 0 to 100, the “near certainty” that Winship and Jackson require for proof beyond a reasonable doubt would probably be reached at least above 90, preferably at 95.
I think it unlikely that appellate courts today would approve a reasonable doubt instruction that uses a number. Nevertheless, writing for the future, I believe that the most effective way to tell jurors that the reasonable doubt standard means near certainty that the defendant is guilty is to give the following instruction:
Proof beyond a reasonable doubt requires evidence of such persuasive force that you are convinced of the defendant’s guilt to a very high degree of certainty. One way to think about that degree of certainty is that if certainty ranged from 0 to 100, proof beyond a reasonable doubt would be reached when your degree of certainty was at least 95.
Then the instruction should stop. No mention of “a doubt based on reason.” No mention of “moral certainty.” No mention of “hesitate to act.” No mention of “probability.” Just “a very high degree of certainty” with a numerical value to make the concept meaningful.
A. The Standard of Review
In 1960, ten years before Winship made the reasonable doubt standard a constitutional requirement for conviction in trial courts, the Supreme Court considered for the first time the appellate review issue of whether a conviction was obtained without due process of law because of insufficient evidence. In Thompson v. City of Louisville,44 the “Shuffling Sam” case (so named because a charge of loitering was brought against a man standing alone on the dance floor of a café shuffling his feet to the sound of music), the Court ruled that the conviction was unconstitutional because there was no evidence of guilt at all.45
Then in 1977, in Freeman v. Zahradnick,46 another state court prisoner sought Supreme Court review on the ground that the evidence was constitutionally deficient. Although the Court denied his petition for a writ of certiorari, Justice Potter Stewart wrote a dissenting opinion urging a grant of the petition for the specific purpose of considering a constitutional standard for reviewing claims that the reasonable doubt standard has not been met. “If . . . a federal court determines that no rational trier of fact could have found a defendant guilty beyond a reasonable doubt of the state offense with which he was charged, it is surely arguable that the court must hold, under Winship, that the convicted defendant was denied due process of law.”47
Just two years later, in Jackson v. Virginia, the Court heeded Justice Stewart’s plea and for the first time announced how appellate courts should determine whether evidence sufficed to meet the constitutional requirement of proof beyond a reasonable doubt.48 Writing for a bare five-member majority that included Justices William J. Brennan, Jr., Byron White, Harry Blackmun, and Thurgood Marshall, Justice Stewart phrased the constitutional reviewing standard in four different ways. First, he wrote, “[T]he critical inquiry . . . must be . . . to determine whether the record evidence could reasonably support a finding of guilt beyond a reasonable doubt.”49
Second, he wrote that the constitutional reviewing standard was met if “any rational trier of fact could have found the essential elements of the crime beyond a reasonable doubt.”50
Third, phrasing the constitutional reviewing standard negatively, he wrote that due process would be denied “if it is found that . . . no rational trier of fact could have found proof of guilt beyond a reasonable doubt.”51
Fourth, combining words from the “reasonably support” phrase in the first iteration with words from the “no rational trier” phase in the third iteration, he concluded, in affirming the denial of relief, “[W]e hold that a rational trier of fact could reasonably have found that the petitioner committed murder in the first degree under Virginia law.”52
I have no way of knowing whether the wording variations in these four expressions of a constitutional reviewing standard were deliberate. But I am firmly of the view that the second version is seriously deficient and should not be used. The phrase “any rational trier of fact” can be taken to mean that a verdict comports with due process as long as just one rational jury among a hypothetical distribution of several juries hearing the same evidence could have found guilt beyond a reasonable doubt. The Court’s totally gratuitous emphasis on the word “any” by placing it in italics increases the risk of such an interpretation. In addition, this variation lacks the requirement that the jury’s conclusion must be reasonable. Finally, although “irrational” is sometimes used interchangeably with “unreasonable,” “irrational” often conjures a more negative image; the third and fourth versions create the risk that a reviewing court will not reject a verdict unless the court can say that the jurors must have been irrational to render it, which may function as a higher standard.
Justice Stewart’s first version has it exactly right. A guilty verdict should stand only if a reviewing court concludes that the evidence sufficed to permit a jury reasonably to find guilt beyond a reasonable doubt. Justice Stewart’s fourth version, stating the holding of the Court, wisely included the adverb “reasonably,” but would have been better had it omitted the phrase “a rational trier of fact.”
The defect of the “any rational trier of fact” formulation becomes clear when one considers how reviewing courts determine whether evidence in a civil case is so overwhelmingly persuasive or so clearly deficient that the case is not even appropriate for a jury’s consideration. In that situation, where the standard of proof is preponderance of the evidence, reviewing courts instruct trial judges to take the case from the jury “[w]hen the evidence is such that without weighing the credibility of the witnesses there can be but one reasonable conclusion as to the verdict.”53
In the civil context, a verdict for a plaintiff is not affirmed just because “any rational trier of fact could have found” the elements of the plaintiff’s claim established by a preponderance of the evidence. Conviction in a criminal trial, of course, requires proof by the higher reasonable doubt standard, but the jury’s determination that this standard has been met should be upheld only if the reviewing court determines that the jury acted reasonably in applying that standard.
Justice Stewart’s promulgation of a constitutional standard for reviewing guilty verdicts provoked a sharp rebuke from Justice John Paul Stevens. In an opinion in which Chief Justice Warren E. Burger and then-Justice William Rehnquist joined, Justice Stevens concurred in the judgment upholding Jackson’s conviction, but his opinion was really a dissent from Justice Stewart’s opinion. Justice Stevens caustically referred to the Court’s standard for appellate review of claims that the evidence was insufficient to prove guilt beyond a reasonable doubt as “this new brainchild”54 and warned that it “threatens serious harm to the quality of our judicial system.”3
Justice Lewis F. Powell, Jr., did not participate.
B. What Standard of Review Appellate Courts Apply
Unfortunately, federal courts of appeals have repeatedly used Justice Stewart’s second version of a constitutional standard for reviewing claims that the reasonable doubt standard has not been met. As the following table indicates,55 by the end of 2018, the “any rational trier of fact” formulation had been applied in the overwhelming majority of criminal appeals challenging the constitutional sufficiency of the evidence and the “could reasonably support a finding of guilt” formulation had rarely been applied.
| Circuit | “any rational trier of fact” | “could reasonably support a finding of guilt” |
|---|---|---|
| First | 96 | 0 |
| Second | 833 | 23 |
| Third | 491 | 5 |
| Fourth | 735 | 4 |
| Fifth | 564 | 9 |
| Sixth | 2,207 | 22 |
| Seventh | 797 | 7 |
| Eighth | 169 | 5 |
| Ninth | 2,342 | 13 |
| Tenth | 512 | 2 |
| Eleventh | 201 | 0 |
| D.C. | 133 | 2 |
| Total | 9,080 | 92 |
C. What Standard of Review Appellate Courts Should Apply
From the foregoing discussion, it should be clear that I favor having appellate courts reviewing criminal convictions “determine whether the record evidence could reasonably support a finding of guilt beyond a reasonable doubt.”56 They should avoid the “any rational trier” formulation, which the Supreme Court itself did not invoke when it stated its holding in its own review of a criminal conviction in Jackson.
I recognize that whatever formulation is used will rarely result in reversing a conviction for lack of sufficient evidence. That is entirely appropriate. Most federal trial judges can be expected to reject a verdict of guilty only where a jury could not reasonably conclude that guilt has been established beyond a reasonable doubt. And the three-tiered reconsideration of state court convictions by federal courts on habeas corpus review will also almost always suffice to reject an insufficiently supported verdict of guilt.
But on rare occasion, a guilty verdict not supported by constitutionally sufficient evidence might be upheld on direct review if a reviewing court applies the “any rational trier” standard formulation. United States v. Clark 57 provides an example. In that case, the evidence established that a small quantity of crack cocaine was found in a police cruiser, wedged into the space between the back of the rear-seat cushion and the bottom of the rear-seat back rest, moments after a defendant, handcuffed with his hands behind his back, had been placed in the back of the cruiser. No traces of cocaine were found on the defendant’s hands or clothing. There were only three possibilities: the handcuffed defendant had somehow managed to secrete the cocaine without leaving a trace on him; the cocaine had been placed where it was found by a previous occupant of the cruiser; or someone else (the police?) had placed the cocaine there. A jury accepted the first possibility and found the defendant guilty of narcotics possession.
Reversing the conviction, I wrote for a divided panel, “We cannot say it is an absolute impossibility for a person with his hands securely handcuffed behind his back to extract a substantial quantity of crack cocaine from his person or clothing and wedge it into the space where it was found without leaving a trace of cocaine on his fingers or clothing, but we can say that the possibility of such an occurrence is so exceedingly remote that no jury could reasonably find beyond a reasonable doubt that it happened.”58
I cannot be certain I would have reached the same result if I had asked whether “any rational trier of fact could have found the essential elements of the crime beyond a reasonable doubt.”59
When a case presents a serious issue as to whether the evidence sufficed to permit a jury to have faithfully observed the reasonable doubt standard in finding a defendant guilty, a reviewing court will be more likely to enforce that standard when it asks whether the jury could reasonably have found guilt beyond a reasonable doubt.
A jury charge quantifying the degree of certainty necessary to find guilt “beyond a reasonable doubt” and conscientious appellate review to assure that a jury reasonably found guilt “beyond a reasonable doubt” can help assure that this venerated phrase is taken seriously.
